ISSN: 2252-8814
IJAAS
International Journal of
Advances in Applied Sciences
Advances in Applied Sciences (IJAAS) is a peer-reviewed and open access journal dedicated to publish significant research findings in the field of applied and theoretical sciences. The journal is designed to serve researchers, developers, professionals, graduate students and others interested in state-of-the art research activities in applied science areas, which cover topics including: chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences.
Editor-in-Chief: Qing Wang, National Institute of Advanced Industrial Science and Technology (AIST), Japan Co-Editor-in-Chief: Chen-Yuan Chen, National Pingtung University of Education, Taiwan, Province of China Bensafi Abd-El-Hamid, Abou Bekr Belkaid University of Tlemcen, Algeria Guangming Yao, Clarkson University, United States Habibolla Latifizadeh, Shiraz (SUTECH) University, Iran, Islamic Republic of EL Mahdi Ahmed Haroun, University of Bahri, Sudan
Published by:
Institute of Advanced Engineering and Science (IAES) Website: http://iaescore.com/journals/index.php/IJAAS/ Email: info@iaesjournal.com, IJAAS@iaesjournal.com
Information for Authors International Journal of Advances in Applied Sciences (IJAAS) is an interdisciplinary journal that publishes material on all aspects of applied and theoretical sciences. The journal encompasses a variety of topics, including chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences. Submission of a manuscript implies that it contains original work and has not been published or submitted for publication elsewhere. It also implies the transfer of the copyright from the author to the publisher. Authors should include permission to reproduce any previously published material.
Ethics in publishing For information on Ethics in publishing and Ethical guidelines for journal publication (including the necessity to avoid plagiarism and duplicate publication) see http://www.iaescore.com/journals/index.php/IJAAS/about/editorialPolicies#sectionPolicies
Paper Submission You must prepare and submit your papers as word document (DOC or DOCX). For more detailed instructions and IJAAS template please take a look and download at: http://www.iaescore.com/journals/index.php/IJAAS/about/submissions#onlineSubmissions The manuscript will be subjected to a full review procedure and the decision whether to accept it will be taken by the Editor based on the reviews. Manuscript must be submitted through our on-line system: http://www.iaescore.com/journals/index.php/IJAAS/ Once a manuscript has successfully been submitted via the online submission system authors may track the status of their manuscript using the online submission system.
IJAAS
International Journal of
Advances in Applied Sciences
High impedance fault detection in distribution system Kavaskar Sekar, Nalin Kant Mohanty
95-102
Frequency regulation of modern power system using novel hybrid DE-DA algorithm Sayantan Sinha, Ranjan Kumar Mallick
103-116
Enhanced performance of PID load frequency controller for power systems Dola Gobinda Padhan, Suresh Kumar Tummala
117-124
Solar irradiance forecasting using fuzzy logic and multilinear regression approach: a case study of Punjab, India Sahil Mehta, Prasenjit Basak
125-135
Solar panel monitoring and energy prediction for smart solar system Isha M. Shirbhate, Sunita S. Barve
136-142
Performance evaluation and comparison of diode clamped multilevel inverter and hybrid inverter based on PD and APOD modulation techniques N. Susheela, P. Satish Kumar
143-153
Solar energy storage and release application of water-phase change material- (SnO2-TaC) and (SnO2–SiC) nanoparticles system Farhan Lafta Rashid, Aseel Hadi, Ammar Ali Abid, Ahmed Hashim
154-156
A model free dissolved oxygen controller for industry effluent using hybrid variables measuring technique P. Kingston Stanley, Sanjeevi Gandhi A., D. Abraham Chandy
157-163
Subsynchronous resonance oscillations mitigation via fuzzy controlled novel braking resistor model Mohamed Fayez, M. Mandour, M. El-Hadidy, F. Bendary
164-170
Responsibility of the contents rests upon the authors and not upon the publisher or editors.
IJAAS
Vol. 8
No. 2
pp. 95-170
June 2019
ISSN 2252-8814
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 95~102 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp95-102
95
High impedance fault detection in distribution system Kavaskar Sekar1, Nalin Kant Mohanty2 1Department
of Electrical and Electronics Engineering, Panimalar Engineering College, India 2 Department of Electrical and Electronics Engineering, Sri Venkateswara College of Engineering, India
Article Info
ABSTRACT
Article history:
High impedance faults (HIFs) present a huge complexity of identification in an electric power distribution network (EPDN) due to their characteristics. Further, the growth of non-linear load adds complexity in HIF detection. One primary challenge of power system engineers is to reliably detect and discriminate HIFs from normal distribution system load and other switching transient disturbances. In this study, a novel HIF detection method is proposed based on the simulation of an accurate model of an actual EPDN study with real data. The proposed method uses current signal alone and does not require voltage signal. Wavelet transform (WT) is used for signal decomposition to extract statistical features and classification of HIF into Non-HIF (NHIF) by Neural Networks (NNs). The simulation study of the proposed method provides good, consistent and powerful protection for HIF.
Received Aug 30, 2018 Revised Mar 20, 2019 Accepted Apr 17, 2019 Keywords: Discrete wavelet transform High impedance fault Neural networks Non-linear load
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: Kavaskar Sekar, Department of Electrical and Electronics Engineering, Anna University, Guindy, Chennai, Tamil Nadu 600025, India. Email: kavaskarsekar@gmail.com
1.
INTRODUCTION The majority of electric power distribution lines in India are overhead lines. These lines are subject to disturbance such as an energized conductor falls to the ground or connected to high impedance objects due to dissimilar environments. This condition is called High impedance fault (HIF), and it draws little fault current which is less compared to the threshold values of conventional protection means. Hence, some of these HIFs may not be detected. Further, threshold values cannot set as lower values, since the normal load distribution will lead to nuisance tripping. Besides, HIFs are accompanied by electric arcs. These arcs produce an asymmetric, unpredictable, and random current signal. Most of the electric power distribution networks (EPDN) are in close proximity to the thickly populated area. If EPDN is operated with HIF leads to endangering the human lives and their properties or material goods. Therefore, with the goal of increasing the performance and the safety of EPDN, several papers have been published in the past few decades to solve HIF problem is well documented in [1]. The redefining process of detection yet to be completed due to an increase in non-linear loads (NLLs). A method based on harmonic content presented in [2, 3]. The setback of such method is to set threshold values which reduces the performance of the method. Time-frequency analysis [4, 5] produces better performance in HIF detection. The modified Fast Fourier Transform (FFT) method employs the relative relation between the third, fifth and seventh harmonic current in [6]. The presence of NLLs taken into account and reliable detection shown in the system studied. However, the percentage of false tripping and computational work is an obstacle for practical implementations. Time-domain based method called mathematical morphology [7, 8] is used for detection of HIF, which is good for a balanced system. When the system is unbalanced, these methods show less performance. Wavelet Transform (WT) has been used in HIF detection. WT is analyzing the signal with frequency component and their position in time. More than 20 years such methods used in protection Journal homepage: http://iaescore.com/online/index.php/IJAAS
96
ISSN: 2252-8814
applications. The algorithm in [9] uses dynamic features extracted by stationary WT and a support vector machine based decision-making system. Arcing currents related to different high impedance surface are generated in the laboratory setup, and these signals are decomposed using Discrete Wavelet Transform (DWT) [10]. The authors used DWT to examine high- and low-frequency voltage components at different points [11] and an average of absolute difference of extracted voltage signal [12] of EPDN. In [13] voltage and current signals are used to detect and locate HIF respectively. This method uses the absolute sum of high-frequency components for both detection and location. Two different HIF detection scheme are discussed in [14], both utilizing WT decomposition. In the first method, principal component analysis (PCA) for feature reduction and Neural Network (NN) for feature classification. In the second method, a Genetic Algorithm (GA) for feature reduction and Bayes clasifier for classification. The change in current signal created by HIF and other transient events has been used in [15] with NLLs in the system. A DWT used to decompose the current signals and extracts features to train NNs. An evolving NN [16] and continually online trained NN method [17] was shown as the proper approach to HIF detection since it is a time-varying problem. A DWT with data mining based classification in [18]. These research articles disclose useful properties and different detection methods of HIF. However, the majority of the above-mentioned articles fail to consider NLLs except in [6, 15, 18]. These three articles are showing less variation in NLLs. Since NLLs such as television, computer, fluorescent lamps, etc., constantly increasing year by year in the distribution grids and NLLs and HIFs characteristics have closely resembled each other, which will make the existing methods less effective. Hence, an enhanced detection method for HIFs with a massive variation of NLLs is Proposed. The rest of the paper has been prepared as follows. Section 2 explains the test system and also the characteristics of HIF model engaged for simulation. The proposed methodology has been thoroughly discussed in Section 3, including the decomposition of a signal, feature extraction, feature selection, and classification. Section 4 has results and discussion and the paper concluded with the main highlights of the work in section 5. 2. TEST SYSTEM 2.1. System studied The proposed method is verified through an actual EPDN in India (Chennai) shown in Figure 1 was modeled with a sim-power-system block set (MATLAB). The system data are given in [18]. Distribution lines are modeled as lumped parameters. Busbar input is modeled by Thevenin's voltage and equivalent impedance. 2.2. HIF model and characteristics The HIF Emanuel’s model [4] shown in Figure 2 used in this work. The HIF current signal during different parameters of the model is shown in Figure 3, which shows the non-linear behaviour, asymmetry in the signal, fewer current and random behavior of fault current. This is due to low- and high-frequency content of the signal is changing with time and hence the signal is termed as a non-stationary signal. 2.3. Test conditions For the classification HIF and non-HIF (NHIF), a total of 540 HIF and 460 NHIF cases are described in Table 1. 3.
PROPOSED METHODOLOGY The proposed HIF detection approach is shown in Figure 4. The process starts with the simulation of EPDN and accessing current signal. The DWT decomposes the current signal and features such as energy, entropy, skewness, sum, standard deviation, and kurtosis are extracted. These features are training two layer feed forward NNs to classify the case belongs to HIF or NHIF. 3.1. Signal decomposition using wavelet transform Wavelet is an effective tool to analyze non-stationary signal with the capability of multiple resolutions in time and frequency. DWT is a flexible signal processing method broadly used in power system engineering. DWT has been successful in stability analysis of power system engineering. The mathematical equation for DWT represented in (1). 𝐷𝑊𝑇(
, )
=
√
∑x(k) ψ(
)
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 95 – 102
(1)
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
97
where x(k) is the discrete signal in terms of coefficients. ψ(.) is mother wavelet, m and n are time scale parameters. k is discrete time of coefficients. 2m is scaling parameter. k 2m is shifting parameter. is the energy normalization component. √
Figure 1. EPDN studied
Figure 2. HIF model
(a)
(b) Figure 3. HIF current waveform (a) R1=10000 Ω, R2=9000 Ω,V1=5000 V,V2= 5050 V. (b) R1= 150 Ω, R2= 160 Ω, V1=2000 V, V2=2050 V. Low pass filters (LPF) and high pass filters (HPF) is used in DWT to decompose a signal into a lowfrequency and high-frequency component. The decomposition process starts by passing a signal through LPF and HPF. The LPF produces an approximate coefficient which has high scaled and low-frequency decomposition. In contrast, the HPF produces detailed coefficient which has low scaled and high-frequency High impedance fault detection in distribution system (Kavaskar Sekar)
98
ISSN: 2252-8814
decomposition. At once the first level of decomposition is completed, the sampling frequency is reduced by half of its value. To calculate the next DWT level, LPF output (approximation) is decomposed. The Daubechies basis db4, five level decomposition, and the sampling rate of 25.6 kHz with the data window length of three cycles are chosen in this work. Table 1. Test conditions Event HIF Load switching NLL switching Capacitor Switching Inrush current
Simulated conditions Resistances varied from 150 Ω to 12 kΩ and DC voltage from 1.5 kV to 10.5 kV randomly Changes in load level: 0% - 25%, 25% - 50%, 50%-75%,75%-100% and 100% -110% both forward and reverse conditions. The ratio of linear to non-linear load changes 1 to 0.5 in steps of 0.05 both in forward and reverse conditions. Switching conditions on and off with load variation of 25%, 50%, 75 %, 100% and 125%. No load switching of 33 /11 kV transformer
Total 540 90 216 90 64
Figure 4. The proposed HIF detection approach The decomposition of phase a current due to NLLs and capacitance switching at inception time of 0.065 s is shown in Figure 5 and Figure 6. It is found that the behavior of both NLL and capacitance has low and high-frequency parts in all wavelet detail coefficient D5 to D1. The decomposition of phase a current during HIF is shown in Figure 7. A small magnitude of change in frequency component is present since the initiation of HIF in detail coefficient D5 alone. 3.2. Selection of wavelet coefficient From the DWT analysis of the current signal, there were one approximation coefficient (A5) and five detailed coefficients (d1-d5) acquired. The comparison of different wavelet detail and approximation coefficient has been made to select wavelet coefficient for HIF Detection. It is observed that detail coefficients d4 and d5 offers diverse characteristics during different power system events and hence selected for feature extraction. 3.3. Features using wavelet coefficients d4 and d5 Six different features were estimated using wavelet coefficients d4 and d5 are described below: a. Standard deviation: It represents the deviation of a signal from its mean. b. Energy: The total energy content of the current signal. (2)
Energy = ∫ i (t) dt c.
Kurtosis: Bigger kurtosis point represents more outlier in the signal. Kurtosis =
(
(
)
)(
)(
)
∑
(
)
−(
(
) )(
)
- Standard deviation, 𝑥̅ – Mean, n – no of sample data d.
Skewness: It is a measure of the irregularity of the probability distribution about its mean.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 95 – 102
(3)
Int. J. of Adv. in Appl. Sci. Skewness =
(
ISSN: 2252-8814
)(
)
∑
(
)
99
(4)
Figure 5. DWT of current signal under non-linear load switching
Figure 6. DWT of current signal under capacitance switching
Figure 7. DWT of current signal under HIF High impedance fault detection in distribution system (Kavaskar Sekar)
100 e.
ISSN: 2252-8814
Entropy: It gives complexity of the signal. (5)
Entropy = − ∑ p log p Where pj is an energy probability distribution of WT detail coefficient (d4 and d5) f.
Sum: Sum of all points of signal.
3.4. NN implementation The NN is a powerful tool used in pattern recognition and classification [19]. Consequently, the proposed work uses NN to learn the input and output relationship from feature input vector. As mentioned in the feature extraction process, the input to NN comprises 12 elements (6 features of detail coefficients d4 and d5). The output layer has two neurons i.e. either HIF or NHIF and 10 hidden layers. A two-layer feedforward NN has been used in this work as shown in Figure 8. The sigmoid activation function used for hidden layer and softmax for the output layer. The NN is trained with scaled conjugate back-propagation that updates weight and bias values. Also, the number of epochs was determined by experimentation. The number of epochs for the training was set around 1000 and a mean squared error rate of e -03 was used [20]. The dataset is randomly divided into three parts: a training data set of 70%, validation of 15% dataset, and the remaining 15% for testing.
Figure 8. NN model for HIF detection 4.
RESULTS AND DISCUSSIONS A set of 1000 features obtained by changing the simulation conditions as given in Table 1. Thus a wide range of studies has been carried out to demonstrate the proposed method. Conversely, it is not possible to report all information because of space constraint. The proposed method is evaluated through the following three parameters: - Dependability: Predicted HIF against total HIF conditions. - Security: Predicted Non-HIF against total non-HIF conditions. - Accuracy: Actual predicted against the total number of conditions considered. - Speed: One cycle power frequency current / No of cycles that it takes to detect the fault. The confusion matrix of the NN during training, validation, and testing are presented in Figure 9. During training, all 370 HIF test cases are correctly classified to produce dependability of 100%. Two NHIF test case is misclassified as HIF to produce 99.4% security with the overall accuracy of 99.7%. During validation of 150 cases, dependability observed as 100%, and 97% of security with the overall accuracy of 98.7%. Finally, during testing of 150 cases, all HIF cases predicted correctly and only one NHIF cases are classified as HIF and produce dependability, security, and accuracy as 100%, 98.4%, and 99.3% respectively. It is very clear from the confusion matrix that the method has an excellent detection rate. Nevertheless, there was a huge amount of non-linear load variation, all the HIF cases are detected. Power system more vulnerable to noise, it is necessary to investigate the proposed method with different Signal to Noise Ratio (SNR). Table 2 compares the performance under noisy atmosphere. It noted that the performance is similar to the one observed in normal case for 10 dB of SNR on the signal. For 20 dB of SNR, HIF detection rate alone same. However, the accuracy and security are reduced to 98% and 95.16% respectively. Additional SNR reduces all the three performance indices. Though the accuracy of 20 dB is less, the HIF detection rate is excellent, and hence the proposed method is suitable for SNR of 20 or less.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 95 – 102
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
101
Figure 9. The confusion matrix of NNs Table 3 shows the impact of the sampling rate of the proposed work. There was substantial upgrading demonstrated by the proposed methodology with sampling rate from 256 to 512. The detection rate diminishes as the sampling rate value decreases. Hence, the proposed scheme is better with 512 samples per cycle. Finally, the performance of the proposed HIF approach is compared with previously published articles as presented in Table 4. It is observed that all three performance indices are better than [4, 7, 10, 14]. The speed of detection is compared with the references mentioned in Table 4. The speed of detection is low compared to [4] and as good as with other references listed in Table 4. However, it is noted that HIFs are low current, during which the distribution network are not in stress. As a result, the HIF detection speed is not a critical concern. In other hand, the security index of the HIF detection algorithm is important as well as its dependability index. Table 2. Impact of SNR on the proposed method Performance indices Accuracy Security Dependability
10 dB 99.33% 98.38% 100%
20 dB 98.00% 95.16% 100%
30 dB 93.33% 88.70% 96.59%
Table 3. Impact of sampling rate on the proposed method Performance indices Accuracy Security Dependability
512 99.33% 98.38% 100%
420 96.00% 95.16% 98.88%
256 86.66% 80.64% 90.90%
Table 4. Comparison of the proposed method with previously published work Method Ref [4] Ref [7] Ref [14] Ref [10] The proposed method
Accuracy (%) 93.6 97.3 96 99.33
Security (%) 81.5 96.3 100 68 98.38
Dependability (%) 100 98.3 90 72 100
Speed 1 0.04 0.25 0.33 0.33
High impedance fault detection in distribution system (Kavaskar Sekar)
102
ISSN: 2252-8814
5.
CONCLUSION In this paper, a new methodology based on the WT NNs is presented. The proposed method efficiently distinguishing HIF form other normal patterns. The proposed method comprehensively studied an actual electric power system with real data using MATLAB with a huge variation in power system operating conditions. The method has an exceptional detection rate even under wide variation in NLLs. This indicates that the proposed scheme is highly consistent and secure detection for HIFs with a dependability index of 100%. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]
Ghaderi, H., Mohammadpour, HA, Ginn, HL, and Shin, YJ, “High Impedance Fault detection -a review,” Electr. Power Syst. Res, vol. 146, pp. 376-388, 2017. Emanuel, D., Cyganski, J., Orr S., Shiller, and E. Gulachenski, “High impedance fault arcing on sandy soil in 15 kV distribution feeders: contributions to the evaluation of the low frequency spectrum,” IEEE Trans. Power Deliv, vol. 5(2), pp. 676-686, 1990. Esmaeil, D. and Jamal, B., “High Impedance Fault Detection in Power Distribution Networks with Use of Current Harmonic-Based Algorithm,” Indonesian Journal of Electrical Engineering and Informatics, vol. 3(4), pp. 216-223, 2015. Ghaderi, H., Mohammadpour, HA, Ginn, HL, and Shin, YJ, “High impedance fault detection in distribution network using time-frequency based algorithm,” IEEE Trans. Power Deliv, vol 30(3), pp. 1260-1268, 2015. Samantaray, SR., Panigrahi, BK., and Dash, PK., “High-impedance fault detection in power distribution networks using time-frequency transform and probabilistic neural network,” IET Gener. Transm. Distrib, vol. 2(2), pp. 261–270, 2008. Soheili, A., Sadeh, J., and Bakhshi, R., “Modified FFT-based high-impedance fault detection technique considering distribution non-linear loads: simulation and experimental study,” International Journal of Electrical Power and Energy Systems, vol. 94, pp. 124–140, 2018. Sarlak, M. and Shahrtash, SM., “High-impedance fault detection using combination of multilayer perceptron neural networks based on multi resolution morphological gradient features of current waveform,” IET Gener. Transm. Distrib, vol 5(5), pp. 588-595, 2011. Kavaskar, S. and Mohanty NK., “Combined mathematical morphology and data mining based high impedance fault detection,” Energy Procedia, vol. 117, pp. 417-423, 2017. Hamid Mortazavi, S., Moravej, Z., and Mohammad, SS., “A hybrid method for arcing faults detection in large distribution networks,” International Journal of Electrical Power and Energy Systems, vol. 94, pp. 141–150, 2018. Jichao, C., Toan, P., Trevor, B., Eliathambi, A., and Daming, Z., “Detection high impedance fault using current transformer for sensing and identification based method on features extracted using wavelet transform,” IET Gener. Transm. Distrib, vol 10(12), pp. 2290-2298, 2016. Santos, WC, Lopes, FV., Brito, NSD., and Souza, BA., “High-impedance fault identification on distribution networks” IEEE Trans. Power Del, pp. 32(1), pp. 23-32, 2017. Bakar, AHA., Ali, M., Tan, C., Mokhlis, H., Arof, H., and Illias, H., “High-impedance fault location in 11kV underground distribution systems using wavelet transforms” International Journal of Electrical Power and Energy Systems, vol. 55, pp. 723-730, 2014. Mahari, H. and Seyed, “High impedance fault protection in transmission lines using a WPT-based algorithm,” International Journal of Electrical Power and Energy Systems, vol. 67, pp. 537-545, 2015. Sedighi, AR., Haghifam, MR., and Malik, OP., “Soft computing application in high impedance fault detection in distribution system,” Electr. Power Syst. Res, vol. 76(1-3), pp. 136-144, 2005. Baqui, I, Zamora, I., Mazón, J., and Buigues, G., “High impedance fault detection methodology using wavelet transform and artificial neural networks,” Electr. Power Syst, Res, vol. 81(7), 1325-1333, 2011. Sergio, S., Pyramo, C., Maury, G., Alcyr, L., Franciele, A., and Daniel, L., “High impedance fault detection in power distribution systems using wavelet transform and evolving neural network,” Electric Power Systems Research, vol. 154, pp. 474-483. 2018. Patrick, EF, Adriano, PM., Jean, PR., and Ghendy, C. “Non-linear high impedance fault distance estimation in power distribution systems: A continually online-trained neural network approach,” Electric Power Systems Research, vol. 157, pp. 20-28, 2018. Kavaskar, S., Mohanty, NK, and Sahoo, AK. “High-impedance fault detection using wavelet transform,” IEEE conference on 2018 Technologies for Smart-City Energy Security and Power, pp. 1-6, Bhubaneswar, 2018. Bishop, CM, Neural Networks for Pattern Recognition. Oxford University Press, 1996. Adewole, AC., Tzoneva, R., Shaheen, B., “Distribution network fault section identification and fault location using wavelet entropy and neural networks,” Applied Soft Computing, vol 46, pp. 296–306, 2016.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 95 – 102
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 103~116 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp103-116
103
Frequency regulation of modern power system using novel hybrid DE-DA algorithm Sayantan Sinha1, Ranjan Kumar Mallick2 1
Department of Electrical Engineering, Siksha ‘O’ Anusandhan Deemed to be University, India 2 Department of Elelctrical and Electronics Engineering, Siksha ‘O’ Anusandhan Deemed to be University, India
Article Info
ABSTRACT
Article history:
An attempt has been made to regulate the frequency of an interconnected modern power system using automatic generation control under a restructured market scenario. The system model considered consists of a thermal generation plant coupled with a gas turbine plant in both areas. The presence of deregulated market scenario in an interconnected power system makes it too vulnerable to small load disturbance giving rise to frequency and tie line power imbalances. An attempt has been made to introduce a novel Tilted Integral derivative controller to minimize the frequency and tie line power deviations and restrict them to scheduled values. A maiden attempt has been made to tune the controller gains with the help of a novel hybrid optimization scheme which includes the amalgamation of the exploitative nature of the Differential evolution technique and the explorative attributes of the Dragonfly Algorithm. This hybrid technique is therefore coined as Differential evolution- dragonfly algorithm (DE-DA) technique. Use of some standard benchmark fucntions are made to prove the efficacy of the proposed scheme in tunig the controller gains. The supremacy of the proposed TID controller is examined under two individual market scenarios and under the effect of a step load disturbance. The robustness of the controller in minimizing frequency deviations in the systems is broadly showcased. The superiority of the controller is also proved by comparing it with pre published results.
Received Aug 30, 2018 Revised Mar 20, 2019 Accepted Apr 20, 2019 Keywords: AGC Benchmark functions DE DA hybrid Deregulated TID
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: Sayantan Sinha, Department of Electrical Engineering, Siksha ‘O’ Anusandhan Deemed to be University, Khandagiri Marg, 2, Sum Hospital Rd, Bhubaneswar, Odisha 751030, India. Email: sayantansinha51@gmail.com
1.
INTRODUCTION In recent days, the power system is in astate of transition from centralised control to restructured market scenario. The restructured market consists of GENCOs (Generation companies) DISCOs (Distribution companies), TRANSCOs (Transmission companies) and a control operator named as Independent Service Operator (ISO). The coordinated control of a power system of such a huge scale is tedious and much more complicated. The prime facie objective of AGC in an interconnecped power system is to ensure that the deviations in frequency and the power flow in the tie-lines are restricted within nominal values. In a deregulated market scenario, the ISO plays an ancillary role in ensuring the stable operation of the power system [1, 2]. The main role of Automatic Generation control is to maintain an equity between the load demands and the generation of each area. It also pays a keen attention to the fact that the frequency deviation and the tie line power variations should stay within specified limits [3]. In the recent years many researches have been made based on the performance of automatic generation control in the power system world like analysis of the power flow taking AGC into consideration in multi-area interconnected power grid taking into consideration the deregulated market scenario [4]. Journal homepage: http://iaescore.com/online/index.php/IJAAS
104
ISSN: 2252-8814
Paper [5] deals with the Utilisation of ultra-capacitor in load frequency control under restructured STPPthermal power systems using WOA optimised PIDN-FOPD controller. Paper [6] concentrates on automatic generation control of two-area hydro-thermal system considering governor dead band under deregulated environment. In paper [7], a work has been done on an algorithm technique known as symbiotic organisms search for AGC in the interlinked power environment which includes wind farms. A recent paper [8] on AGC describes multiple unit of multiple area of deregulated power network by considering an algorithm method known as novel-quasi oppositional harmony search. A research has been made on the judgment of the influence of unreliability on AGC systems [9]. A method of modeling technique and a survey has been made on system stability of AGC on radio systems in smart grids [10]. Paper [11] describes the safety games for threat minimization in automatic generation control. In previous years, a work has been done on a model based strike diagnosis and reduction for automatic generation control [12]. From the study of literature, it was found that the activity of the whole power system network depends on various optimization techniques, structure of the controller. So, the innovations of various optimization techniques are always a welcoming step for the role enhancement of automatic generation control of the entire power system. Many works have been done on various algorithm techniques and methods such as a flower pollination technique based AGC of interlinked power environment and cross characterized ‘gbest’-mentored gravitational search and marking search algorithm for AGC of multiple area power system [13, 14]. Several strategies have been made for control for automatic generation control in multi terminal DC grids [15] and other strategies such as in [16], divided model predictive control strategies with demand to power system automatic generation control and in paper [17], distributed automatic generation control using horizontal-based method for large perforation in generation of wind. In [18], an approach has been made to AGC with different nonlinearities by using two degrees of PID controller. Paper [19] brings to light the implementation of an ANFIS based controller for the Load frequency studies in deregulated market. Load and frequency control for an interconnected power system using fractional order controllers were effectively discussed in [20]. In this work an attempt has been made to develop a novel controller with the gains tuned by a hybrid optimization technique. The main objectives of the proposed work are: design of a novel tilted integral derivative controller for the AGC of a multi-source power system under deregulated environment as proposed in [21]; optimization of the gains of the proposed controller with a hybrid DE-DA technique and comparison with the previously published results [21]; testing of the hybrid technique in some standard benchmark functions and establishing the superiority of the optimization scheme; analysis of the system dynamic parameters and comparison with the pre published results [21] and to establish the superiority of the proposed DE-DA optimized TID controller over DE optimized PID controller. 2.
SYSTEM CONSIDERED The system proposed in this work is a two area system having equal power ratings 2000 MW each. The linearized transfer function model of the system is clearly depicted in Figure 1. Each area is inclusive of a thermal unit and a gas generating unit. The thermal unit is coupled with a reheat turbine in order to increase its efficiency. The system parameters are listed down in Appendix A. Due to the presence of more than one GENCO in the system, the apf (ACE participation factor) is to be effectively chosen. APF stands for those coefficients that distribute the ACE (Area control error) among the GENCOs. This paper incorporates the concept of a mutual contract that exists between the Distribution companies (DISCOs) and the generation companies (GENCOs) of an interconnected power system. This mutual contract can exist in various combinations and is well explained mathematically by the Distribution Participation Matrix (DPM). In a DPM, the GENCOs are denoted by the rows and the DISCOs are labeled in the columns. Every individual entry in the DPM gives us a picture of the fraction of the load that is in the contract between the DISOC and the corresponding GENCO. The actual tie line power flow can be mathematically expressed as Ptieactual
2T12 f1 f 2 s
(1)
The tie line power flow in a deregulated environment can be explained as: ∆𝑃
= ∑
∆𝑃
= 𝑃
∑
𝑐𝑝𝑓 ∆𝑃 − ∑
∑
𝑐𝑝𝑓 ∆𝑃
− 𝑃
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 103 – 116
(2) (3)
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
105
where Pexp and Pimp stands for the net power expected to flow from area 1 according to the DISCO demands and the power actually supplied to the DISCOs respectively. This gives rise to a deviation in tie line power which is mathematically expressed as: ∆𝑃
= ∆𝑃
(4)
− ∆𝑃
The area control error in deregulated scheme can me expressed as: 𝐴𝐶𝐸 = 𝛽 ∆𝑓 + ∆𝑃
(5)
𝐴𝐶𝐸 = 𝛽 ∆𝑓 + 𝛼 ∆𝑃
(6)
1 R 1
1 sT cr 1 sT F
1 sX G 1 sY G
1 c g sb g
1 1 sTcd
1 1s.Tg1
1 s.Kr1.Tr1 1 s.Tr1
1 1 s.Tt1
Kp1 1s.Tp1
Δf1 2* pi*T12 s
a12
a12 1 1s.Tg2
1 c g sb g
1s.Kr2.Tr2 1s.Tr2
1 1 s.Tt 2
1 sX G 1 sY G
1 sT cr 1 sT F
1 1 sTcd
Kp1 1s.Tp1
Δf 2
1 R2
B2
Figure 1. Linearised transfer function model of the system [18] Where the area coefficient is labeled by α12 . β1 And β 2 stands out as the frequency bias constants of the respective areas. In the proposed system there are more than one GENCOs and so the Area Control Error needs to be shared between all the GENCOs proportional to the contributions for AGC. 3.
CONTROLLER STRUCTURE The Proportional Integral Derivative (PID) controller is the most trusted feedback controller used so far. The PID controller accepts an error value as its input. The error is basically the difference between the system variable to be controller and a standard reference point. The control action is exercised using three gain parameters: Proportional term [P]: mainly governed by error occurring at time t Integral term [I]: governed by the summation of all past error Derivative term [D]: involved in predicting errors occurring at t+1. Frequency regulation of modern power system using novel hybrid DE-DA algorithm (Sayantan Sinha)
106
ISSN: 2252-8814
The control action of a PID control can be mathematically represented as: Y(s) = 𝐾 + 𝑠𝐾 +
(7)
Where Kp stands for the proportional gain, Ki stands for the integral gain and Kd denotes the differential gain. This PID is a low order system but has its applications in broad areas. This type of controller mainly finds its application in SISO( single input single output) system. MIMO (Multi input multi output) systems basically are divided into a number of SISO loops and a PID controller is employed for individual of them. The robustness of the PID controller is the main reason for its wide spread acceptance in industrial control. However for an optimum operation of the PID controller, the selection of the above mentioned gains are of prime concern. 3.1. Tilted integral derivative controller The main objective of a Tilted Integral Derivative controller is to provide a feedback action as effective as the PID controller but with the results very close to the theoretical response. The TID controller has the proportional compensator of the PID controller replaced using a ‘tilt’ compensation. Mathematically 1
the ‘tilt’ can be expressed by s n . The tilt compensator mainly generates a frequency dependent feedback gain with a tilt or a shape inclination towards the theoretical or conventional compensation value. As a result the compensation came to be termed as Tilted Integral Derivative controller. The value of ‘n’ for the tilt function is between 2 and 3. When compared to the conventional PID controller, the coefficients of the respective transfer functions mainly has a value of 0,+1 or -1. On contrary the frequency coefficient in case of tilted controller has a coefficient of (1/n). This effectively helps to improve the compensating action of the controller. The schematic diagram of the Tilted integral derivative controller is given in Figure 2.
1 1 s n
Figure 2. The tilted integral derivative controller 4. OPTIMIZATION TECHNIQUE PROPOSED 4.1. Differential evolution First coined by storn and price [22], the differential evolution is mainly known for its simplicity, efficacy and a tremendous high sense of reliability. It is by principle different from genetic algorithm because the technique of Genetic Algorithm still relies on crossover whereas Differential evolution mainly focuses on mutation process for the generation of new solutions. The successive difference between the two solutions of a randomly generated population mainly brings about the process of mutation. The optimization technique starts with the generation of a random population of size N. Each and every member of the population is to be maintained within limits. The workflow of the optimization technique mainly includes three processes namely mutation, crossover and selection. In this optimization scheme, two randomly generated populations are taken into considerationi.e the old population and the new population in which the members are generated from the old one by the process of mutation and crossover. The cross over operation done by mixing the mutant vector parameters with the target vector gives rise to a trial vector. The target vector in the next generation is substituted by the trial vector. The evolutionary pathway of DE is shown below:
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 103 – 116
Int. J. of Adv. in Appl. Sci. a.
ISSN: 2252-8814
107
Initialization of a random population of size NP consisting of real parameter vectors of size D. If the upper bound of the population is
X Uj and the lower bound is X JL then it is to be ensured that the randomly
U
L
generated populations stay between the limits X J , X J . b.
The mutation process usuallydeals with a vector from the old population to act as the target vector for the new generation. The mutation process is mainly done by calculated the weighted difference between two randomly generated vectors and adding it to another randomly generated vector. The process can be mathematically expressed as:
𝑉𝑖,𝐺+1 = 𝑋𝑟1,𝐺 + 𝐹𝐶(𝑋𝑟1,𝐺 − 𝑋𝑟2,𝐺 ) c.
(8)
Where FC is a constant which varies from (0,2). The main advantage of henew vector generated by mutation is that it increases the diversity of the search space by a large extent. The crossover operation usually needs the addition of the mutant vector to the target vector and formation of a new vector. It basically involves three parents for the formation of their offspring’s. Mathematically it can be expressed as:
Vj,i,G1ifrandj,i CRorj Irand Uj,i,G1 .....(9) X j,i,G1ifrandj,i CRorj Irand
(9)
In order to maintain the population size throughout the entire length of the process, the selection operation is necessary. This operation involves the target vector Xi,G to be compared with Vi,G+1 and the better fitness possessing vector attains entry to the new generation. 4.2. Dragonfly algorithm Brought to picture by Mirjalili [23], the dragonfly algorithm draws its inspiration from the steady streaming activities of the dragonflies. Supposedly small creatures and harming almost all other insects, the most astonishing fact about the Dragonflies are their strictly adhered social behaviour. The reason for swarming behaviour displayed by the Dragonflies is for the purpose of hunting and migration. The process of hunting is discussed as feeding or static swarm and that of migration is termed as dynamic or migratory swarm [24]. These two behavioural traits replicate the two main phases of optimization: exploration and exploitation. These two phases can be expressed as follows: Separation: the term implies the tendency of individuals to avoid collision from other nearby individuals in their path of motion. Alignment: The term stands for the matching of velocities of one individual with other in the entire population. Cohesion: The term implies the tendency of each individual to travel towards the centre part of the group i.e. towards the central solution. Attraction towards a food: This term mathematically implies the attractive behaviour of the dragonflies towards any random source of food. Distraction from enemy: the term correctly hints at the unique ability of the dragonflies to distract their enemy to a separate course and saving the entire population. Taking into consideration the above behaviour of the flies, they are mathematically modelled as:
(10) Where Si is the separation vector and X is the position of the current individual and Xj is the position of the jth individual.
(11) Frequency regulation of modern power system using novel hybrid DE-DA algorithm (Sayantan Sinha)
108
ISSN: 2252-8814
Where Ai is the alignment vector and Vj is the velocity of the current individual that is to be matched with the rest of the swarm.
(12) Where Ci is the cohesion vector, N indicates the number of particles, Xj indicates the position of the current particle and X indicates the central solution as discussed above.
Fi X X
(13)
Where X+ and X denotes the current individual position and that of the source of food respectively.
Ei X X
(14)
Where X- and X stands for the position of the current dragonfly and its enemy respectively. The step vector for the updation of the dragonfly position can be calculated as the summation of all these factors and can be mathematically expressed as: X t 1 sS i aA i cC i fFi eE i w X i
(15)
After the generation of the step vector, the position of the dragonflies is updated as: X t 1 X t X t 1
(16)
In the absence of any neighboring immediate solution., the dragonflies position is put to updation using a random walk pattern namely called the Levy flight. Thus the position can be mathematically expressed as:
X t 1 X t Levy (d ) X t
(17)
Where d stands for the distance that is to be covered by Levy flight phenomena. 5.
HYBRID OPTIMIZATION SCHEME A novel attempt has been made to bring about the amalgamation of the exploitative abilities of an evolutionary computation scheme namely the Differential Evolution and the explorative abilities of a swarm intelligence namely a Dragonfly Algorithm. The technique so developed was coined as Differential Evolution - Dragonfly Algorithm (DE-DA). The entire code has been developed in MATLAb 2013 b platform. Use of some standard unimodal and multimodal benchmark fucntions has been used for testing the efficacy anf the supremacy of the desined novel hybrid DE –DA technique. In order to prevent any slight bit of anomaly in the comparisons, the function evaluation parameters are considered the same for all the considered fucntions. The hybrid codes are simulated for 30 times and the results were put to statistical analysis. Table 2 summarizes these results. For each method the worst, mean, median, best, and standard deviation from the 30 independent runs were calculated and compared. In order to establish the robustness of the proposed hybrid technique, the standard deviation and the mean of the fitness values obtained over 30 iterations are calculated and observed. Lesser value of standard deviation indicates the equity that exists in the solutions over 30 runs. This infers that an algorithm with lesser value of standard deviation is more likely to be consistent.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 103 – 116
Int. J. of Adv. in Appl. Sci. Function name Sphere
Test function n
F1 ( x )
x
2 i
i 1 n
Schwefel
n
F2 ( x ) x i x i i 1
F3 ( x )
Schwefel
F4 (x) maxx i ,1 i x}
Step
i
(
i 1
N
S
fo pt 0
30
[100,100]n
30
[10,10]n
0
x j )2
30
[100,100]n
0
30
[100,100]n
0
30
[100,100]n
0
30
[32,32]n
0
30
[500,500]n
0
30
[50,50]n
0
j1
n
F5 ( x )
109
i 1
Rotated hyper ellipsoid
n
ISSN: 2252-8814
([ x
i
0 .5]) 2
i 1
Ackley
F6 ( x ) 20 exp( 0.2 Griewank function Penalised
n F7 x i sin( x i i 1 F8 ( x )
Function name F1
F3
F4
F5
n n 10 sin 2 (y i ) ( y i 1) 2 [1 10 sin 2 (3y i 1 )] u ( x i ,10,100,4) n i 1 i 1
yi 1
F2
1 n 2 1 n x i ) exp( cos 2 x i ) 20 e n i 1 n i 1
xi 1 4 METHOD DE DA DE-DA DE DA DE-DA DE DA DE-DA DE DA DE-DA DE DA DE-DA
BEST 12.5817 18.1680 7.5605 1.95e+04 1.26e+04 228.6165 1.16e+3 1.12e+03 47.8718 0.2221 0.0553 0.0248 45.4651 40.8339 6.0574
DE DA DE-DA DE DA DE-DA DE DA DE-DA
14.1712 6.2233 3.2943 6.4211 3.4524 0.6638 14.0331 3.2985 0.1777
WORST 72.6387 58.8918 55.6477 1.07e+07 8.001e+07 1.57e+07 1.26e+04 1.22e+04 2.19e+03 3.0569 10.1392 1.4754 150.5548 107.4640 41.2643
MEAN 35.2792 35.3926 30.0650 1.92e+06 4.96e+06 9062e+05 6.194e+03 4.27e+03 748.8607 0.8643 2.2868 0.3595 102.1089 71.5784 17.1603
STD 16.5974 10.0163 12.9463 2.311e+06 1.445e+07 2.9932e+06 5.76e+03 4.56e+03 871.0934 0.6937 2.3062 0.3832 25.2523 21.2027 9.9018
MULTIMODAL FUNCTIONS F6
F7
F8
20.2920 19.9485 16.1654 124.8288 75.4961 36.7725 7.454e+7 1.3e+7 2e+6
18.6768 15.6799 9.4173 49.3711 30.1153 6.7834 7.9e+6 1.3e+6 1.65e+5
1.6130 3.1122 3.3874 39.0981 20.4420 9.1282 1.66e+5 2.7e+6 4.3e+5
6.
ANALYSIS FOR DE-DA The best values, worst values, mean values and standard deviation values for the above mentioned eight benchmark functions in Table 1. Table 2 clearly indicates that the values of the statistical parameters are the least for the DE-DA hybrid technique. The results were obtained for 100 iterations. Added to this the DEDA also accounts for the lowest value of standard deviations among the three techniques. This low value of standard deviation implies that the solutions obtained for the 30 iterations are approximately constant and this is a test for consistency of the proposed optimization scheme. Figure 3, Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9, and Figure 10 are showing the convergence curve of DE, DA and DE-DA in 8 functions. Frequency regulation of modern power system using novel hybrid DE-DA algorithm (Sayantan Sinha)
110
ISSN: 2252-8814
Figure 3. Convergence curve of DE, DA and DE-DA for function 1
Figure 4. Convergence curve of DE, DA and DE-DA for function 2
Figure 5. Convergence curve of DE, DA and DE-DA for function 3
Figure 6. Convergence curve of DE, DA and DE-DA for function 4
Figure 7. Convergence curve of DE, DA and DE-DA for function 5
Figure 8. Convergence curve of DE, DA and DE-DA for function 6
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 103 – 116
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
Figure 9. Convergence curve of DE, DA and DE-DA for function 7
111
Figure 10. Convergence curve of DE, DA and DE-DA for function 8
7.
APPLICATION OF DE-DA TO AGC PROBLEM The proposed hybrid Differential evolution - Dragonfly Algorithm is effectively implemented for obtaining the optimal gain parameters of the Tilted Integral Derivative controller. The hybrid algorithm is said to have a better performance than the De and DA as shown in Table 2. Table 3 lists down the optimal values of the controller gains. The application of the hybrid algorithm to the AGC is governed by the following pseudo code. Pseudo code for the hybrid optimization: Generation of the initial population and specification of the Differential evolution parameters like cross over ratio and mutation frequency. Computation of the fitness functions of each of the population members. Generation of offsprings using mutation technique. Calculation of the fitness function of each generated offspring. Selection of the best offspring by cross over technique that includes the parent. Replacement of the parents in the total population with the off springs and give rise to a final population. The final DE population acts as the initial population for Dragonfly Algorithm. Initialization of the step vectors X i
Computation of the fitness values of all the dragonflies. Updation of the food source and the enemy position. Updation of cohesion vector, separation vector, alignment vector, food distance vector and the enemy attack vector. Based on condition of at least one dragonfly in the neighbourhood of the current dragonfly, updation of the position is done. Checking of the new particles under boundary conditions. Jump over to step 2 until maximum iteration is reached. Terminate the loop. 8.
RESULTS AND ANALYSIS The discussed two area power system model is put to simulation with the help of MATLAB/ Simulink. The tilted integral derivative controller is employed in this case as the secondary controller for minimizing the Area Control Error (ACE) to zero. The gains of the TID controller are set to optimal values with the help of a novel hybrid DE-DA technique. The objective functions employed in the tuning process are as follows: t IAE f1 f 2 p tie dt 0
t ITAE f f f 2 p tie tdt 0
t ISE f12 f 2 p 2 dt 2 tie 0
(15)
(16)
(17)
Frequency regulation of modern power system using novel hybrid DE-DA algorithm (Sayantan Sinha)
112
ISSN: 2252-8814
t ITSE f12 f 2 p 2 tdt 2 tie 0
(18)
The values of each of the above objective functions during the tuning process is noted and scripted in Table 2. The table clearly indicates that the value of the objective function J is the minimum in the case of ISE (Integral square error) and henceforth all the tuning process in the paper has been done considering ISE as the cost function. Table 1. Optimized system parameters by hybrid DE-DA technique Market scenario Base case Billateral contract
Kt1 0.2401 0.2556
Kd1 0.2802 0.1503
Ki1 0.1590 0.3057
Kt2 0.1539 0.1995
Kd2 0.0952 0.1976
Ki2 0.1302 0.2974
Table 2. Objective function values for base case and bilateral transactions Market scenario Base case Billateral transaction
ITAE 1.9137 1.6242
Objective function IAE ITSE 0.5937 0.0782 0.5898 0.0837
ISE 0.0398 0.0409
The analysis is done on two cases: Case 1: the base case In this case the GENCOs and DISCOs come into mutual participation for the electricity market over a common area. This implies DISCOs of one area can only come in contract with the GENCOs of the respective areas. This mutual contract scenario can be represented with the help of Distribution Participation Matrix (DPM). 0 .5 0 .5 0 .5 0 .5 DPM 0 0 0 0
0 0 0 0 0 0 0 0
(19)
The area participation factor apf is taken equal for all the GENCOs in this particular case. So apf1 = apf2 = apf3 = apf4 =0.5 The simulation is carried out inclusive of a SLP of 0.01 pu in area 1. The demands of DISCOs are generally fixed at 0.1 MW. Figure 11, Figure 12 and Figure 13 represents the frequency deviation in area 1 ( f1 ), area 2 frequency deviation ( f 2 ) and tie line power flow deviation ( p tie ) respectively. Table 3 clearly gives us the values of settling time, maximum overshoot and minimum undershoot for the DE DA optimized TID controller. Effective comparison has been done with the settling time, maximum overshoot and minimum undershoot of DE optimized PID controller when applied to the same physical system under base case. From the tabular comparison it can be easily inferred that the DE-DA tuned TID controller is much more robust and effective than the De tuned PID controller. Figure 14, Figure 15, Figure16 and Figure 17 elaborates the power output pattern of GENCO 1, GENCO 2, GENCO 3 and GENCO 4 respectively.
Figure 11. Frequency deviation of area 1
Figure 12. Frequency deviation of area 2
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 103 – 116
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
Figure 13. Tie line power deviation
Figure 14. Power supplied by GENCO 1
Figure 15. Power supplied by GENCO 2
Figure 16. Power supplied by GENCO 3
113
Figure 17. Power supplied by GENCO 4 Case 2: Billateral transaction In this case the DISOCs and the GENCOs of both area come to a mutual contract participation and the scenario is expressed by the following DPM. 0.5 0.2 DPM 0 0.3
0.25 0 0.3 0.25 0 0 0.25 1 0.7 0.25 0 0
(20)
The area participation factor in this case is taken as reference to [18]. apf1=0.75; apf2=1-apf1= 0.25; apf3=apf4 = 0.5. The given system is simulated under bilateral contract scenario subject to a step load perturbation of 0.01 pu. Figure 18, Figure 19 and Figure 20 clearly portrays the frequency deviation in area 1 ( f1 ), area 2 frequency deviation ( f 2 ) and tie line power flow deviation ( p tie ) respectively. Table 4 lists down the values of the settling time, maximum overshoot and minimum undershoot Frequency regulation of modern power system using novel hybrid DE-DA algorithm (Sayantan Sinha)
114
ISSN: 2252-8814
for the hybrid optimized TID controller and is put to comparison with DE optimized PID controller [18]. Figure 21, Figure 22, Figure 23 and Figure 24 clearly portrays the power output of GENCO 1, GENCO 2, GENCO 3 and GENCO 4 respectively. The comparison clearly indicates the superiority of DE-DA optimized TID controller over DE optimized PID controller under bilateral contract scenario.
Figure 18. Frequency oscillations in area 1
Figure 20. Tie line power variations
Figure 22. Power supplied by GENCO 2
Figure 19. Frequency oscillations in area 2
Figure 21. Power supplied by GENCO 1
Figure 23. Power supplied by GENCO 3
Figure 24. Power supplied by GENCO 4
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 103 – 116
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
115
Table 3. Dynamic system parameters under base case scenario Controller type
System data
Settling time 5.52
Maximum overshoot 0.0409
Minimum undershoot -0.2574
4.51
0.012
-0.0946
p tie
1.88
0.0057
-0.0315
f1 f 2
4.4831
0.3500
-0.0090
4.420
0.9400
-0.0055
p tie
7.0473
0.0057
-0.0315
f1 f 2
PID[18]
TID
Table 4. Dynamic system parameters for bilateral scenario Controller type
System data
f1 PID[18]
TID
f 2
Settling time 10.87
Maximum overshoot 0.0988
Minimum undershoot -0.4664
11.47
0.0683
-0.2158
p tie
3.08
0.05
-0.0185
f1
7.6598
0.1149
-0.1551
f 2
7.1017
0.1096
-0.0975
p tie
9.4850
0.0083
-0.0348
9.
CONCLUSION Whenever a modern interconnected power system is considered for study under restructured market scenarios, the power system becomes highly sensible to slight load disturbances. As of concern the maintainence of frequency within scheduled values should be of a major importance. This is effectively attained by the Tilted Integral Derivative controller in the proposed work. Development of a hybrid optimization technique, the DE-DA technique is successfully verified and tested with standard benchmark functions. This technique is thereafter used to provide optimal values to the controller for minimizing the ACE to zero. The hybrid tuned controller has proved to be effective in minimizing frequency as well as tie line power deviations in a shortest time period possible. In order to justify the supremacy of the controller, the results are compared with pre published results. The dynamic performance indices of the system is analysed in both the cases and is listed down in respective tables for clear inference. It can be clearly established that the values of settling time, maximum overshoot and minimum undershoot is considerably less for both market scenarios taken into consideration. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8]
Dong, X., Sun, H., Wang, C., Yun, Z., Wang, Y., Zhao, P., Ding, Y., and Wang, Y., “Power Flow Analysis Considering Automatic Generation Control for Multi-Area Interconnection Power Networks,” IEEE Transactions on Industry Applications, vol 53(6), pp. 5200-8, 2017. Kumar, J., Ng, KH., and Sheble, G., “AGC simulator for price-based operation. I. A model,” IEEE Transactions on Power Systems, vol. 12(2), pp. 527-32, 1997. Kumar, J., Ng, KH., and Sheble, G., “AGC simulator for price-based operation. II. Case study results,” IEEE Transactions on Power Systems, vol. 12(2), pp. 533-8, 1997. Sinha, N., Lai, LL., and Rao, VG., “GA optimized PID controllers for automatic generation control of two area reheat thermal systems under deregulated environment,” Electric Utility Deregulation and Restructuring and Power Technologies, 2008. DRPT 2008, pp. 1186-1191, 2008. Saha, A. and Saikia, LC., “Utilisation of ultra-capacitor in load frequency control under restructured STPP-thermal power systems using WOA optimised PIDN-FOPD controller,” IET Generation, Transmission & Distribution, vol. 11(13), pp. 3318-31, 2017. Raju, M., Saikia, LC., Sinha, N., and Saha, D, “Application of antlion optimizer technique in restructured automatic generation control of two-area hydro-thermal system considering governor dead band," Power and Advanced Computing Technologies (i-PACT), pp. 1-6, 2017. Hasanien, HM. and El-Fergany, AA, “Symbiotic organisms search algorithm for automatic generation control of interconnected power systems including wind farms,” IET Generation, Transmission & Distribution, vol. 11(7), pp. 1692-700, 2016. Shiva, CK. and Mukherjee, V., “Automatic generation control of multi-unit multi-area deregulated power system using a novel quasi-oppositional harmony search algorithm,” IET Generation, Transmission & Distribution, vol. 9(15), pp. 2398-408, 2015.
Frequency regulation of modern power system using novel hybrid DE-DA algorithm (Sayantan Sinha)
116 [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24]
ISSN: 2252-8814
Apostolopoulou, D., Domínguez-García, AD., and Sauer PW., “An assessment of the impact of uncertainty on automatic generation control systems,” IEEE Transactions on Power Systems, vol. 31(4), pp. 2657-65, 2016. Liu, S., Liu, PX., and El Saddik, A., “Modeling and stability analysis of automatic generation control over cognitive radio networks in smart grids,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 45(2), pp. 223-34, 2015 Law, YW, Alpcan, T., and Palaniswami, M., “Security games for risk minimization in automatic generation control,” IEEE Transactions on Power Systems, vol. 30(1), pp. 223-32, 2015. Sridhar, S. and Govindarasu, M., “Model-based attack detection and mitigation for automatic generation control,” IEEE Transactions on Smart Grid, vol. 5(2), pp. 580-91, 2014. Jagatheesan, K., Anand, B., Samanta, S., Dey, N., Santhi, V., Ashour, AS., and Balas VE, “Application of flower pollination algorithm in load frequency control of multi-area interconnected power system with nonlinearity,” Neural Computing and Applications, vol 28(1), 475-88, 2017. Khadanga, RK. and Kumar, A., “Hybrid adaptive ‘gbest’-guided gravitational search and pattern search algorithm for automatic generation control of multi-area power system,” IET Generation, Transmission & Distribution, vol. 11(13), pp. 3257-67, 2016. McNamara, P., Meere, R., O'Donnell, T., and McLoone, S., “Control strategies for automatic generation control over MTDC grids,” Control Engineering Practice, vol. 54, pp. 129-39, 2016. Venkat, AN., Hiskens, IA., Rawlings, JB. and Wright, SJ., “Distributed MPC strategies with application to power system automatic generation control,” IEEE transactions on control systems technology, vol. 16(6), pp. 1192-206, 2008. Variani, MH. and Tomsovic, K., “Distributed automatic generation control using flatness-based approach for high penetration of wind generation,” IEEE Transactions on Power Systems, vol. 28(3), pp. 3002-9, 2013. Ibrahim, AN., Shafei, MA., and Ibrahim, DK., “Linearized biogeography based optimization tuned PID-P controller for load frequency control of interconnected power system,” Power Systems Conference (MEPCON), pp. 1081-1087, 2017. Selvaraju, RK. and Somaskandan, G., “ACS algorithm tuned ANFIS-based controller for LFC in deregulated environment,” Journal of Applied Research and Technology, vol. 15(2), pp. 152-66, 2017. Gorripotu, TS., Sahu, RK., and Panda, S., “AGC of a multi-area power system under deregulated environment using redox flow batteries and interline power flow controller,” Engineering Science and Technology, an International Journal, vol. 18(4), pp. 555-78, 2015. Hota, PK. and Mohanty B., ”Automatic generation control of multi source power generation under deregulated environment,” International Journal of Electrical Power and Energy Systems, vol. 75, pp. 205-14, 2016. Storn, R. and Price, K., “Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces,” Journal of global optimization, vol. 11(4), pp. 341-59, 1997 Mirjalili, S., “Dragonfly algorithm: a new meta-heuristic optimization technique for solving single-objective, discrete, and multi-objective problems,” Neural Computing and Applications, vol. 27(4), pp. 1053-73, 2016. KS SR, Murugan, S., “Memory based hybrid dragonfly algorithm for numerical optimization problems,” Expert Systems with Applications, vol. 83, pp. 63-78, 2017.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 103 – 116
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 117~124 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp117-124
117
Enhanced performance of PID load frequency controller for power systems Dola Gobinda Padhan, Suresh Kumar Tummala EEE Department, Gokaraju Rangaraju Institute of Engineering and Technology, India
Article Info
ABSTRACT
Article history:
A novel control structure for designing a PID load frequency controller for power systems is presented. The controller with a single tuning parameter is designed based on a desired closed-loop complementary sensitivity function and Pade approximation. Comparative analysis demonstrates that proposed PID controllers improves the settling time and reduces overshoot effectively against small step load disturbances. Also, the performance and robustness of the controllers have been analyzed and compared. Simulation results show significantly improved performances when compared with recent results.
Received Sep 8, 2018 Revised Mar 27, 2019 Accepted Apr 29, 2019 Keywords: Complementary sensitivity Kharitonov’s theorem Load frequency control (LFC) PID
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: Dola Gobinda Padhan, EEE Department, Gokaraju Rangaraju Institute of Engineering and Technology, Survey No. 288 Nizampet Road, Krishnaja Hills, Bachupally, Kukatpally, Hyderabad, Telangana 500090, India. Email: dg.padhan@griet.ac.in
1.
INTRODUCTION Frequency deviation in Power System due to variation between generation and load shall be rectified within a fraction of seconds resulting in stability and security. Load Frequency Control (LFC) of an extensive power framework can be alluded as the issue of controlling the recurrence by directing the created units with reaction to change in stack [1]. For framework soundness, LFC must furnish recurrence with zero enduring state mistakes and tie-line trade varieties, high damping of recurrence motions and diminishing overshoot of the unsettling influence. The objectives specified are conveyed effectively in past works by various creators utilizing Fuzzy rationale PI and PID controllers [2, 3], ideal control [4, 5]. Variable structure control [6, 7], versatile and self-tuning control [8, 9]. Down the line, different tuning rules have picked up the consideration for the previously mentioned goals in which Internal Model Control (IMC) [10] is one among them. The LFC PID controller configuration utilizing Laurent arrangement is clarified by Padhan and Majhi [11]. Double PI controller tuning utilizing swam enhancement calculation is introduced in [12]. The two-degree-of-freedom internal model control scheme suggested by Tan [10] consists of two controllers with two tuning parameters where simultaneous tuning of the two parameters is difficult. In practice, a simple control structure with a fewer number of tuning parameters is desirable. The proposed control structure (see Figure 1) for LFC design consists of only one controller (Gc). Kasireddy et.al designed a PID controller for LFC through reduced model order [13].
Journal homepage: http://iaescore.com/online/index.php/IJAAS
118
ISSN: 2252-8814
Figure 1. Proposed control structure In Figure 1, G and Gme−θm represent the power system dynamics and its model, respectively. For LFC, controller design is inconvenient because G results in higher order plant models, which are approximated by lower order transfer functions with time delay using a relay-based identification method. This paper has been alienated into 6 sections. Modeling of power system dynamics with necessary derivations discourse in section 2. In section 3, the PID controller design method is discussed followed by Section 4 in which the simulation results are presented. Section 5 deals with Robustness analysis and performance of a power system using Kharitonov’s rectangles followed by conclusions in section 6. 2.
MODELING OF POWER SYSTEM DYNAMICS Figure 2 shows single area power systems with a linear model. From Figure 2 it can be noticed that the power is supplied to the single area by a single generator. There are two types of turbine used for a generation: (a) non-reheated (NRT) and reheated (RT).
Figure 2. Single area power system The plant model used for LFC without droop characteristics is (1)
G = G g G t Gp
Where Gg, Gt, and Gp are the dynamics of the governor, turbine, and load & machine, respectively. For a reheated turbine,
Gt =
cTr s 1 Tr s 1 TT s 1
Where Tr is a constant and c is the portion of the power generated by the reheat turbine in the total generated power. For non-reheated turbine Tr = 0. The plant model used for LFC with droop characteristic is
G=
GgGt Gp
1 GgGt Gp / R
From (1) and (2) can be represented by the second-order transfer function model
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 117 – 124
(2)
Int. J. of Adv. in Appl. Sci.
G=
ISSN: 2252-8814
ke m s T1s 1 T2 s 1
119
(3)
State space equations in the Jordan canonical form become Ax(t) bu(t m ) x(t)
(4)
y(t) cx(t)
(5)
Where 1 T 1 A 0
0 k ; b = 1 ; c = 1 1 1 T1 T2 1 T2
When a relay test is performed with symmetrical relay of height ±h, then the expression for the limit cycle output for 0 ≤ t ≤ θm is
y(t) ce At x(0) cA 1 e At I bh
(6)
Let the half period of the limit cycle output be τ. Then the expression for the limit cycle output for θm ≤ t ≤ τ is
y(t) ce A( t m ) x(m ) cA 1 e A(t m ) I bh
(7)
The condition for a limit cycle output can be written as
y(0) cx(0) y( ) 0
(8)
Substitution of t = τ in (7) and use of (6) gives the initial value of the cycling states
x(0) I e A
1
A 1 2e A ( m ) e A I bh
(9)
When tp is the time instant at which the positive peak output occurs and tp ≥ θm, then the expression of the peak output Ap becomes
Ap c e
A( tp m )
x(m ) A 1 e
A(t p m )
I bh
(10)
and the expression for the peak time becomes t p m
1 e / T1 T1T2 ln T1 T2 1 e / T2
(11)
Substitution of A, b, and c in (9) and (10) give
T1 1 e / T2
2e
( m )/ T1
A p kh 2 1 e / T1
e / T1 1 T2 1 e / T1
T1 T1 T2
1 e
/ T2
T2 T1 T2
1
2e
( m )/ T2
e /T2 1 0
(12)
(13)
Enhanced performance of PID load frequency controller for power systems (Dola Gobinda Padhan)
120
ISSN: 2252-8814
The (11-13) are solved simultaneously to estimate θm, T1 and T2 from the measurements of , Ap and tp. The relentless state gain k is thought to be known from the earlier or can be assessed from a stage flag test. Care has been taken to explain the arrangement of non-direct conditions, so intermingling may not occur to a false arrangement. 3.
PID CONTROLLER DESIGN The nominal complementary sensitivity function for load disturbance rejection can be obtained as
GGc 1 GGc
T=
(14)
To reject a step change in the load of the power system, the asymptotic constraint should be satisfied so that the closed loop internal stability can be achieved [3]. lim s
1 1 , T1 T2
1 T 0
(15)
The desired closed-loop complementary sensitivity function is proposed as
s T 2
2
e
1s 1
s 1
4
m s
(16)
Where β is the only tuning parameter for obtaining the desired performance of the power system. As there always exists a trade-off between the nominal performance and robust performance, β must be tuned according to the desired choice. α1 and α2 can be obtained from (15) and the constraint as 4 4 m / T2 T12 1 e m / T1 1 T22 1 1 e T1 T2 and 1 T2 T1
4 2 T22 1 em /T2 1 T2 1 T2
(17)
Using (14), (15) and second order Pade´ approximation for the time delay term, we get Gc
6 2 s2 1s 1 l2 s2 l1s 1
(18)
m m kmo s 2 s2 1 s 1 m m o o
Where
l1
6T2 6T1 4m
6 m0 6m 24 61
l2
6T T
1 2
4T1m 4T2 m m2
6
m1 362 m2 6 2 16m 21m m2 24 2 m 2 2 m 4m2 243
The (18) can be written in the form of a PID controller with lead/lag filter as
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 117 – 124
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
a s 2 a1s 1 1 Gc K c 1 Td s 2 2 T2 s b 2 s b1s 1
121
(19)
Where
Kc
61 km0
a2 l2
Ti 1
Td
2 1
a1 l1
b2
m2 m0
b1
m1 m0
4.
SIMULATION RESULTS Consider a power system with a non-reheated and a reheated turbine whose model parameters are given by KP = 120, TP = 20, TT = 0.3, TG = 0.08, R = 2.4, Tr = 4.2 and c = 0.35 [11]. The identified models and controller settings (see Table 1) for the power system with non-reheated and reheated turbines are obtained using (11-13, 17). The Nyquist plots of the identified and actual models are shown in Figure 4 to illustrate the accuracy of the identification method. To get stable and robust response, β values in Table 1 are obtained from extensive simulation studies. Figure 3 and Figure 5 show the frequency change of the power system following a load demand ΔPd = 0.01. The stability robustness is tested by changing the parameters of the system by 50%. From the simulation results, it is evident that the proposed method gives significantly improved performances than the Tan’s method. Table 1. Control parameters for identified model Model Type
Identified Model 120e0.4626s 28.4952s 1 0.2202s 1
Kc=2.0245, Ti=0.5005, Td=0.1332, a1=29.0238, a2=15.1661, b1=28.6982, b2=5.77239, β=0.01
NRT (D)
250e0.05s 2.028s 12.765s 106.2
Kc=0.7192, Ti=0.2075, Td=0.1159, a1=0.9212, a2=0.1411, b1=0.1515, b2=0.0234, β=0.07
RT (WD)
120e 0.541s 23.2137s 1 0.9057s 1
Kc=3.6549, Ti=0.5797, Td=0.2355, a1=24.4801, a2=29.7725, b1=24.0884, b2=20.2681, β=0.01
RT (D)
235.3e0.035s 1.79s2 16.9s 100
Kc=1.0619, Ti=0.2107, Td=0.1828, a1=1.154, a2=0.1323, b1=0.1973, b2=0.0231, β=0.065
NRT (WD)
2
Control Parameters
(a) Nominal systems
(b) Parameters of the system change by 50% Figure 3. Frequency deviation of the closed loop system with non-reheated turbine Enhanced performance of PID load frequency controller for power systems (Dola Gobinda Padhan)
122
ISSN: 2252-8814
Figure 4. Nyquist plots for the power system with non-reheated turbine
Figure 5. Frequency deviation of the closed loop system with reheated turbine 5.
ROBUSTNESS ANALYSIS AND PERFORMANCE In this section, Robustness of the system has been analyzed using Kharitonov’s Theorem. Closedloop characteristic equation CL (s) and denominator of the closed-loop transfer function T(s) are the polynomials that make the control system stable. Considering the forward-path and feedback-path transfer functions G(s) and H(s), characteristic equation is CL (s) = 1+G(s)H(s) = 0
CL (s) an sn an 1sn 1 ...... a1s a0
(20)
For simplicity, assume that the leading coefficient an is constant and the coefficients have been normalized so that an = 1. The polynomial coefficients can then be expressed as ai amin ,aimax , i
i=0,1.......n-1
(21)
so, the characteristic equation becomes CL (s) sn an 1sn 1 ...... a1s a0
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 117 – 124
(22)
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
123
According to Kharitonov’s Theorem, an nth-degree interval polynomial family described by (1a) and (1b) is robustly stable if and only if each of the four Kharitonov polynomials is stable, that is, all the roots of those polynomials have strictly negative real parts. For the system G(s)
120 0.08s 1 0.3s 1 20s 1
The characteristic equation is
CL (s) 0.48s3 7.624s2 20.38s 121 0
(23)
For ±10% variations in the coefficients of the polynomial, The intervals of the polynomial will become
a3 0.528,0.432
a2 8.3864,6.8616
a1 22.418,18.342
a0 133.1,108.9
Figure 6 shows Kharitonov’s rectangles rotate around the origin in a counter-clockwise direction to satisfy the monotonic phase increase property of Hurwitz polynomials. For clarity, the graph is zoomed in Figure 7 to show the zero-exclusion point. As the Kharitonov’s rectangles do not pass through the origin, it is concluded that the closed loop system guarantees the robust stability.
Figure 6. Response of the system with ±10% variations in the coefficients of polynomial on the complex plane
Figure 7. Response of the system showing zero exclusion point Enhanced performance of PID load frequency controller for power systems (Dola Gobinda Padhan)
124
ISSN: 2252-8814
Table 2 shows it is observed that the proposed method gives less Integral Absolute Error (IAE), Integral Squared Error (ISE) and Integral Time Absolute Error (ITAE) as compared to Tan’s method, so closed loop performance is improved. If we compare the total variations of the control signals, the results of both methods are almost same. Thus, with same control signals, the proposed method gives comparatively fewer errors. Table 2. Various errors and total variations Type of Model NRT (WD)-Tan NRT (WD)-Proposed NRT (D)-Tan NRT (D)-Proposed RT (WD)-Tan RT (WD)-Proposed RT (D)-Tan RT (D)-Proposed
Integral Time Absolute Error 0.5164 0.5147 0.5303 0.521 2.002 2.058 2.006 1.966
Integral Absolute Error 0.1061 0.1008 0.106 0.1007 0.2007 0.2061 0.2007 0.2006
Integral Squared Error 0.001494 0.001184 0.001363 0.001175 0.002254 0.002273 0.01223 0.002264
Total variations 0.0199 0.0127 0.0214 0.0189 0.0735 0.07906 0.0729 0.0797
6.
CONCLUSION The Load Frequency Characteristics of a single-area power system with non-reheated and reheated turbines have been deliberated. The proposed method is flexible and gives satisfactory performance in nominal as well as the perturbed case. The proposed PID controller with a new control structure and a single tuning parameter (β) gave better performance than Tan’s controller. By showing the zero exclusion point by Kharitonov’s rectangles, it guarantees the robust stability for closed loop power systems. The proposed scheme can easily be extended to multi-area power systems. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]
R. K. Cavin, M. C. Budge, and P. Rasmussen, “An optimal linear system approach to load frequency control,” IEEE Trans. Power App. Systems, vol. 90(6), pp. 2472-2482, 1971. C. E. Fosha and O. I. Elgerd, “The megawatt-frequency control problem: A new approach via optimal control theory,” IEEE Trans. Power App. Systems, vol. 89(4), pp. 563-567, 1972. N. N. Bengiamin and W. C. Chan, “Variable structure control of electric power generation,” IEEE Trans. Power App. Systems, vol. 101(2), pp. 376-380, 1982. M. A. Sheirah and M. M. Abd-EI-Fattah, “Improved load frequency self-tuning regulator.,” Int. J. Control, vol. 39(1), pp. 143-158, 1984. P. Kundur, Power System Stability and Control. McGraw Hill, New York 1994. C. T. Pan and C. M. Liaw, “An adaptive controller for power system load-frequency control,” IEEE Trans. Power Systems, vol. 4(1), pp. 122-128, 1989. J. Talaq and F. Al-Basri, “Adaptive fuzzy gain scheduling for load frequency control,” IEEE Trans. Power Systems, vol. 14(1), pp. 145-150, 1999. M. F. Hossain, T. Takahashi, M. G. Rabbani, M. R. I. Sheikh, and M. Anower, “Fuzzy-proportional integral controller for an AGC in a single area power system,” Proc. 4th Int. Conf. Electrical and Computer Engineering, Dhaka, Bangladesh, pp. 120-123, 2006. S. Majhi, “Relay based identification of processes with time delay,” Journal of process control, vol. 17, pp. 93-101, 2007. W. Tan, “Unified tuning of PID load frequency controller for power systems via IMC,” IEEE Transactions on Power Systems, vol. 25(1), pp. 341-350, 2010. D. G. Padhan and S. Majhi, “A New Control Scheme for PID Load Frequency Controller of Single-area and Multiarea Power Systems,” ISA Transactions, vol. 52, pp. 242–251, 2013. M. Elsisi, M. Soliman, M. A. S. Aboelela, and W. Mansour, “Dual Proportional integral controller of Two-Area Load Frequency control based gravitational search algorithm,” Telkomnika Indonesian Journal of Electrical Engineering and Computer Science, vol. 15(3), pp. 397-406. 2015. Idamakanti Kasireddy, Abdul Wahid Nasir, and Arun Kumar Singh., “Non-integer IMC based PID design for load frequency control of power system through reduced model order,” International Journal of Electrical and Computer Engineering, vol. 8(2), pp. 837-844, 2018.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 117 – 124
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 125~135 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp125-135
125
Solar irradiance forecasting using fuzzy logic and multilinear regression approach: A case study of Punjab, India Sahil Mehta, Prasenjit Basak Electrical and Instrumentation Engineering Department, Thapar Institute of Engineering and Technology, India
Article Info
ABSTRACT
Article history:
The accurate forecasting of solar irradiance depends on various uncertain parameters like time of day, temperature, wind speed, humidity, and atmospheric pressure. All these play an important role in calculating PV power output. In this paper, a novel approach for forecasting of solar irradiance using flexible and accurate fuzzy logic and robust multi-linear regression approach has been proposed considering the above mentioned five variables. Based on the simultaneous consideration of those five variables, the solar irradiance is forecasted using the proposed methodology at a particular location in India, and the results are compared with the real time measured value of solar irradiance at that location on the days for which solar irradiance are forecasted. The proposed method is validated by comparing the results with real time data. The error analysis of the fuzzy logic based proposed system shows the root mean square error of 10.011 and mean absolute percentage error of 1.703%, while compared with real time data measured by instruments pyranometer, anemometer etc. The same results are found better while compared with the results obtained using multilinear regression approach.
Received Sep 11, 2018 Revised Apr 5, 2019 Accepted May 2, 2019 Keywords: Forecasting Fuzzy logic Microgrid planning Multilinear regression Solar irradiance
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: Prasenjit Basak, Electrical and Instrumentation Engineering Department, Thapar Institute of engineering and Technology, 147004 Patiala, Punjab, India. Email: prasenjit@thapar.edu
1.
INTRODUCTION In the present energy scenario, the depletion of the natural fossil fuels, increased environmental pollution and the effect of ageing of the installed power plants are becoming a challenge for the society. Thus, integrating the renewable energy sources (REs) with the grid is one of the key solutions to such problems. But this integration also leads to some major challenges and problems like the uncertain nature of these resources especially solar in terms of power generation is a critical issue. These sustainable forms of energy depends and vary according to the weather conditions like temperature, atmospheric pressure, humidity, wind speed and also the time of the day. Apart from these parameters, the solar power generation mainly depends upon the solar irradiance which further depends upon the various weather based parameters. Thus, the foremost step in solar power prediction systems is the forecasting of Global Horizontal Irradiance (GHI) [1]. Focusing on the benefits of RES forecasting, in a recent time, a number of studies on solar irradiance forecasting have been proposed, most of which require the historical data of the dependent parameters. In [2], number of existing techniques for forecasting, their advantages and computation requirements has been discussed. Also, the various challenges faced by increased penetration of renewable energy resources in power system as well as microgrid have been briefly discussed. In [3], a study based on demand response of an isolated microgrid under the impact of uncertainties in renewable energy forecasts and its performance has been studied. The result shows less influence of uncertainty in solar than that in wind power on the dispatch cost at the generating end and for the management of the flexible loads. The energy management system for an isolated Journal homepage: http://iaescore.com/online/index.php/IJAAS
126
ISSN: 2252-8814
microgrid with PV, ESS and electric vehicles using the hybrid solar irradiance forecasting methodology for a short period of time has been presented in [4]. The proposed methodology, based on the data available online from satellite and the data of clear sky for the selected microgrid location uses Heliosat algorithm for determining the cloud cover from the satellite data. In [5], the problem of short term forecasting the estimated PV power for planning of a microgrid has been solved using neural networks. The historical data of parameters like PV production, global irradiance for learning and training available at the test facility has been considered. In [6], using the Markov switching model, solar forecasting for a remote microgrid has been presented. The proposed method uses the previous data available on the online portals for solar irradiance and clear sky irradiance, whereas using Fourier expansion, linear model for different conditions have been created for sunny, mildly cloudy and extremely cloudy days. Overcoming the problem of uncertainty in microgrid integrated with renewable sources of energy, holt-winters forecasting method has been used in [7]. The proposed forecasting methodology has been incorporated in the microgrid simulation with assumed residential load and battery storage system as back up source. The results of the work give an idea of reduced cost by forecasting the accurate value of solar in terms of energy management system (EMS) for a microgrid. The performance of intelligent neural network (NN) tool and statistical linear regression tool for forecasting the PV output has been analyzed in [8]. Using the previous online available data for solar irradiance, temperature, training and testing of the multilayer feed forward neural network have been done. The results show artificial NN (ANN) tool to be more accurate and flexible whereas statistical regression being more robust and easy to construct. In [9], short term solar irradiance forecasting focusing on the variable nature of the weather conditions as well as dependent radiation. The work compares seasonal autoregressive (SAR) linear models for solar irradiance forecasting with the nonlinear models, i.e. nonlinear autoregressive (NAR) based on computational intelligence technique and regression tree (RT). In [10], long term load demand forecasting using a regression based approach has been presented. The work focuses on various variables like historical data of the local power utility company, the present number of consumers and development plans of the set location. Short term load forecasting for big data, i.e. huge historical data for various input parameters affecting the forecasting has been considered in [11]. Being more deterministic and robust multi linear regression methodology has been used for accurate forecasting. Various sensors and data science applications, historical data for load forecasting has been used. Focusing on the high matrix formulations, least number of parameters are considered with big data. The results of the work show 3.99% of the MAPE, considered to be under the tolerable limits. In [12], the forecasting of the electricity consumption in Thailand, as a case study, has been presented using artificial neural network and multi linear regression. Using the ANN technique the minimum cost of generation is calculated. Using the multi linear regression algorithm, depending on historical data of various parameters, electricity consumption has been calculated. In [13], dependent upon various meteorological parameters like temperature, dew point, humidity, visibility, wind speed and weather summary, short term solar irradiance forecasting has been proposed. Using the available meteorological data from various online sources, the forecasting problem has formulated as a structured output prediction problem. The proposed algorithm of long-short term memory has been compared with least square regression, persistence algorithm and different NN techniques. Based on historical data for cloud cover i.e. input parameters like azimuth, zenith angle and time of day. In [14], forecasting of solar irradiance using Labview software and regression based numerical algorithm has been proposed. Initially, dependent upon the parameters, solar irradiance has been forecasted and further combined with the local available temperature PV power output has been forecasted. [15] Presents multi-linear regression based solar power probabilistic forecasting, working with various inputs available at the weather stations. The system performance depends upon the cloud movement or the clear sky index is rated as better performance under the clear sky conditions. The results show that the accuracy of the system increases with the increase in historical data. In [16], the combined effect of relative humidity (RH) and air temperature (AT) on solar irradiance forecasting have been explored. Using the multivariate linear regression analysis carried out in different temperature range, the results come out to be more accurate and reliable in comparison to single variable dependent forecasting. Also, forecasting using the flexible fuzzy logic approach has been in trend among various authors in the recent years. In [17], focusing on the future aspects of energy planning, various techniques like neural networks, genetic algorithm and specially fuzzy logic approach along with the its application and advantages over the other; in terms of forecasting have been discussed. Overcoming the problem of uncertainty in weather conditions, in [18] solar radiation forecasting using fuzzy and neural net approach has been done. The results show the higher accuracy level and thus its application in the grid connected PV system. Similar to [18], considering the meteorological parameters like sky information, clearness index and temperature; solar radiation forecasting using a functional fuzzy logic approach has been suggested in [19]. Forecasting of solar energy using the fuzzy logic approach have been done focusing on the need of available solar energy data in [20] considering the sunshine per hour, temperature, latitude and longitude and time of year. With the increasing need of accurate forecasting using a fuzzy logic approach, forecasting of solar irradiance has been suggested in [21] normalizing the inputs and output to Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 125 – 135
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
127
eliminate convergence problems. The results show a high accuracy level of forecasted value with respect to measured value for short term forecasting. In a recent trend, i.e. planning and designing of a microgrid with installed RES, in [22] the fuzzy logic approach for forecasting of renewable energy based sources considering the data of meteorological parameters of a set location has been shown. The comparative results show better accuracy of the fuzzy logic approach in comparison to the neural network and vector autoregressive forecasting models. With the increasing need of forecasting in microgrid applications, in [23] the focus is on energy management system, i.e. uncertain renewable energy based source and variable load. For this purpose, a real microgrid system in Chile with wind and PV and variable load have been considered and day ahead forecasting using the fuzzy logic approach has been carried out. Hence, for the planning of a microgrid considering particular longitude and latitude, it is very necessary to have the accurate forecasted value of energy generation through the available renewable energy resources. Thus, from the literature survey using the historical data available at the online portals, the individual effect of few meteorological parameters such as temperature, sky movement, cloud cover, humidity and wind on solar irradiance forecasting has been studied. It is observed that the combined effect of five parameters such as temperature, wind, humidity, time of day and the atmospheric pressure at the location has not been considered simultaneously for solar irradiance forecasting because of complex system characteristics. On the basis of above discussion the main objective of the present work have been considered as solar irradiance forecasting and power calculation using the high dimension fuzzy logic and multilinear regression approach. In high dimension fuzzy logic approach a large number of rules (1200) combining the effect of above mentioned parameters are formed to get more reliable and accurate results whereas, the robust multilinear regression (MLR) approach has also been implemented to develop the mathematical relation between all the parameters and forecasted solar irradiance for different time intervals. For the purpose of validation, the forecasted values of solar irradiance, as the output of both fuzzy logic and multilinear regression approach, are compared with the real time solar irradiance using Pyranometer device. The work presented in this paper is organized as follows: The introduction of the article is covered in section 1 followed by description of measuring devices and data collection methodology in section 2. The design and development of the high dimensional fuzzy logic system and multilinear regression approach have been discussed in section 3. The results and discussion for the proposed approach is presented in section 4 and the conclusions in section 5 followed by references. 2.
MEASURING DEVICES AND DATA COLLECTION Focusing on the major objective of work, the data collection of various parameters is initiated. The parameters are identified as temperature in °C, wind speed in m/s, humidity in percentage, time of day in hours: minutes, atmospheric pressure in hectopascals (hpa) and solar irradiance in W/m 2. The advantage of measuring and collecting the actual real time data at a specific location is the high accuracy and system reliability. In Section 2.1 (A), 2.1(B) and 2.1 (C) the devices used for the measurement of such parameters at the set location at 30.356008, 76.372142 (30° 21′ 21.63″ N and 76° 22′ 19.71″ E) in the Thapar University campus have been discussed along with the data logging system in 2.1 (D). 2.1. Measuring devices 2.1.1. Pyranometer The Pyranometer device; Hand-Held terminal model (HHT-13) used for the measurement of global horizontal irradiance is shown in Figure 1 (a). It is a portable electronic device to which the GHI sensor shown in Figure 1 (b) can be attached. The major benefit of using the HHT-13 terminal model for measuring the GHI is the data logging facility. The real time GHI for a specific location can be stored as well as logged in the personal laptops using the installed software. 2.1.2. Anemometer The multipurpose device used to measure various meteorological parameters is the Lutron AM4214SD Hot wire Anemometer shown in Figure 1 (c). This device is used to measure various parameters like wind speed in m/s as well as in km/hr, temperature in °C and °F; and humidity in percentage in the real time. The benefit of such a device is the online monitoring as well as the data logging option which offers flexible operation of the device. The figure shows the anemometer device connected to one of the two separate probes which are used for the measurement purpose. Probe at the first port is used to measure the wind speed in m/s or km/hr where as at port two the probe for measuring humidity and the temperature can be connected.
Solar irradiance forecasting using fuzzy logic and multilinear regression approach … (Sahil Mehta)
128
ISSN: 2252-8814
2.1.3. Barometer Highly sensitive multipurpose electronic digital device capable of measuring the atmospheric pressure in hpa has been used in the presented work. Shown in Figure 1 (d), the device is capable of representing the altitude in meter (m), compass for direction and the atmospheric pressure.
(a)
(b)
(c)
(d)
Figure 1. Real time data measuring devices (a) Pyranometer device, (b) GHI sensor, (c) Anemometer device, (d) Barometer device 2.1.4. Data logging system Using the above mentioned devices; data have been logged for necessary experiment. Figure 2 shows the block layout of the data logging system. Hence, focusing on the measured parameters; through anemometer device connected to two different probes (humidity and temperature probe, wind speed probe) various meteorological parameters like wind speed in m/s and km/hr, temperature in °C and °F and humidity in percentage can be logged. The device stores the data in an external SD card through which the data in the form of Microsoft excel sheet can be transferred to the laptop memory.
Figure 2. Block diagram of data logging system The atmospheric pressure has been measured by the electronic barometer device. The other major parameter is the solar irradiance which is measured using the Pyranometer device and GHI sensor. Pyranometer device connected via USB cable to the personal laptop through hand held terminal software (HHT-13) has been well utilized to download the recorded data of solar irradiance in W/m2. Thus, based on the above layout of data logging system, Table A1 (submitted as supplementary sheet) shows the set of logged data measured at the set location of Patiala, Punjab, India (30.356008, 76.372142). The data have been measured using the above discussed devices (Anemometer, Barometer and Pyranometer). The table shows, the recorded value of relative temperature in °C, relative humidity in %, the wind speed in m/second, atmospheric pressure at the set geographical location in hectopascals (hpa) and the solar irradiance in W/m2. It is to be noted that the data is collected in the real time with 1 minute interval, whereas here a random sample set of data has been given for the 3 different days for which data are measured. Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 125 – 135
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
129
3.
PROPOSED METHODOLOGY Being user defined, simple and reliable, the fuzzy logic can also be termed as the simplified form of the classical logics involved in forecasting. The working of fuzzy logics is based on the non-linear mapping of input and output variables resulting in flexible output, i.e. rule viewer window, whereas multilinear regression approach truly depends upon the historical data of all inputs and gives the result in mathematical equations. 3.1. High dimensional based fuzzy logic approach Figure 3 shows the layout of the designed system with five uncertain weather based inputs which are temperature, humidity, wind speed, time of day, and atmospheric pressure and its output as solar irradiance. The inputs to the system, i.e. uncertain parameters are fuzzified in respective membership function in the range of [0, 1].
Figure 3. Layout of proposed fuzzy logic model (FIS) Based on the observed minimum and maximum value for each parameter distribution of the linguistic variables has been done. The membership functions for various inputs and output variable is shown in Figure 4, where the x-axis presents the parameter dependent value. For example; dependent upon the minimum and maximum values of temperature axis has been defined, whereas the divisions have been marked on the basis of observations and level of variations in Figure 4.
(a)
(b)
(c)
(d)
Figure 4. Input/output membership functions for designed FIS system (a) Temperature, (b) Humidity, (c) Wind speed, (d) Solar irradiance Solar irradiance forecasting using fuzzy logic and multilinear regression approach … (Sahil Mehta)
130
ISSN: 2252-8814
The membership function of the temperature as the input given to the FIS system shown in Figure 4 (a) is distributed in five linguistic variables, i.e. Very Low (VL), Low (L), Medium (M), High (H) and Very High (VH). Similar to Figure 4 (a), other dependent variables, i.e. humidity and wind speed are mentioned in Figure 4 (b) and Figure 4 (c) respectively with different number of linguistic variables. The output of the system, i.e. solar irradiance has been shown in Figure 4 (d). With the above proposed system, solar irradiance is forecasted depending upon different input parameters. 3.2. Multilinear regression approach Multiple linear regression model shows the relationship between multiple independent variables and a dependent variable. The mathematical model of the approach can be represented as (1)
𝑦 = 𝛽 + 𝛽 𝑥 + 𝛽 𝑥 + ⋯+ 𝛽 𝑥 + 𝜖
where, y is the dependent variable, 𝑥 , 𝑥 ,……., 𝑥 are k independent variables, 𝛽 is the regression coefficients and 𝜖 is represented as the error term. For multiple observations, we can write as 𝑦 = 𝛽 +𝛽 𝑥
+𝛽 𝑥
+⋯+𝛽 𝑥
+𝜖
𝑦 = 𝛽 +𝛽 𝑥
+𝛽 𝑥
+ ⋯+ 𝛽 𝑥
+𝜖
𝑦 = 𝛽 +𝛽 𝑥 + 𝛽 𝑥 +⋯+𝛽 𝑥 +𝜖 𝑦 = 𝛽 +𝛽 𝑥
+𝛽 𝑥
+⋯+𝛽 𝑥
+𝜖
These equations can be represented in matrix form as follows (2)
𝑦 = 𝑋𝛽 + 𝜖 where, 𝑦 𝑦 y= ⋮ 𝑦
1 𝑥 1 𝑥 𝑋= ⋮ ⋮ 1 𝑥
𝑥 𝑥 ⋮ 𝑥
⋯ 𝑥 ⋯ 𝑥 ⋯ ⋮ ⋯ 𝑥
𝛽 𝛽 𝛽= ⋮ 𝛽
𝜖 𝜖 𝜖= ⋮ 𝜖
The matrix X and y represents the historical data of all independent and dependent variables respectively where as using the least square method, 𝛽 of (1) can be calculated by the following equation: 𝛽 = (𝑋 𝑋) 𝑋
𝑦
(3)
From above known regression coefficient 𝛽, output, i.e. forecasted solar irradiance can be calculated from the multiple linear regression model as below: 𝑦 = 𝑋𝛽
(4)
For the measured historical data of various parameters, 𝑦 is the forecasted value of y and the difference of those two is the error in forecasting. After collecting future independent variable matrix 𝑋 , the forecasted independent variable 𝑦 is calculated as below. 𝑦 =𝑋 𝛽
(5)
Dependent upon the number of independent variables and a number of historical datasets, i.e. for a big set of data, number of rows are huge in X and y increasing the complexity in the linear regression which results in time consuming and complex solving. Hence multi-core parallel processing is used for big matrix transpose, multiplication and inverse operations which have been used in [24-26]. In the present work, five independent variables such as time of day, temperature, wind speed, humidity and atmospheric pressure and the dependent variable as the forecasted solar irradiance; the job of data handling and processing has been done for reducing the complexity of parallel processing, using the Matlab software. Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 125 – 135
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
131
The step by step procedure for calculating the mathematical expression between independent and dependent parameters is given below: Step 1: Measure and log the data of the independent and dependent variables and develop y and X matrices. For forecasting the dependent variable, read future data of independent variables only. Calculate correlations and exclude unrelated variables. Step 2: Use (3) to derive regression coefficients 𝛽 using Matlab software. Step 3: Forecast dependent variable using regression coefficients 𝛽 and future independent variables using (5) depending on the requirement. 3.3. Model validation and error computation Many methods for error calculation have been applied to the forecasting techniques for the need of evaluating the forecasted value. The most used methods include mean bias error (MBE), mean absolute error (MAE) root mean square error (RMSE), normalized root mean square error (nRMSE), mean square error (MSE) and Mean Absolute Percentage Error (MAPE) [3]. The input for the methods is the forecasted value of the parameter at a particular time calculated using a set of data along with the other input which is the real or actual value of the parameter w.r.t. the given time. In the presented work, the performance evaluation of the developed fuzzy system and multilinear regression approach for forecasting of solar irradiance is determined by calculating Mean Bias Error (MBE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) given by (6) to (8). MBE = ∑
(6)
[Y − Y ]
RMSE =
∑
MAPE =
∑
(7)
(Y − Y )
(8)
Where, Y is the predicted or the forecasted value and Y is the actual value measured in real time and N is the size of the test data set. The choice of method depends upon the decision maker.
100
40
80 Humidity (%)
50
60
30
10
20
0
0 Time steps (10minute interval) Day 1 Day 2 Day 3
(a)
1 38 75 112 149 186 223 260 297 334 371 408 445 482 519 556 593 630
40
20
1 38 75 112 149 186 223 260 297 334 371 408 445 482 519 556 593 630
Temperature (°C)
4. CASE STUDY, RESULTS AND ERROR EVALUATION 4.1. Case study: Location selection and data measurement With the availability of various measurement devices, the real time data for 3 complete days have been monitored and logged. For the purpose of forecasting and its benefit in the planning of a microgrid, the location of TIET, Patiala (Punjab), India with longitude and latitude (30.356008, 76.372142) has been considered. Figure 5 shows the variation in recorded parameters with the course of time and the solar irradiance respectively. For 3 consecutive days, the data for different parameters including relative temperature, humidity, wind speed, atmospheric pressure and solar irradiance were recorded using the above mentioned devices. Figure 5 (a) to 5 (c) represents the variation in temperature, humidity and wind speed from morning to evening for 3 days, whereas Figure 5 (d) shows uncertain solar irradiance dependent upon various input parameters for the same time.
Time steps (10minute interval) Day 1
Day 2
Day 3
(b)
Solar irradiance forecasting using fuzzy logic and multilinear regression approach … (Sahil Mehta)
ISSN: 2252-8814 1200 Solar Irradiance (W/m2 )
Wind Speed (m/s)
5
1000
4
800
3
600
2
400
1
200 0
1 38 75 112 149 186 223 260 297 334 371 408 445 482 519 556 593 630
0 Time steps (10minute interval) Day 1 Day 2 Day 3
1 38 75 112 149 186 223 260 297 334 371 408 445 482 519 556 593 630
132
Time steps (10minute interval) Day 1 Day 2 Day 3
(c)
(d)
Figure 5. Plot for various parameters (a)Temperature (b) Humidity (c) Wind speed (d) Solar irradiance 4.2. High dimensional fuzzy logic based approach Corresponding to the five input parameters; time of day, temperature, humidity, wind speed and atmospheric pressure, high dimensional fuzzy logic systems have been developed. Considering the linguistic variables of each parameter the calculated number of rules are 1200, out of which sample rules are given in Table A2 (submitted as supplementary material). For the results of the developed system, the input values: time of day- evening hours, temperature 31.6 °C, wind 2.61 m/s, humidity 41.9% and atmospheric pressure 1018.1 has been considered as shown in Figure 6.
Figure 6. Rule viewer of fuzzy logic approach The rule viewer window for the designed high order fuzzy logic system shows the crisp value of all input variables and its corresponding output i.e. forecasted solar irradiance. Thus, the result obtained from the developed system, which is forecasted solar irradiance for a particular value of time of day, temperature, humidity, wind speed and atmospheric pressure; is given by 275 W/m2 whereas; the actual solar irradiance observed was 273 W/m2. 4.3. Multilinear regression based forecasting From Table 1, the (9) to (15) represents the mathematical expressions for different time intervals of the day, i.e. very short term time interval based forecasting represented by (9) to (13) where the complete day has been divided into 4 time intervals; 07.00AM to 10.00AM, 10.00AM to 01.00PM, 01.00PM to 04.00PM and 04.00PM to 06.00PM. The advantage of such very short time interval is high forecasting accuracy. Also, (13) and (14) represent short term forecasting where the complete day from 07.00AM to 06.00PM has been divided in 2 intervals. Similarly (15) represents long term based forecasting expression. Figure 7 shows the forecasting results of the proposed methodologies with both the fuzzy and multilinear regression approach. Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 125 – 135
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
133
Table 1. Output of MLR approach for various time intervals Time interval 7.00AM - 10.00AM 10.00AM - 1.00PM 1.00PM - 4.00PM 4.00PM - 6.00PM 7.00AM – 1.00PM 1.00PM – 6.00PM 7.00AM– 6.00PM
Proposed equation 6182.467+(3.763*Ti)+(2.779*Tp)+ (0.978*H)-(6.139*W)-(6.16*P) 2568.037+(1.12* Ti)-(0.303* Tp)- (0.686*H)-(5.725*W)-(1.911*P) 141.528-(2.636* Ti)+(2.38* Tp)- (1.727*H)+(5.588*W)+(1.723*P) -7435.73-(3.669* Ti)-(5.671 Tp)-(2.207*H)+(3.283*W)+(9.969*P) 19890+(1.553* Ti)+(16.46* Tp)-(1.441*H)+(37.121*W)-(19.684*P) 1214-(2.966* Ti)+(2.011* Tp)-(1.797*H)+(22.172*W)+(0.761*P) 68146.894-(1.074* Ti)+(53.995*Tp)-(5.134*H)-(1.328*W)-(67.469*P)
1500 Solar Irradiance (W/m2 )
Solar Irradiance (W/m2)
1200 1000
1000
800 600
500
400 200 0
1 13 25 37 49 61 73 85 97 109 121 133 145 157 169 181 193
1 13 25 37 49 61 73 85 97 109 121 133 145 157 169 181 193
0
Time steps (10minute interval) Actual Irradiance Fuzzy Forecasted
Time steps (10minute interval) Actual Irradiance
(a)
(b) 1500 Solar Irradiance (W/m2 )
1500
1000
1000
500
500
1 14 27 40 53 66 79 92 105 118 131 144 157 170 183 196
0 Time steps (10minute interval) Actual Irradiance
0 1 13 25 37 49 61 73 85 97 109 121 133 145 157 169 181 193
Solar Irradiance (W/m2 )
Eq. No. (9) (10) (11) (12) (13) (14) (15)
-500
Time steps (10minute interval)
Actual Irradiance
(c)
MLR forecasted Long Term
(d)
Figure 7. Plot for forecasted and actual irradiance (a)Actual vs. Fuzzy output (b) Actual vs. MLR Very short interval output (c) Actual vs. MLR short interval output (d) Actual vs. MLR long interval output Figure 7 (a) represents the plot for the fuzzy logic based model with the actual output, i.e. solar irradiance, Figure 7 (b), Figure 7 (c) and Figure 7 (d) shows the results for the multilinear based forecasting approach. The analysis of the results shows almost overlapped plots for the fuzzy based system (Figure 7a); whereas with an increase in the time interval of forecasting, the variation between actual and forecasted irradiance increases, compared to Figure 7 (b), as shown in Figure 7 (d). 4.4. Error evaluation Using (6) to (8), error analysis in the proposed fuzzy logic system and multilinear regression approach has been presented in Table 2. Table 2. Error evaluation of proposed methodologies Error evaluation
Fuzzy
MBE RMSE MAPE (%)
-1.395 10.011 1.703
MLR EQ. 9 -0.910 28.258 10.536
MLR EQ. 10 6.950 26.236 2.583
Proposed methodology MLR EQ. MLR EQ. 11 12 16.028 3.316 32.786 29.166 3.952 14.746
MLR EQ. 13 10.806 50.003 12.461
MLR EQ. 14 0.158 31.048 8.194
MLR EQ. 14 0.690 104.671 22.710
Solar irradiance forecasting using fuzzy logic and multilinear regression approach … (Sahil Mehta)
134
ISSN: 2252-8814
The results show very low RMSE and MAPE error in the flexible fuzzy approach in comparison to the multilinear regression approach. Also, dividing the total time of the day in various time intervals, i.e. very short term (MLR EQ. 9 to MLR EQ. 12), short term (MLR EQ. 13, MLR EQ. 14) and long term interval (MLR EQ. 15); the analysis shows less error in very short term based forecasting compared to long term forecasting. Thus, the future work will focus on the expansion of fuzzy logic in the forecasting of renewable energy sources and comparing it with an artificial neural network approach taking all the basic meteorological parameters in consideration. 4.5. Power calculation For the purpose of calculating the power output of solar PV array, (9) shows the relation between all the major parameters, i.e. at both Standard Test Condition (STC) and the forecasted irradiance [27, 28]. P,
=Y f
, ,
[1 + α (T − T )]
(9)
Where, P , is forecasted power output in kW, Y is the rated capacity of the PV array, i.e. the power output of the panel under standard test conditions in kW, f is the derating factor of the system in %, G , is the forecasted solar irradiance at time T in kW/m2, G , is solar irradiance at STC (1000 W/m2), α is the temperature coefficient of power in %/°C, T and T represents the cell temperature and reference temperature in °C. Based on (9), Figure 8 shows the comparison plot of power output for different cases which are; actual irradiance, fuzzy forecasted irradiance and MLR forecasted irradiance. As per the rated standard test conditions, Y is taken as 25kW, f is taken as 0.66, T is taken as 25 °C and temperature coefficient of power i.e. α as 0.0167.
Figure 8. Plot for forecasted power (dependent upon actual, fuzzy and MLR solar irradiance) 5.
CONCLUSION In this paper, considering relative temperature, relative humidity, wind speed, time of the day and atmospheric pressure as independent uncertain input variables, a high dimensional fuzzy logic system along with multilinear regression approach for short term solar irradiance forecasting has been proposed. For the purpose of real time data measurement of parameters at the set location, i.e. 30.356008, 76.372142 at the Library lawns, Thapar Institute of Engineering and Technology, Patiala (India), various devices like Pyranometer, anemometer and barometer devices have been used. The results of the proposed fuzzy logic based system dependent upon the designed rule base of 1200 rules show crisp value for forecasted solar irradiance with an error of 10.011 RMSE and 1.703% MAPE in comparison to the real time observed solar irradiance. For the multilinear regression based approach, different mathematical equations dependent upon the time intervals have been produced, which shows the least error in case of very short term forecasting. Comparing the error analysis of multilinear regression based approach with the fuzzy logic based approach shows the later to be more acceptable. Thus, depending on the forecasted values of solar irradiance and the standard test conditions of solar panel, solar energy generation and the variations present in it has been shown. REFERENCES [1]
E. Lorenz, T. Scheidsteger, J. Hurka, D. Heinemann, and C. Kurz, "Regional PV power prediction for improved grid integration," Progress in Photovoltaics: Research and Applications, vol. 19, no. 7, pp. 757–771, 2011.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 125 – 135
Int. J. of Adv. in Appl. Sci. [2] [3] [4] [5] [6] [7] [8] [9] [10]
[11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28]
ISSN: 2252-8814
135
S. Chakraborty, M. D. Weiss, and M. G. Simoes, "Distributed intelligent energy management system for a singlephase high-frequency AC microgrid," IEEE Transaction on Industrial Electronics, vol. 54, no. 1, pp. 97–109, 2007. D. Neves, M. C. Brito, and C. A. Silva, "Impact of solar and wind forecast uncertainties on demand response of isolated microgrids," Renewable Energy, vol. 87, pp. 1003–1015, 2016. E. Duverger, C. Penin, P. Alexandre, F. Thiery, D. Gachon, and T. Talbert, "Irradiance forecasting for microgrid energy management," 2017 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT-Europe), Torino, pp. 1-6, 2017. E. Corsetti, A. Guagliardi, and C. Sandroni, "Recurrent neural networks for very short term energy resource planning in a microgrid," Mediterranean Conference on Power Generation, Transmission, Distribution and Energy Conversion (MedPower 2016), Belgrade, pp. 1-9, 2016. A. Shakya et al., "Solar Irradiance Forecasting in Remote Microgrids Using Markov Switching Model," in IEEE Transactions on Sustainable Energy, vol. 8, no. 3, pp. 895-905, 2017. R. Darbali-Zamora, C. J. Gómez-Mendez, E. I. Ortiz-Rivera, H. Li, and J. Wang, "Solar irradiance prediction model based on a statistical approach for microgrid applications," 2015 IEEE 42nd Photovoltaic Specialist Conference (PVSC), New Orleans, LA, pp. 1-6, 2015. S. I. Sulaiman, T. K. Abdul Rahman, I. Musirin, and S. Shaari, "Artificial neural network versus linear regression for predicting Grid-Connected Photovoltaic system output," 2012 IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems (CYBER), Bangkok, pp. 170-174, 2012. A. D. Orjuela-Cañón, J. Hernández, and C. R. Rivero, "Very short term forecasting in global solar irradiance using linear and nonlinear models," 2017 IEEE Workshop on Power Electronics and Power Quality Applications (PEPQA), Bogota, pp. 1-5, 2017. K. R. M. Supapo, R. V. M. Santiago, and M. C. Pacis, "Electric load demand forecasting for Aborlan-Narra-Quezon distribution grid in Palawan using multiple linear regression," 2017IEEE 9th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment and Management (HNICEM), Manila, pp. 1-6, 2017. Y. Saber and A. K. M. R. Alam, "Short term load forecasting using multiple linear regression for big data," 2017 IEEE Symposium Series on Computational Intelligence (SSCI), Honolulu, HI, pp. 1-6, 2017. N. Jaisumroum and J. Teeravaraprug, "Forecasting uncertainty of Thailand's electricity consumption compare with using artificial neural network and multiple linear regression methods," 12th IEEE Conference on Industrial Electronics and Applications (ICIEA), Siem Reap, pp. 308-313, 2017. Xiangyun Qing and Yugang Niu, "Hourly day-ahead solar irradiance prediction using weather forecasts by LSTM," Energy, vol. 148, Pp. 461-468, 2018. O. Ceylan, M. Starke, P. Irminger, B. Ollis, and K. Tomsovic, "A regression based hourly day ahead solar irradiance forecasting model by labview using cloud cover data,” in International Conference on Electrical and Electronics Engineering (ELECO), Bursa, pp. 406-410, 2015. M. Abuella and B. Chowdhury, "Solar power probabilistic forecasting by using multiple linear regression analysis," Southeast Con 2015, Fort Lauderdale, FL, pp. 1-5, 2015. U. Nalina, V. Prema, K. Smitha, and K. U. Rao, "Multivariate regression for prediction of solar irradiance," International Conference on Data Science & Engineering (ICDSE), Kochi, pp. 177-181, 2014. L. Suganthi, S. Iniyan, and A. A. Samuel, "Applications of fuzzy logic in renewable energy systems – A review," Renewable Sustainable Energy Reviews, vol. 48, pp. 585–607, 2015. S. X. Chen, H. B. Gooi, and M. Q. Wang, "Solar radiation forecast based on fuzzy logic and neural networks", Renewable Energy, vol. 60, pp. 195-201, 2013. R. S. Boata and P. Gravila, "Functional fuzzy approach for forecasting daily global solar irradiation,” in Atmospheric Research, vol. 112, pp. 79–88, 2012. M. Rizwan, M. Jamil, S. Kirmani, and D. P. Kothari, "Fuzzy Logic based Modelling and Estimation of Global Solar Energy using Meteorological Parameters," in Energy, vol. 70, pp. 685-691, 2014. A. Chugh, P. Chaudhary, and M. Rizwan, "Fuzzy logic approach for short term solar energy forecasting," in Annual IEEE India Conference (INDICON), New Delhi, pp. 1-6, 2015. M. B. Mbarek and R. Feki, "Using fuzzy logic to renewable energy forecasting: a case study of France", in International Journal of Energy Technology and Policy, vol. 12, no. 4, pp. 357-376, 2016. D. Saez, F. Avila, D. Olivares, C. Canizares, and L. Marin, "Fuzzy prediction interval models for forecasting renewable resources and loads in microgrids," in IEEE Transactions on Smart Grid, vol. 6, no. 2, pp. 548-55, 2015. H. S. Hippert, C. E. Pedreira, and R. C. Souza, "Neural networks for short-term load forecasting: a review and evaluation," in IEEE Transactions on Power Systems, vol. 16, no. 1, pp. 44-55, 2001. C. Y. Tee, J. B. Cardell, and G. W. Ellis, "Short-term load forecasting using artificial neural networks," in North American Power Symposium, Starkville, MS, USA, pp. 1-6, 2009. K. Liu et al., "Comparison of very short-term load forecasting techniques," in IEEE Transactions on Power Systems, vol. 11, no. 2, pp. 877-882, 1996. S. Arshdeep and P. Basak. "The present energy scenario and need of microgrid in India," Foundations and Frontiers in Computer, Communication and Electrical Engineering: Proceedings of the 3rd International Conference, India, pp. 313-317, 2016. Q. Deng, X. Gao, H. Zhou, W. Hu, "System modeling and optimization of microgrid using genetic algorithm,"2011 2nd International Conference on Intelligent Control and Information Processing, Harbin, pp. 540-544, 2011.
Solar irradiance forecasting using fuzzy logic and multilinear regression approach … (Sahil Mehta)
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 136~142 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp136-142
136
Solar panel monitoring and energy prediction for smart solar system Isha M. Shirbhate, Sunita S. Barve School of Computer Engineering & Technology, MIT Academy of Engineering, India
Article Info
ABSTRACT
Article history:
Solar Energy is established as an alternative source of energy known as renewable energy. In a developing country like India, the perspective of Solar Energy is important, as it supports a limitless source of energy. Monitoring and prediction of photo-voltaic energy generation help to reduce the energy loss and empower to utilize more energy. Solar energy prediction is challenging as it depends on the fluctuating solar radiations and climate conditions. The problem statement is to monitor solar panels and predict energy generation for energy management procedure. In this paper, the Internet of Things and Machine Learning algorithms are used as a powerful tool for developing a smart solar system. The metro-logical data such as humidity, temperature and photovoltaic panel data is used as input to forecast solar power generation. For prediction, we examine time-series of solar energy data with Hidden Markov Model. This model considers the probabilistic correlation between previous values to next value in time-series. Experimental results shows that individual panel dead state is located successfully and timeseries based solar energy prediction emulate the actual power generation.
Received Sep 12, 2018 Revised Apr 5, 2019 Accepted May 5, 2019 Keywords: Internet of things (IoT) Prediction Solar energy Solar radiation Time-series
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: Isha M. Shirbhate, School of Computer Engineering & Technology, MIT Academy of Engineering, Alandi, Pune, Maharashtra, India. Email: shirbhate.isha@gmail.com.in
1.
INTRODUCTION Solar energy is the best promising natural-renewable energy source for a green future, it has a lesser ecological impact as compared to other usual energy sources. Oil, Coal, Gases are conventional but nonrenewable sources of energy, they will run out very soon if exploited at huge scale. Moreover, the burning of these fossil fuels increases the concentration of greenhouse gases results in global warming [1]. Solar energy does not affect greenhouse atmosphere. Currently, photovoltaic (PV) solar energy characterizes as the thirdlargest source of renewable energy subsequent to hydro and wind [2]. The global investment in photo-voltaic solar systems increased because of transformation in favorable government policies and a reduction in the cost of the PV component. These aspects give rewards for investment in solar systems which is more motivating. PV system faces some problems after installation like dirty surface, damaged connections, and junction box failure results in a reduction in all-over system performance. IoT is a topology of the linked smart sensors and software that allows the things to assemble and transmit data using the internet. IoT based appliance is very supportive for monitoring the system at cheap cost. Solar panel's energy is affected by environmental conditions. Using sensors we can directly collect details about environmental conditions like humidity and temperature of that location [3-4]. The real-time monitoring system using IoT shows the current status of generated energy, humidity, temperature, a fault at the panel level and details of dead panels [5]. A defect can be detected by collecting and analyzing each sensor's data. Electrical parameters observation generally creates a huge quantity of data. The collected data used by artificial Journal homepage: http://iaescore.com/online/index.php/IJAAS
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
137
intelligence for power forecasting principles [6]. Electrical solar power production is decentralized in nature and it's difficult to monitor a large amount of data [7]. Real-time research of solar system is a very complicated task, as it demands a precise PV emulator that may specifically reproduce the non-linear character of PV cells. It also conflicts realization cost, accuracy, efficiency, complexity, sensitivity with varying environmental conditions [8]. According to author Arun G. Phadkeet [9] by Wide Area Monitoring (WAM), it is possible to assemble dimensions remotely from different places of power systems. Monitoring PV panels perform supervision along with fault recognition. The fault finding process assists for the recognizing diagnosis of the solar plant system, which depends on the study of voltage and current factors anticipated from monitored data of a PV [10]. This facilities improves WAM to modern energy system protection. The proper management meets a new scheme which gives an advance result of the adaptive system. To develop a smart solar plant system, it is essential to get accurate PV power forecasting results for efficient energy management. The messy nature of ecological conditions and the surprising weather circumstances makes accurate solar power prediction enormously difficult [11]. Prediction of solar power is straightforwardly connected to the weather predictions. To make well-organized consumption of energy system, it is helpful to forecast the reports of solar energy generation. The exact forecasting of solar irradiance digression enhances the quality of progression with civilizing PV energy supervision circumstances [12]. Maximum Power Point Tracking (MPPT) for solar systems assist prediction by measuring and estimating voltage and current. The Adaptive Neuro-Fuzzy (ANFIS) model utilizes PV cells temperature and solar radiation with a wavelet denoising model to obtain filtered constraints. MATLAB is easy to revise the processing of the MPPT [13]. The problem statement of the proposed system is to monitor solar panels and predict the future energy generation using Hidden Markov Model algorithm for energy management in IoT base environments. The objectives of the proposed system are as follows: to develop a real-time solar panel monitoring system for more energy utilization; and to develop a prediction model for solar energy management procedure. 2.
SYSTEM ARCHITECTURE The proposed smart solar system starts with monitoring panel level PV plant. The solar energy monitoring system is implemented by using IOT. Figure 1 shows the system architecture of a proposed system. We can observe the solar panel and multiple sensors and PV panels are connected with the middleware electronic board that is Raspberry Pi 3. That board is a central observing structure which collects data from sensors. A temperature and humidity of the location are sensed by sensors and power values are sensed by PV panel. Monitoring system displays the current status of temperature, humidity and generating energy power at certain time intervals. The system displays the entire information on the web pages and subsequently, it stores data on the cloud.
Figure 1. Architecture diagram of proposed system Solar panel monitoring and energy prediction for smart solar system (Isha M. Shirbhate)
138
ISSN: 2252-8814
Prediction system works on historical data of solar power collected by monitoring system. To make advance adaptive forecast system we apply a Hidden Markov Model (HMM) which works as time-series prediction. The system states definition and transition probabilities for constructing HMM. The proposed method gives an accurate prediction of solar energy generation. 3. RESEARCH METHOD 3.1. Monitoring system The monitoring system measure and display the collected data by sensors and solar panels. We monitored temperature, humidity and generating solar power using smart devices. It is essential for a monitoring system that the customers must know the potential and practical boundaries of devices, operational deviations and objective of data analysis. As we know python is a largely used, advanced, interpreted, highlevel, dynamic programming language, so we use python programming to complete this task. Monitoring constraints are then uploaded on the cloud using a python program. Flask works as a light-weight framework, which is written in Python. The Flask uses stretchy cast off to read the sensor values. We consider the DHT11 sensor that shows temperature with humidity as weather information from real-time situations. The system can work on the multiple solar panels which had the heterogeneous configuration. 3.2. Prediction system In the proposed system, we work on HMM for finding the most probable time-series. The time-series prediction is calculated by estimating probability of state transition and observation sequences for the predetermined state which gives precision in prediction. This method takes historical records of power as a input. Markov chain property says that the possibility of every subsequent state depends just on earlier state [14, 15]. The HMM have 𝑁 states, known as 𝑠 , 𝑠 , … , 𝑠 where discrete time-period 𝑡 = 0,1, … On each timeperiod, the system is in the state which is freely available called 𝑁 states. These states with time 𝑡 is identify like 𝑄 , by 𝑄 ∈ {𝑠 , 𝑠 , … , 𝑠 }. At each and every time-period, the state 𝑄 form, one resulting symbol according to Observation and Emission Probabilities allocation. We find hidden and observed states called as a state probability vector 𝜫, inter-state transition probability matrix 𝑨 and emission probability matrix 𝑩. Markov chain can translate calculated data to the various states, so the series of energy generation values is translated into the number of states. Therefore, State transition probabilities are {𝑠 , 𝑠 , … , 𝑠 } It signifies probability of one state to another state as a state transition probability matrix 𝐴= 𝑎
, 𝑎 = Р(𝑠 |𝑠 )
Here, 𝑖 represents a number of rows and 𝑗 number of columns. The observation probability allocation signify for state 𝑗 is 𝑉𝑚, 𝑚 = 1, … . , 𝑀 in the state 𝑠 𝐵 = 𝑏 (𝑣 ) , 𝑏𝑖(𝑣 ) = Р(𝑣 |𝑠 ) As quantity of symbols are two thus 𝑚 = 2 With vector initial probability matrix Π = (Π ), Π = Р(𝑠 ) At last, Model is shown by 𝜆 = (𝐴, 𝐵, Π). The purpose of energy prediction is examining a Hidden Markov Model by considering observed time- series. Maximum likelihood estimation method estimate 𝜆∗ which maximizes a likelihood of training series, 𝑂𝑇𝑠 = {𝑂𝑇𝑠 } for improvement in Р(𝑂𝑇𝑠|𝜆). Autonomous and supremely spotted samples of, 𝑂𝑇𝑠 = {𝑂𝑇𝑠 } are anxious from the probability allocation Р(𝑂𝑇𝑠|𝜆). The objective is to get a value of 𝜆 which originates 𝑂𝑇𝑠 to Р(𝑂𝑇𝑠|𝜆) as likely as probable. 𝑂𝑇𝑠 ≅ Р(𝑂𝑇𝑠|𝜆) The probability variation method is precise with identified 𝜆 = (𝐴, 𝐵, Π). Recalculate 𝜆 for the entire steps with likelihood probability of frequency of state 𝑠 following state 𝑠 . This variation step is reiterate till convergence for the time-interval of which Р(𝑂𝑇𝑠|𝜆) not reduces. Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 136 – 142
Int. J. of Adv. in Appl. Sci. Steps 1: 2: 3: 4: 5:
ISSN: 2252-8814
139
Algorithm Observe Time-Series in Sequence 𝑂𝑇𝑠 = {0′𝑠, 1′𝑠} Named as different observation symbols for states N 𝑉 = {0, 1} Calculate approximately values of Markov model to maximize probability Р(𝑂𝑇𝑠|𝜆) 𝜆 = (𝐴, 𝐵, Π) Predict Energy generation depends on past observation time-series Р(𝑂𝑇𝑠, 1|𝜆) < Р(𝑂𝑇𝑠, 0|𝜆) 𝑡ℎ𝑒𝑛 𝑆(𝑡 + 1) = 0 Р(𝑂𝑇𝑠, 1|𝜆) ≥ Р(𝑂𝑇𝑠, 0|𝜆) 𝑡ℎ𝑒𝑛 𝑆(𝑡 + 1) = 1 Produce predicted observation time-series for 𝑡 = 𝑇 + 1 𝑡𝑜 2𝑇 𝑃𝑂𝑇𝑠 = {𝑂𝑇𝑠} using {𝑂𝑇𝑠}
The algorithm initially examines all the records from the dataset and spites every characteristic. At step 2, it mentions each attribute as a unique symbol i.e. 0 𝑠 and 1 𝑠. The sampling approach work with the help of HMM model in step 3. Then next we do prediction of observation time-series of moving ratio of input. Lastly, in conclusion step 5 gives the prediction output. This process is called as an observable Markov Model. It gives a set of number of states at the specific time period, where every state communicates to the next state [16]. Thus the presented model seems preventive for several problems. We can expand the idea of a Markov Model to embrace the sets while observation sequence is a probabilistic method of states. HMM is a vital version of Markov method, yet HMM is a self-governing procedure. Here, states communicate to a group of same probability allocation of observation. While working on this algorithm we found two challenging problems that are mention. 3.2.1. Likelihood problem The first problem is to calculate the likelihood sequence of a distinct observation sequence that is a new observation sequence and a set of models. It finds a model which gives the best sequence. The best possible sequence linked with known observations using state sequence and Observed Time-Series Р(𝑂𝑇𝑠|𝑄) = Π Р(𝑂𝑇𝑠| 𝑠 ) ⨯ Π Р( 𝑠 | 𝑠
)
We used an efficient algorithm called the forward algorithm. The forward algorithm is a kind of dynamic programming algorithm. It computes the observation probability by summing over the probabilities of all possible hidden-state paths that could generate the observation sequence. 3.2.2. Decoding problem HMM model, which contains hidden variables, the challenge is to determine which sequence is the underlying source of some sequence of observations. It is called the decoding task. The decoder finds the besthidden sequence, this problem solved by the Viterbi algorithm. It is possible to alter the model constraints 𝜆 = (𝐴, 𝐵, Π) for maximize Р(𝑂𝑇𝑠|𝜆). The Viterbi Algorithm score finest from the a particular path at initial time 𝑡 observations, trim with state 𝑠 , presented as: 𝛿(𝑖) = 𝑚𝑎𝑥Р(𝑠 , 𝑠 , … , 𝑠 = 𝑖, 𝑂𝑇𝑠|𝜆) 4.
RESULTS AND DISCUSSION The results are split up into three parts. First part is a performance evaluation of the monitoring system. The second part is an analysis and comparison of the prediction model and the third part is a case-study on a proposed method. 4.1. A solar plant monitoring system The system shows a status dashboard of monitoring panels with energy generation level. It monitors the temperature in degree celsius and humidity in the percentage. Solar energy differs by environmental seasons [3], Figure 2 shows graphs for hourly temperature and solar energy generation in sunny, cloudy and rainy days. The system represents graphs for analyzing the performance in a standard method to provide the understandable view of information to the users. The dashboard is accessible to the users from anywhere and anyplace. If the one-time connection to the server is successful, the status of the system is constantly passed to the web server for monitoring the parameters of solar Power Conditioning Unit (PCU). We can examine the intense values generated by solar PCU. The system is able to analyze past average ratio of energy generation monthly and yearly. In addition, the system also forecasts the panel's dead state and the error finding. Solar panel monitoring and energy prediction for smart solar system (Isha M. Shirbhate)
140
ISSN: 2252-8814
Rainy
Sunny
Cloudy
Rainy
Cloudy
1.5
80
Power(W)
Temperature(C)
Sunny 100
60 40 20
1 0.5
0
0
5 6 7 8 9 10 11 12 13 14 15 16 17 18
5 6 7 8 9 10 11 12 13 14 15 16 17 18
Time(Hours)
Time(Hours)
(a)
(b)
Figure 2. (a) Temperature and (b) Power in different weather conditions 4.2. Solar energy prediction We identified that the consequences of prediction are affected by the environmental situations and location of PV plant [6]. In the prediction model, we used an advanced Hidden Markov Model to predict solar energy generation. Considering the real-time statistics measured by the monitoring system, we get better results. Following Table 1 shows the actual values and predicted values, for few hourly sample cases. Accurateness of prediction is calculated with help of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). We compare the proposed algorithm with Linear Regression [17] as shown in Table 2. If rate among predicted and actual values are close to zero by RMSE, then we find betterquality output at low error. Figure 3 and Figure 4 show day wise and month-wise, actual with predicted PV power generation for different environmental conditions. Table 1. Different sample cases on hourly data Hour 1 2 3 4
Observation Seq 1-0-1-……..0 0-1-0-.…….1 1-1-0-.…….1 0-1-1-.…….0
States 2 2 2 2
Actual Output 1.44 1.50 1.82 1.55
Proposed Model 1.56 1.57 1.67 1.65
Table 2. Comparison between proposed method and another prediction method Sr. No. 1 2
Actual Predicted
5
MAPE 8.442 13.440
6
2 1
Power(W)
3
3 2 1
0
10 20 Time(day)
30 (a)
5 4 3 2 1
0
0
Actual
7
4
Power(W)
Power(W)
RMSE 0.642 1.501
Actual Predicted
5
4
Method Proposed Method Linear Regression
0
0
10 20 Time(day)
30 (b)
0
10 20 Time(day)
30 (c)
Figure 3. Actual and predicted PV power generation in (a) Cloudy, (b) Rainy and (c) Sunny month
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 136 – 142
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
141
Figure 4. Actual and predicted PV power generation for (a) Cloudy, (b) Rainy and (c) Sunny day 4.3. Case study The proposed methodology is tested on a database of the large-scale system, which is located at MIT Academy of Engineering, Alandi, Pune campus as shown in Figure 5. There are six distributed solar plants, one single panel is with 315watt peak capacity and whole solar system having the 435kW capacity. The system added with irradiance sensor, temperature sensor and wind speed sensor for monitoring the weather situations. We consider the hourly power data of one plant to predict the solar power. Figure 6 shows the actual and predicted power generation of the solar plant.
Figure 5. Large-scale solar plant at MIT Academy of Engineering, Alandi
Figure 6. Actual and predicted PV power generation of large-scale solar plant
Solar panel monitoring and energy prediction for smart solar system (Isha M. Shirbhate)
142
ISSN: 2252-8814
5.
CONCLUSION Solar energy is trustworthy and sustainable hence the utilization of this smart solar system is reliable, adequate and cost-efficient. The proposed smart solar system comprises both Monitoring and Prediction. We have used the Internet of Things (IoT) for monitoring by considering parameters like Temperature, Humidity and Solar Power. Internet of Things is providing practically expert methods that offers required outcome. The designed system monitors panel level PCU and anticipates the error findings. We can analyze the weekly or monthly performance of panels. On the other hand, Prediction uses Hidden Markov Model for forecasting the solar power. We have developed an hourly prediction system. Considering historical records, the proposed model is able to predict accurate power generation in time- series method. Numerical results show that the proposed model achieves better prediction accuracy than the simple Linear Regression model.
REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17]
I. Gherboudj and H. Ghedira, "Assessment of solar energy potential over the United Arab Emirates using remote sensing and weather forecast data," Renewable and Sustainable Energy Reviews, vol. 55:1210-1224, 2016. Moreno-Garcia, E. Palacios-Garcia, V. Pallares-Lopez, I. Santiago, M. Gonzalez-Redondo, M. Varo Martinez, and R. Real-Calvo, "Real-time monitoring system for a utility-scale photovoltaic power plant," Sensors, vol. 16(6), p. 770, 2016. S. R. Madeti and S. N. Singh, "Monitoring system for photovoltaic plants: A review," Renewable and Sustainable Energy Reviews, vol. 67, pp. 1180-1207, 2017. B. Ando, S. Baglio, A. Pistorio, G. M. Tina, and C. Ventura, "Sentinella: Smart monitoring of photovoltaic systems at panel level," IEEE Transactions on Instrumentation and Measurement, vol. 64(8), pp. 2188-2199, 2015. F. O. Hocaoglu and F. Serttas, "A novel hybrid (Mycielski-Markov) model for hourly solar radiation forecasting," Renewable Energy, vol. 108, pp. 635-643, 2017. S. Daliento, A. Chouder, P. Guerriero, A. M. Pavan, A. Mellit, R. Moeini, and P. Tricoli, "Monitoring, diagnosis, and power forecasting for photovoltaic fields: A review," International Journal of Photoenergy, 2017. C. Stegner, M. Dalsass, P. Luchscheider, and C. J. Brabec, "Monitoring and assessment of PV generation based on a combination of smart metering and thermographic measurement," Solar Energy, vol. 163, pp. 16-24, 2018. S. Senthilkumar, G. P. C., “Hidden Markov Model based channel selection framework for cognitive radio network,” Computers and Electrical Engineering, vol. 65, pp. 516-526, 2018. A. G. Phadke, P. Wall, L. Ding, and V. Terzija, "Improving the performance of power system protection using wide area monitoring systems," Journal of Modern Power Systems and Clean Energy, vol. 4(3), pp. 319-331, 2016. S. Silvestre, L. Mora-Lopez, S. Kichou, F. Sanchez-Pacheco, and M. Dominguez-Pumar, "Remote supervision and fault detection on OPC monitored PV systems," Solar Energy, vol. 137, pp. 424-433, 2016. C.Wan, J. Zhao, S. Member, and Y. Song, "Photovoltaic and solar power forecasting for smart grid energy management," Journal of Power and Energy Systems, vol. 1(4), pp. 38-46, 2015. S. Mohanty, P. K. Patra, S. S. Sahoo, and A. Mohanty, "Forecasting of solar energy with application for a growing economy like India: Survey and implication," Renewable and Sustainable Energy Reviews, vol. 78, pp. 539-553, 2017. A. Chikh and A. Chandra, "An optimal maximum power point tracking algorithm for PV systems with climatic parameters estimation," IEEE Transactions on Sustainable Energy, vol. 6(2), pp. 644-652, 2015. M. J. Sanjari and H. B. Gooi, "Probabilistic forecast of pv power generation based on higher order markov chain," IEEE Transactions on Power Systems, vol. 32(4), pp. 2942-2952, 2017. S. S. Barve, "Dynamic channel selection and routing through reinforcement learning in cognitive radio networks," IEEE International Conference on Computational Intelligence and Computing Research, 2012. W. J. I. Li and Jinbo Pedrycz, "Multivariate time series anomaly detection: A framework of hidden markov models," Applied Soft Computing Journal, vol. 60(27), pp. 229-240, 2017. N. Sharma, P. Sharma, D. Irwin, and P. Shenoy, "Predicting solar generation from weather forecasts using machine learning," Smart Grid Communications (SmartGridComm), 2011 IEEE International Conference, pp. 528-533, 2011.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 136 – 142
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 143~153 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp143-153
143
Performance evaluation and comparison of diode clamped multilevel inverter and hybrid inverter based on PD and APOD modulation techniques N. Susheela, P. Satish Kumar Department of Electrical Engineering, Osmania University, India
Article Info
ABSTRACT
Article history:
The popularity of multilevel inverters have increasing over the years in various applications without use of a transformer and has many benefits. This work presents the performance and comparative analysis of single phase diode clamped multilevel inverter and a hybrid inverter with reduced number of components. As there are some drawbacks of diode clamped multilevel inverter such as requiring higher number of components, PWM control method is complex and capacitor voltage balancing problem, an implementation of hybrid inverter that requires fewer components and less carrier signals when compared to conventional multilevel inverters is discussed. The performance of single phase diode clamped multilevel inverter and hybrid multilevel inverter for seven, nine and eleven levels is performed using phase disposition, alternate phase opposition disposition sinusoidal pulse width modulation techniques. Both the multilevel inverter are implemented for the above mentioned multicarrier based Pulse Width Modulation methods for R and R-L loads. The total harmonic distortion is evaluated at various modulation indices. The analysis of the multilevel inverters is done by simulation in matlab/simulink environment.
Received Aug 29, 2018 Revised Apr 1, 2019 Accepted May 8, 2019 Keywords: Diode clamped multilevel inverter Hybrid inverter Sinusoidal pulse width modulation Total harmonic distortion
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: N. Susheela, Department of Electrical Engineering, University College of Engineering, Osmania University, Hyderabad, Telangana, India. Email: nsusheela2007@yahoo.com
1.
INTRODUCTION The multilevel inverter (MLI) is now proven technology for medium/high voltage high- power applications such as such as marine drives, variable-speed drives, reactive power compensation, steel rolling mills and other applications. The concept of power conversion in multilevel inverters (MLI) is to synthesize a staircase voltage waveform from several lower voltage DC sources which approaches the sinusoidal wave with reduced harmonic distortion has got several advantages and have drawn tremendous interest in high power high voltage applications [1-4]. In case of multilevel inverters the semiconductors are wired to form series type connection so that the operation at higher voltages is possible. The switching losses and the switching frequency can be reduced by staggering the switching because the switches are not truly series connected. Conventional multilevel inverters include neutral point clamped (diode-clamped) inverter, flying capacitor (capacitor clamped) inverter and cascaded H-bridge inverter. The major drawback of multilevel inverters is the Journal homepage: http://iaescore.com/online/index.php/IJAAS
144
ISSN: 2252-8814
higher number of power semiconductor switches needed that complicates the overall system [5-8]. Using lower rated switches in the multilevel inverter can reduce the cost of active semiconductors compared to two level inverters. Associated gate drive circuits are required for each active semiconductor which increases the complexity. In [9], the symmetrical topology which is called as reversing voltage topology is implemented for single phase seven level inverter using phase disposition method. This multilevel inverter topology requires less number of components when compared to conventional multilevel inverters. A multilevel inverter with reversing voltage component has many advantages as the levels increase when compared to conventional multilevel inverters. The hybrid topology eliminates the diodes and capacitors that are used in diode clamped inverters, capacitors used in flying capacitor inverters and also reduces the switches and carrier signals required than in cascaded inverters, diode clamped, and flying capacitors inverters. An approach of utilizing high-power devices with low-switching-frequency reduces voltage distortion of output but has got current harmonics which is a major drawback [10-13]. There are asymmetrical methods of using different values of voltage source which requires more number of power switches and diodes with different rating. Some topologies suffer from the capacitor balancing problems. Whereas in case of hybrid MLI, the voltage sources used have equal values and has many advantages compared with the methods discussed above. It utilizes less number of switches and carriers and also operates the switching devices at line frequency. The different multicarrier PWM methods are reported to minimize total harmonic distortion (THD). Advanced MLI topologies have been proposed recently such as hybrid multilevel inverter, soft switching inverter and generalized MLI. These are extensively used in applications like FACTS, tractions and industrial drives [14-19]. The phase disposition (PD) and alternate phase opposition disposition (APOD) sinusoidal pulse width modulation (SPWM) methods are used to drive the single phase DCMLI and hybrid MLI for different levels. The general MLI inverter structure for various levels is shown in in Figure 1.
Figure 1. MLI inverter structure for two, three and n-levels 2.
MULTILEVEL INVERTERS The analysis of diode clamped multilevel inverter and hybrid inverter are discussed based on sinusoidal pulse width modulation techniques. 2.1. Diode clamped MLI In order to generate seven levels by SPWM, six carriers and a sinusoidal reference signal for modulator are needed for DCMLI. The arrangement of the carriers for APOD and PD techniques can be seen in Figure 2 and Figure 3.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 143 – 153
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
145
Figure 2. APOD PWM technique for single phase 7 level DCMLI
Figure 3. PD PWM technique for single phase 7 level DCMLI 2.2. Hybrid MLI The hybrid MLI has two stages in which one stage is level generation and other is polarity generation stage. The first stage requires high-frequency switches which produces the required positive levels. The output polarity can be obtained by the second stage that has low frequency switches. The hybrid MLI eliminates higher number switches that are required to generate output levels. The single phase 7 level DCMLI usiing SPWM requires 6 carriers, but three carriers are sufficient for hybrid MLI. The 7 level hybrid MLI requires only 3 carriers and a sinusoidal reference. Figure 4 and Figure 5 represents the carrier arrangement using APOD and PD SPWM techniques.
Figure 4. APOD PWM technique for single phase 7 level hybrid inverter
Perfomance evaluation and comparison of diode clamped multilevel inverter based… (N. Susheela)
146
ISSN: 2252-8814
Figure 5. PD PWM technique for single phase 7 level hybrid inverter 3.
IMPLEMENTATION OF DCMLI AND HYBRID MLI Figure 6 and Figure 7 depicts the simulation models for producing gating signals in a single phase 7 level DCMLI using APOD and PD SPWM techniques.
Figure 6. Model for gating signals of 7 level DCMLI using APOD method
Figure 7. Model for gating signals of 7 level DCMLI using PD method
Figure 8 shows the simulation model using APOD method for generating gating signals in level generation stage of 7 level hybrid MLI. The simulation model for producing gating signals in polarity generation stage of 7 level hybrid MLI is shown in Figure 9.
Figure 8. Model for gating signals in level generation part
Figure 9. Model for gating signals in polarity generation part
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 143 – 153
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
147
4.
RESULTS ANALYSIS The results obtained for DCMLI and hybrid MLI are eloberated for various loads. The performance of the inverters using LC filter are also discussed. The results obtained using APOD technique is presented below for 0.9 modulation index. 4.1. DCMLI for 7, 9 and 11 levels The performance of DCMLI is shown below for various levels using different loads. The results of DCMLI for 7 levels are depicted in Figure 10 to Figure 12.
Figure 10. Waveforms of voltage and current in 7 level DCMLI for R load
Figure 11. Waveforms of voltage and current in 7 level DCMLI for R load with filter
Figure 12. Voltage and current waveforms of 7 level DCMLI for R-L load
Perfomance evaluation and comparison of diode clamped multilevel inverter based… (N. Susheela)
148
ISSN: 2252-8814
The 9 level DCMLI results can be seen from Figure 13 to Figure 15 using APOD method. Figure 16 to Figure 18 presents the results of DCMLI for 11 levels.
Figure 13. Waveforms of voltage and current in 9 level DCMLI for R load
Figure 14. Waveforms of voltage and current in 9 level DCMLI for R load with filter
Figure 15. Waveforms of voltage and current in 9 level DCMLI for R-L load
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 143 – 153
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
149
Figure 16. Waveforms of voltage and current in 11 level DCMLI for R load
Figure 17. Waveforms of voltage and current in 11 level DCMLI for R load with filter
Figure 18. Waveforms of voltage and current in 11 level DCMLI for R-L load 4.2. Hybrid MLI for 7, 9 and 11 levels The performance of hybrid MLI are presented in this section for various levels at different loads. Figure 19 to Figure 21 shows the results for single phase 7 level hybrid MLI. Whereas the results for single phase 9 level hybrid MLI are shown from Figure 22 to Figure 24. The waveforms obtained for single phase hybrid MLI are shown from Figure 25 to Figure 27 for 11 levels.
Perfomance evaluation and comparison of diode clamped multilevel inverter based… (N. Susheela)
150
ISSN: 2252-8814
Figure 19. Waveforms of voltage and current in 7 level hybrid MLI for R load
Figure 20. Waveforms of voltage and current in 7 level hybrid inverter for R load with filter
Figure 21. Waveforms of voltage and current in 7 level hybrid MLI for R-L load
Figure 22. Waveforms of voltage and current in 9 level hybrid MLI for R load Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 143 – 153
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
151
Figure 23. Waveforms of voltage and current in 9 level hybrid MLI for R load with filter
Figure 24. Waveforms of voltage and current in 9 level hybrid MLI for R-L load
Figure 25. Waveforms of voltage and current in 11 level hybrid MLI for R load
Figure 26. Waveforms of voltage and current in 11 level hybrid inverter for R load with filter Perfomance evaluation and comparison of diode clamped multilevel inverter based… (N. Susheela)
152
ISSN: 2252-8814
Figure 27. Waveforms of voltage and current in 11 level hybrid MLI for R-L load 4.3. Comparison of DCMLI and hybrid MLI Table 1 to Table 3 summarizes the comparative analysis of single phase diode clamped MLI and hybrid MLI for seven, nine and eleven levels at various modulation indices (ma) using PD and APOD SPWM techniques. It is clear that the THD decreases as the modulation index is increased. At any level, the THD of hybrid inverter is less when compared to DCMLI. The THD is further reduced using LC filter. Table 1. THD (%) comparison using PD method for R load Number of Levels 7 level
9 level
11 level
Without Filter DCMLI Hybrid MLI 18.67 12.29 17.52 11.48 15.77 11.08 13.97 9.84 13.19 9.05 12.81 8.47 11.97 8.4 10.48 7.38 10.02 7.15 10.11 6.8 9.54 6.65 8.25 6.05
ma 0.85 0.9 0.95 1 0.85 0.9 0.95 1 0.85 0.9 0.95 1
DCMLI 3.97 3.83 3.53 3.28 3.55 3.42 3.18 3.24 2.5 2.84 2.57 3.12
Filter Hybrid MLI 1.13 1.05 1 0.92 0.88 0.87 0.85 0.8 0.79 0.8 0.83 0.77
Table 2. THD (%) comparison using APOD method for RL load Number of Levels 7 level
9 level
11 level
Without Filter DCMLI Hybrid MLI 18.5 12.21 17.58 11.56 15.74 11.12 13.86 9.7 13.01 8.96 12.88 8.53 12.02 8.37 10.46 7.44 9.77 7.34 10.22 6.93 9.75 6.52 8.38 6.02
ma 0.85 0.9 0.95 1 0.85 0.9 0.95 1 0.85 0.9 0.95 1
Filter DCMLI 3.18 3.01 2.93 2.79 2.85 2.7 2.78 2.82 2.54 2.96 2.61 2.84
Hybrid MLI 1.1 1.02 0.99 0.86 0.81 0.8 0.73 0.69 0.72 0.71 0.70 0.66
Table 3. Current THD (%) for R-L load 7-Level Hybrid DCMLI inverter 4.81 4.8 4.81 4.83 4.79 4.79 4.79 4.77
PD method 9-Level Hybrid DCMLI inverter 4.81 4.79 4.79 4.81 4.79 4.81 4.8 4.79
11-Level Hybrid DCMLI inverter 4.75 4.77 4.77 4.75 4.74 4.75 4.76 4.79
7-Level Hybrid DCMLI inverter 4.83 1.79 4.81 4.79 4.81 4.81 4.81 4.71
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 143 – 153
APOD method 9-Level Hybrid DCMLI inverter 4.81 4.82 4.82 4.82 4.82 4.83 4.81 4.83
11-Level Hybrid inverter 4.79 4.79 4.78 4.79 4.8 4.8 4.78 4.78
DCMLI
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
153
5.
CONCLUSION The single phase DCMLI and hybrid MLI are implemented for various levels at different modulation indices using PD and APOD SPWM control techniques.When compared to DCMLI, the hybrid MLI needs lesser number of high frequency switches for any number of levels. The hybrid MLI has many features compared to DCMLI in terms of the required switches, control requirements, cost, reliability and efficiency. The switches needed for various voltage levels for single phase hybrid MLI are lower compared to classical MLI’s. The hybrid MLI can be preferred for applications like STATCOM, HVDC and FACTS. The low rated dc sources are required for hybrid MLI so the photovoltaic arrays and fuel cells can be utilized. The THD of DCMLI and hybrid MLI are analyzed at various modulation indices. It is observed that the THD reduces as there is an increase in modulation index. Different types of pulse width modulation control strategies can also be used to hybrid multilevel inverters. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19]
Nabae, I. Takahashi and A. Akagi, “A new neutral-point clamped PWM inverter,” IEEE Trans. Ind. Appl., vol. 19, pp. 518–523, 1981. G. Mondal, K. Gopakumar, P. N. Tekwani, and E. Levi, “A reduced switch- count five-level inverter with commonmode voltage elimination for an open-end winding induction motor drive”, IEEE Trans. Ind. Electron., vol. 54(4), pp. 2344–2351, 2007. P. M. Bhagwat and V. R. Stefanovic, “Generalized structure of a multilevel PWM inverter,” IEEE Trans. Industry Applications, vol. 19(6), pp. 1057-1069, 1983. A. Rufer, “An aid in the teaching of multilevel inverters for high power applications,” Proc. Rec. IEEE PESC’95, pp. 347–352, 1995. L. M. Tolbert, F. Z. Peng, and T. G. Habetler, “Multilevel converters for large electric drives,” IEEE Trans. Ind. Appl., vol. 35(1), pp. 36–44, 1999. N. Susheela, P. Satish Kumar, and C. H. Reddy, “Performance Analysis of Four Level NPC and NNPC Inverters using Capacitor Voltage Balancing Method,” IEEE Uttar Pradesh Section International Conference on Electrical, Computer and Electronics Engineering, Varanasi, pp. 212-217, 2016. G. Carrara, S. Gardella, M. Marchesoni, R. Salutari, and G. Sciutto, “A new multilevel PWM method: a theoretical analysis,” IEEE Trans.Power Electronics., vol. 7(3), pp. 497–505, 1992. N. Susheela and P. Satish Kumar, “Performance Evaluation of Multicarrier Based Techniques for Single Phase Hybrid Multilevel Inverter using Reduced Switches,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 7, no. 3, pp. 676-686, September 2017. J. Rodriguez, J.-S. Lai, and F. Z. Peng, “Multilevel inverters: A survey of topologies, controls, and applications,” IEEE Trans. Ind. Electron., vol. 49(4), pp. 724–738, 2002. E. Najafi, A. H. M. Yatim, and A. S. Samosir, “A new topology -Reversing Voltage (RV) - for multilevel inverters,” IEEE 2nd International Power, Energy Conference (PECon 08), pp. 604-608, 2008. S. A. Gonzalez, M. I. Valla, and C. F. Christiansen, “Analysis of a cascade asymmetric topology for multilevel converters,” Proc. IEEE ISIE, pp. 1027–1032, 2007. N. Susheela and P. Satish Kumar, “Comparative Analysis of Carrier Based Techniques for Single phase Diode Clamped MLI and Hybrid Inverter with Reduced Components,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 7, no. 3, pp. 687-697, September 2017. E. Beser, B. Arifoglu, S. Camur, and E. K. Beser, “Design and application of a single phase multilevel inverter suitable for using as a voltage harmonic source”, J. Power Electron., vol. 10(2), pp. 138–145, 2010. G. M. Martins, J. A. Pomilio, S. Buso, and G. Spiazzi, “Three-phase low frequency commutation inverter for renewable energy systems,” IEEE Trans. Ind. Electron., vol. 53(5), pp. 1522–1528, 2006. Ehsan Najafi and Abdul Halim Mohamed Yatim, “Design and Implementation of a New Multilevel Inverter Topology,” IEEE Transactions on Industrial Electronics, vol. 59(11), pp. 4148-4154, 2012. E. Babaei, “Optimal topologies for cascaded sub-multilevel converters,” J. Power Electron, vol. 10, no. 3, pp. 251–261, 2010. C. Govindaraju and K. Baskaran, “Analysis and implementation of multiphase multilevel hybrid single carrier sinusoidal modulation,” J. Power Electron, vol. 10(4), pp. 365–373, 2010. R. Stala, “Application of balancing circuit for dc-link voltages balance in a single-phase diode-clamped inverter with two three-level legs,” IEEE Trans. Ind. Electron, vol. 58(9), 4185–4195, 2011. N. Susheela and P. Satish Kumar, “Performance Analysis of FPGA based Diode Clamped Multilevel Inverter Fed Induction Motor Drive using Phase Opposition Disposition Multicarrier Based Modulation Strategy,” International Journal of Power Electronics and Drive System (IJPEDS), vol. 8, no. 4, pp. 1512-1523, December 2017.
Perfomance evaluation and comparison of diode clamped multilevel inverter based… (N. Susheela)
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 154~156 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp154-156
154
Solar energy storage and release application of water - phase change material - (SnO2-TaC) and (SnO2–SiC) nanoparticles system 1
Farhan Lafta Rashid, 2Aseel Hadi, 3Ammar Ali Abid, 4Ahmed Hashim 1
Department of Petroleum, College of Engineering, University of Kerbala, Iraq of Ceramics and Building Materials, College of Materials, University of Babylon, Iraq 3Water Resources Engineering College, University of Al-Qasim Green, Iraq 4Department of Physics, College of Education for Pure Sciences, University of Babylon, Iraq
2Department
Article Info
ABSTRACT
Article history:
The thermal energy storage and release application of water- phase change material - (SnO2-TaC) and (SnO2–SiC) nanoparticles system has been investigated for cooling and heating applications. The water - polyethylene glycolwith (SnO2-TaC) and (SnO2–SiC) nanoparticles have been used. The results showed that the melting and solidification times for storage and release of thermal energy of water - polyethylene glycoldecrease with increase in (SnO2-TaC) and (SnO2–SiC) nanoparticles concentrations. The melting and solidification times decrease with increasing of TaC nanoparticles concentrations to water-polyethylene glycol/SnO2 nanofluid and SiC nanoparticles concentrations to water-polyethylene glycol/SnO2 nanofluid.
Received Sep 15, 2018 Revised Apr 11, 2019 Accepted May 11, 2019 Keywords: Energy storage Meltingtime Polyethylene glycol Solidification time
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: Farhan Lafta Rashid, Department of Petroleum, College of Engineering, University of Kerbala, P.O.Box 1125, Kerbala, Iraq. Email: engfarhan71@gmail.com
1.
INTRODUCTION The extensively used thermal energy storage materials are phase change materials (PCMs) because of their ability of storing and releasing considerable amounts of energy (thermal) during the processes of phase change melting and solidification. Because of environmental regards and ascent of fossil fuels cost, PCMs are attractive increasingly spirited for many applications such as, battery thermal management, electronic cooling, buildings space heating and cooling. In the current study, a cylindrical geometry is selected as it is considered most favorable for heat exchangers, due to its acceptable efficiency in a low volume. The disadvantage of PCMs is low thermal conductivity, which decreases rates of melting and solidification [1]. Phase change materials and thermal energy storage become increasingly substantial subjects during the last two decades for purposes of heating and cooling. When there is delay in time between generating energy and energy demand, a great solution is thermal energy storage. There are three ways to store thermal energy which are sensible, latent, and chemical options. The important norm to choose a PCMs for a specific application is its phase change temperature. Also, other important parameters should also be possessed into account, including high latent heat and thermal conductivity values, in addition tocycling stability [2]. PCMs can be classified into organic and inorganic materials. The phase changes that result in absorbing heat involve conversions from solid to liquid, liquid to vapor, and solid to solid. The change from liquid to solid tends to be prioritized, given energy during the changes of transformation and minimal volumetric. PCMs should also have preferred Journal homepage: http://iaescore.com/online/index.php/IJAAS
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
155
properties such as: economic, thermophysical, chemical, kinetic end environmental feasibility to be used in passive LHTES systems. Other type of PCMs called organic PCMs which can be classified asparaffin and nonparaffin synthesis, like alcohols, fatty acids, glycols and esters [3]. Improved heat transfer techniques in solar energy systems leads to better performance. Among many improvement techniques in heat transfer, using of nanofluids as working fluids in solar collector systems, water heaters, cooling systems, solar still and solar cells [4]. 2.
MATERIALS AND METHODS The water/polyethylene glycol (PEG) with (tin oxide(SnO2)–tantalum carbide (TaC)) and (tin oxide(SnO2)–silicon carbide (SiC))nanoparticlessystems were prepared for thermal energy storage and release by nanofluidswith different concentrations of nanoparticles are water-polyethylene glycol/(SnO2)0.05-x - TaCx nanoparticles and water-polyethylene glycol/(SnO2)0.05-x-SiCx nanoparticles, where x=0.005, 0.01 and 0.015) where SiC and TaC nanoparticles were added each one to SnO2with concentrations are (10, 20, and 30) wt.%. The (SnO2-TaC) and (SnO2-SiC) nanoparticles were added to water with concentration (1.67×10-3 g/mL). The melting and solidification processes during heating and cooling present thermal energy storage and release. Digital device was used to measure the temperature during the heating and cooling processes. 3.
RESULTS AND DISCUSSION The heat transfer of water/PEG with (SnO2-TaC) and (SnO2- SiC) nanoparticles nanofluids was investigated during the processes of melting and solidification as shown in Figure 1 to Figure 4. The time of melting and solidification decreases with increasing of (SnO2-TaC) and (SnO2- SiC) nanoparticles concentrations. Effective dispersion of (SnO2-TaC) and (SnO2-SiC) nanoparticles into base fluid were accelerated the conductive heat transfer during the process of melting and solidification where the nanoparticles form a paths network inside the nanofluids. The water/PEGwith (SnO2-TaC) and (SnO2- SiC) nanoparticles nanofluids could be considered efficient for solar water heating system due to their characteristics of enhanced heat transfer [5-12].
Figure 1. Melting curves of water/PEG-(SnO2-TaC) nanofluids
Figure 2. Melting curves of water/PEG-(SnO2SiC) nanofluids
Solar energy storage and release application of water-phase change … (Farhan Lafta Rashid)
156
Figure 3. Solidification curves of water/PEG(SnO2-TaC) nanofluids
ISSN: 2252-8814
Figure 4. Solidification curves of water/PEG(SnO2-SiC) nanofluids
4.
CONCLUSION The water/PEG with (SnO2-TaC) and (SnO2-SiC) nanoparticles nanofluids have high efficiency for storage and release of thermal energy which can be used for heating-cooling of buildings, automobile engines, etc.The time of melting and solidification for storage and release of thermal energy applications are decreased with increase of the (SnO2-TaC) and (SnO2-SiC) nanoparticles concentrations. REFERENCES [1]
Muath A. Alomair, Yazeed A. Alomair, Hussein A. Abdullah, Shohel Mahmud, and SyedaTasnim, “Nanoparticle Enhanced Phase Change Material in Latent Heat Thermal Energy Storage System: An Experimental Study,” Proceedings of the International Conference of Energy Harvesting, Storage, and Transfer (EHST'17), Toronto, Canada, August 21–23, 2017. [2] Ismaila H. Zarma, Hamdy Hassan, Shinichi Ookawara, and Mahmoud Ahmed, “Thermal Energy Storage in Phase Change Materials: – Applications,Advantages and Disadvantages,” 1st International Cnferecne of Chemical, Energy and Environmental Engineering, Hilton Alexandira Green Plaza, 19-21 March 2017, Egypt, 2017. [3] Maria Elena Arce, Miguel Ange, Alvarez Feijoo, Andres Suarez Garcia, and Claudia C. Luhrs, “Novel Formulations of Phase Change Materials—Epoxy Composites for Thermal Energy Storagem,” Materials, vol. 11, no. 195, 2018. [4] Malleboyena Mastanaiah, K. Hemachandra Reddy, and V. Krishna Reddy, “Thermal Performance Improvement of Flat Plate Solar Collector using Nano Fluids,” International Journal of Mechanical Engineering and Technology, vol. 8, no. 7, 2017. [5] Naheda Humood Al-Garah, Farhan Lafta Rashid, Aseel Hadi, and Ahmed Hashim, “Synthesis and Characterization of Novel (Organic–Inorganic) Nanofluids for Antibacterial, Antifungal and Heat Transfer Applications,” Journal of Bionanoscience, vol. 12, 2018. [6] Hani Najm Obaid, Majeed Ali Habeeb, Farhan Lafta Rashid, and Ahmed Hashim, “Thermal energy storage by nanofluids,” Journal of Engineering and Applied Sciences, vol. 8, no. 5, pp. 143-145, 2013. [7] Farhan Lafta Rashid, Aseel Hadi, Naheda Humood Al-Garah, and Ahmed Hashim, “Novel Phase Change Materials, MgO Nanoparticles, and Water Based Nanofluids for Thermal Energy Storage and Biomedical Applications,” International Journal of Pharmaceutical and Phytopharmacological Research, vol. 8, no. 1, 2018. [8] Ibrahim R. Agool, Kadhim J. Kadhim, and Ahmed Hashim, “Preparation of (polyvinyl alcohol–polyethylene glycol– polyvinyl pyrrolidinone–titanium oxide nanoparticles) nanocomposites: electrical properties for energy storage and release,” International Journal of Plastics Technology, vol. 20, no. 1, pp. 121–127, 2016. [9] Ahmed Hashim, Ibrahim R. Agool, and Kadhim J. Kadhim, “Novel of (Polymer Blend-Fe3O4) Magnetic Nanocomposites: Preparation and Characterization for Thermal Energy Storage and Release, Gamma Ray Shielding, Antibacterial Activity and Humidity Sensors Applications,” Journal of Materials Science: Materials in Electronics, vol. 29, no. 12, pp. 10369–10394, 2018. [10] A. Hashim and A. Hadi, “Synthesis and characterization of novel piezoelectric and energy storage nanocomposites: biodegradable materials–magnesium oxide nanoparticles,” Ukrainian Journal of Physics, vol. 62, no.12, 2017. [11] Ibrahim R. Agool, Kadhim J. Kadhim, and Ahmed Hashim, “Synthesis of (PVA-PEG-PVP-ZrO 2) Nanocomposites For Energy Release and Gamma Shielding Applications,” International Journal of Plastics Technology, vol. 21, no. 2, 2017. [12] S. Harikrishnan and S. Kalaiselvam, “Experimental Investigation of Melting and Solidification Characterization of Nanofluid as PCM for Solar Water Heating System,” International Journal of Emerging Technology and Advanced Engineering, vol. 3, special issue: ICERTSD, pp. 628-635, 2013.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 154 – 156
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 157~163 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp157-163
157
A model free dissolved oxygen controller for industry effluent using hybrid variables measuring technique P. Kingston Stanley1, Sanjeevi Gandhi A.2, D. Abraham Chandy3 1,3
Karunya Institute of Technology and Sciences, India 2 Karpagam College of Engineering, India
Article Info
ABSTRACT
Article history:
The present scenario of the world relays on water scarcity, enormous amount of water is needed for people but the world has pollution which makes a great effect. Water is polluted due to various reasons such as industrial waste, sewage, and global warming and oil pollution and mainly polluted due to industrial effluent waste water. The main objective of this paper is to control the dissolved oxygen of the industrial waste water by measuring turbidity and dissolved oxygen (DO). Turbidity is measured through the intensity of the light through the medium in Nephelometric unit (NTU) and dissolved oxygen is measured in PPM through amperometic electrode method. The controller is designed to improve the quality of the effluent water through the aeration process. The very low concentration of dissolved oxygen (DO) level is too harmful to the aquatic ecosystem and it pollutes the water to the maximum extent. This measurement method of turbidity has an enhanced idea which processed with the software and it is monitored. The controller is designed as such with the condition of inverse proportionality of water turbidity and dissolved oxygen, which contributes an additive advantage to the concept design.
Received Sep 19, 2018 Revised Mar 25, 2019 Accepted Apr 15, 2019 Keywords: Dissolved oxygen Heteropolymer - formazin LabVIEW Turbidity Waste water
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: P. Kingston Stanley, Karunya Institute of Technology and Sciences, Karunya Nagar, Coimbatore, Tamil Nadu 641114, India. Email: kingstonstanley@karunya.edu
1.
INTRODUCTION The dissolved oxygen is very important factor in measuring the water quality. The dissolved oxygen level in the water should be moderate for the aquatic life in the water. The presence of dissolved oxygen is must for the aquatic level but the range should be in some certain level. Frequently varying amount of dissolved oxygen in the water level also harm the aquatic life and pollute the water nominal rate should be processed. Lower the amount of dissolved oxygen also causes water pollution. The cloudy appearance of the water caused the presence of suspended and colloidal matter. In the waterworks field, a turbidity measurement is used to indicate the clarity of water. An improvised model predictive control is being added to already develop PI control which utilizes dissolved oxygen concentration and the air pressure of the aeration system. [1]. Control of the DO concentration in the Benchmark Simulation Model No. 1 (BSM 1) was done using the proposed T-S fuzzy neural network. A fast and efficient real time control of DO concentration in the waste treatment plant is observed by adjusting the learning rate which in turn increases the convergence rate of the system [2]. In any waste water management system oxygen is a key variable that needs to be controlled beside other variables for the simulation prove that the expected level of DO for MPC can be achieved [3]. An increased accuracy is possible when the weak scattered light is measured using a single photon avalanche diode (SPAD) a resolution of 0.1 NTU can be measured within 1s [4]. The Simulations results in provided segment as well as comparison results against a well-known Lyapunov based controller and a classical PID controller [5]. The Nephelometric turbidity values ranges from 40 to 400 NTU. Journal homepage: http://iaescore.com/online/index.php/IJAAS
158
ISSN: 2252-8814
In the activated sludge waste water treatment plant DO is a crucial factor which determines how efficiently the biological processes are taking place. Improvisation of the Direct Model Reference Adaptive control (DMRAC) is relation to DO tracking for the Sequencing Batch Reactor (SBR) is put forward [6]. The first step was to identify dissolved oxygen control loop at various operating points. By using simulation the gain scheduling control strategy of DO concentration was tested; an experimentally pilot plant is also presented in this paper [7]. The successful control strategies usually employ PI control for industry based applications. Effectiveness of GS-PI and MNC control methods are displayed using simulations performed on nonlinear model [8]. The process of sewage treatment is a convoluted biochemical process. The simulation results point out a better response speed, reduced overshoot and also a reduced static error is observed [9]. Waste Water treatment plant in general is a complicated, multivariable, time depended and enormous industrial system which is nonlinear. Air/Oxygen for the aerated tanks is obtained through the aeration system (blowers, pipes, diffusers). Investigation is being carried out for the nonlinear fuzzy PI control system [10]. The outcome of this study is the development of a PC based virtual (VI) system using LABVIEW, which automatically monitors and controls the Dissolved Oxygen concentration (DO). This drastically reduces business costs in terms of automation required and has higher effectiveness with lower electricity consumption [11]. The optimal set-point of dissolved oxygen affects both the batch time and energy savings. PC-based virtual instrument (VI) system using LabVIEW, which can monitor and control the dissolved oxygen (DO) concentration in an automated manner and the quality of DO is been analyzed [12]. The controllers are validated by simulation using real data sets and an ASM2d model of the biological reactor. On the consolatory aspect, three model predictive controllers (MPCs) are designed, tested and the results are compared with a PI controller. A well-established simulation benchmark is used as a test bench for this study [13]. Fuzzy logic based controlled was designed for aeration process [14]. The batch time and the energy savings are affected by the optimal set point of the parameter dissolved oxygen [15, 16]. An investigation is being carried out for the nonlinear model predictive controller and direct reference adaptive controller. Validation of the controller is done using simulated real data sets and an ASM2d model of a biological reactor [17-21]. Various model predictive controllers for dissolved oxygen in an activated sludge wastewater treatment plant (WWTP) are being discussed. This paper proposes a new system that measures turbidity and dissolved oxygen for designing an efficient DO controller. Because turbidity is inversely proportional to DO value of water. Therefore, measuring both variables and designing a controller makes the system output response better. This paper also contributes an enhanced approach of model free control for dissolved oxygen (DO). 2. METHODOLOGY 2.1. Turbidity measurement The turbidity values are measured in NTU (Nephelometric Unit). Earlier concepts measured turbidity based on the amount of light scattered through the sample medium. Instead of measuring scattered rays of light, the other method is suggested to measure the intensity of light transmitted through the sample. It consists of a light source, which is a LASER emitting diode at a wavelength of 670 nm. The light from the LASER passes through the turbid solution and falls on a photo detector. The output signal from the photo detector is given as analog input to the embedded system, where it is processed and the given analog input is calibrated Nephelometric turbidity unit and which is reported in Table 1. This turbidity measurement values are compared with standard turbidity measuring instrument. Figure 1 shows the process to measure the turbidity. Turbidity of the sample considered is inversely proportional to dissolved oxygen concentration. This turbidity measurement helps in designing efficient controller. Different known samples were prepared to calibrate turbidity meter.
Figure 1. Turbidity measurement process Int. J. of Adv. in Appl. Sci. Vol. 7, No. 2, June 2019: 157 – 163
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
159
Table 1. Min and max output voltage value SOLUTION (NTU) 40 400
VOLTAGE (volts) 0.3 4.6
2.2. Solution preparation In order to calibrate the turbidity measuring system the standard formazin (heterocyclic polymer) is prepared by three step processes [20]. Step 1 - Preparation of solution A: Dissolving 1 gm of Hydrazine sulphates in 50ml of water and dilute to 100ml by using distilled water. Step 2 - Preparation of solution B: Dissolving 10 gm of Hexamine in 50ml of distilled water and dilute to 100ml in distilled water. Step 3 - Formazin Preparation: Take 5ml of solution A and it mix with 5ml of solution B and allow to stand 24 hours at 25º C to 30º C temperature and dilute to 100ml of distill water to attain a standard formazin turbid solution of 400NTU. The stock samples were analyzed and tested to calibrate the measuring system. The chemical reactions are shown in Figure 2.
Figure 2. Three steps of solution preparation 2.3. Dissolved oxygen control Based on literature survey most of the industry effluent waste waters have very low dissolved oxygen. Figure 3 shows about the overall block diagram of Dissolved oxygen aeration process control in effluent waste water.
Figure 3. Block diagram of dissolved oxygen aeration process control Dissolved oxygen sensor works based on the amperometric method. Current flow in between the electrodes is proportional to the dissolved oxygen concentration in the water. Analog output from dissolved oxygen sensor is interfaced with Arduino. To acquire the data using LabVIEW, there is an additional Arduino tool installed. The analog signal from the dissolved oxygen sensor is given to the Arduino and it is directly read through LabVIEW without any specialized data acquisition hardware. The air pump is used as actuator device to adjust the DO level in the waste water by varying the input voltage of the pump range from 0-12v. A model free dissolved oxygen controller for industry effluent using hybrid … (P. Kingston Stanley)
160
ISSN: 2252-8814
Figure 4 epics the variation in the DO concentration during the aeration process by varying the pump voltage in range of 2V, 4V and 10V. By changing the pump input voltage, data’s are collected and used for designing gain values of PID controller.
Figure 4. Input voltage to pump 3. RESULTS AND DISCUSSION 3.1. Experimental results for turbidity measurement The results have obtained from the turbidity sensor using the known samples which were prepared. The analog voltage from the turbidity sensor (L14G1) is fed as variable input through the DAQ in LabVIEW to the formula node. The formula node computes the turbidity value and which is displayed on the front panel. The results were shown in embedded LED display as well as in LabVIEW. Using formula node, DAQ and While loop in LabVIEW is used to convert the LED voltage (0-5V) output to Nephelometric (40-400 NTU) Unit. Figure 5 shows the LabVIEW design for calibrating the voltage value from phototransistor into NTU. The output voltage for various values of 10 samples with different turbidity ranges are shown in Figure 6.
Figure 5. LabVIEW for converting voltage to NTU Higher order polynomial equation can improve the curve fitting relation between the input (voltage) and the output (turbidity). Therefore 4th degree linear polynomial formula which is used to convert voltage into turbidity as follows: Y = q1*x^4 + q2*x^3 + q3*x^2 + q4*x + q5; q1 = 1.732 q2 = -10.35 q3 = 21.25 q4 = 21.31 q5 = 31 Int. J. of Adv. in Appl. Sci. Vol. 7, No. 2, June 2019: 157 – 163
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
161
Figure 6. Samples solution vs output voltage 3.2. PID controller design for DO Thus the real time PID controller is implemented in LabVIEW. The set point of 5 PPM is taken to the three samples, the PID controls the process, the steady state is obtained at various load changes of each samples. The PID Controller designed using LabVIEW, is tuned using Ziegler Nicholes method and the gain is changed by noting the response of the system. Figure 7 shows the design of LabVIEW program to control DO level. Design of controllers are always different in real time implementation. Therefore PID controller gain values are selected from trial and error method also and its suits for DO concentration set point tracking. Proportional & Integral controllers are used frequently in automation industry, especially when quickness of the response is not a matter. A control without Derivative mode is used when fast response of the system is not required and great disturbances and noise are existing during operation of the process. As the aeration process makes the DO to change quickly, Proportional, Integral and derivative gain very much need for this process.
Figure 7. LabVIEW program to control DO level using PID controller
Figure 8. PID controller response for domestic wastewater A model free dissolved oxygen controller for industry effluent using hybrid … (P. Kingston Stanley)
162
ISSN: 2252-8814
Figure 9. PID controller response for paper industry wastewater
Figure 10. PID controller response for normal water Figure 8 demonstrates the DO Control in domestic waste water. The waste water must have a minimum of 5 PPM level of DO concentration as per the literature survey. Therefore set point of 5PPM was taken to control the concentration of DO. Figure 9 indicates the DO control for paper industry waste water, the PID controller was able to achieve to the set point value of 5 PPM within 1000 milli-seconds. These experiments are tested with 30L water. The DO control for normal water is shown in Figure 10. The comparing all these three results, normal water reacts quickly to reach the set point of 5 PPM. And these experiments were performed multiple times with fresh waters in the setup. The designed controller responses shows that the controller is able to track DO concentration effectively using hybrid variables rather than measuring single variable based DO control. Through this hybrid variables based automation process, much amount of energy can be saved. 4.
CONCLUSION In industrial effluent and domestic waste water have measured in terms of turbidity and DO concentration. The turbidity sensor and DO sensor were calibrated with help of standard solution. The designed turbidity measurement is another method which is useful to measure colloidal particles of water in fields with a lesser conceptual cost aspect. Many phototransistors were used to improve the sensitivity of measuring turbidity values. The model free controller is designed to control the DO concentration effectively using both variables namely turbidity and DO concentration. The controlling of DO level concentration reduces the risk of polluting the land and can cut down many harmful factors too. The output responses of PID controllers with DO concentration level for domestic waste water; industrial effluent and drinking water were shown. From the responses it was understood the controller is effectively controlling the DO concentration without any overshoot and under shoot. The real time PID controller was successfully implemented in the pilot plant. The adaptive controller will be proposed to reduce the settling time of the process.
Int. J. of Adv. in Appl. Sci. Vol. 7, No. 2, June 2019: 157 – 163
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
163
REFERENCES [1] Gabriel Harja, Grigore Vlad, and Ioan Nascu, “MPC advanced control of dissolved oxygen in an activated sludge wastewater treatment plant,” IEEE International Conference on Automation, Quality and Testing, Robotics, 2016. [Online]. Available: https://ieeexplore.ieee.org/document/7501329/. [2] Wen-Tao Fu, Jun-Fei Qiao, Gai-Tang Han,and Xi Meng , “Dissolved oxygen control system based on the T-S fuzzy neural network,” IEEE International Conference Neural Networks, 2015. [Online]. Available: https://ieeexplore.ieee.org/document/7280506/. [3] Xiaoxin Liu, Yuanwei Jing, Tao Ren, and Siying Zhang, “Model predictive control application in dissolved oxygen control,” IEEE Chinese Control and Decision Conference, 2015. [Online]. Avalaible: https://ieeexplore.ieee.org/document/7162095/. [4] Huanqin Wang, Yixin Yang, Zhe Huang, and Huaqiao Gui, “Instrument for Real-Time Measurement of Low Turbidity by Using Time-Correlated Single Photon Counting Technique,” IEEE Transactions on Instrumentation and Measurement, vol. 64(4), pp. 1075-1083, 2015. [5] Alex Alzate, Adriana Amicarelli, Lina Gómez, and Fernando di Sciascio, “Model based predictive strategy for dissolved oxygen control applied to a batch bioprocess,” IEEE - Workshop Information Processing and Control, 2015. [6] Piotr Hirsch, Robert Piotrowski, and Kazimierz Duzinkiewicz, “Two-step model based adaptive controller for Dissolved Oxygen control in Sequencing Wastewater Batch Reactor,” International Conference on Methods and Models in Automation and Robotics, 2015. [7] Ciprian Vlad, Sergiu Caraman, Daniela Carp, Viorel Mînzu, and Marian Barbu, “Gain Scheduling control of dissolved oxygen concentration in a wastewater treatment process,” Mediterranean Conference Control & Automatio,. 2012. [8] C. Vlad, M. Sbarciog, M. Barbu, S. Caraman, and A. Vande, “Wouwer Indirect Control of Substrate Concentration for a Wastewater Treatment Process by Dissolved Oxygen Tracking,” Journal of control engineering and applied informatics, vol. 14(1), pp. 37-47, 2012. [9] Minghe Li, Jingen Peng, and Jian Wang, “Study on Smith-self-adaptive fuzzy PID controller in dissolved oxygen control of sewage treatment,” IEEE Conference Industrial Electronics and Applications, 2012. [10] Zawadzki, R. Piotrowski, “Nonlinear fuzzy control of the dissolved oxygen in activated sludge processes,” IEEE Conference Emerging Technologies & Factory Automation, 2012. [11] Hsiung Cheng Lin, Liang-Tsung Huang, Lien-Fu Lai, and Yin-Fan Chi, “A remote automated system for a case study of dissolved oxygen monitoring and control,” IEEE International Symposium on Industrial Electronics, 2009. [12] Baiqing Zhou, Juan Liu, and Lu Li, “Evaluation Method for Performance of Reverse Osmosis Antiscalants Based on the Turbidity,” International Conference on Electrical and Control Engineering, 2010. [13] Jun-dong Wang, Pei-yan Li, Yong-ming Zhang, and Wei-gui Qi, “River Water Turbidity Forecasting Based on Phase Space Reconstruction and Support Vector Regression,” Intelligent Computation Technology and Automation (ICICTA), International Conference, 2010. [14] J. Ferrer, M. A. Rodrigo., A. Seco, and J.M. Penya-roja, “Energy saving in the aeration process by fuzzy logic control,” Water Science and Technology, vol. 38(3), pp. 209-217, 2017. [15] Ruey-Fang Yu, Cheng-Nan Chang, and Wan-Yuan Cheng, “Applying real time control to enhance the performance of nitrogen removal in the continuous flow SBR system,” Water science and Technology, vol. 38(3), pp. 271-280, 1998. [16] Young-Hwang Kim, Chang Kyoo Yoo, and In-Beum Lee, “Optimization of biological nutrient removal in a SBR using simulation –based iterative dynamic programming,” Chemical Engineering Journal, vol. 139(1), pp. 11-19, 2008. [17] W. Chotkowski, M. A. Brdys, and K. Konarczak, “Dissolved oxygen control for activated sludge processes,” International Journal of Systems Science, vol. 36(12), pp. 727-736, 2005. [18] Reza Loloee, Per A. Askeland, and Ruby N., “Ghosh Dissolved Oxygen Sensing in a Flow Stream using Molybdenum Chloride Optical Indicators,” IEEE Sensors, 2007. [19] Sanchez and M. R. Katebi, “Predictive control of dissolved oxygen in an activated sludge wastewater treatment plant,” European Control Conference, 2003. [20] Budi Srinivasarao, G. Sreenivasan, and Swathi Sharma, “Comparison of Dynamic Stability Response of A SMIB with PI and Fuzzy Controlled DPFC,” Indonesian Journal of Electrical Engineering and Informatics, vol. 5(3), pp. 199-206, 2017. [21] V. Balaji, Dr. L. Rajaji, and Shanthini K., “Comparison Analysis of Model Predictive Controller with Classical PID Controller for pH Control Process,” Indonesian Journal of Electrical Engineering and Informatics, vol. 4(4), pp. 250-255, 2016.
A model free dissolved oxygen controller for industry effluent using hybrid … (P. Kingston Stanley)
International Journal of Advances in Applied Sciences (IJAAS) Vol. 8, No. 2, June 2019, pp. 164~170 ISSN: 2252-8814, DOI: 10.11591/ijaas.v8.i2.pp164-170
164
Subsynchronous resonance oscillations mitigation via fuzzy controlled novel braking resistor model M. Fayez1, M. Mandour2, M. El-Hadidy3, F. Bendary4 1,2,4
Electrical Engineering Department, University of Benha, Egypt 3 Egyptian Electricity Holding Company (EEHC), Egypt
Article Info
ABSTRACT
Article history:
Subsynchronous resonance (SSR) torsional torque oscillations is a problem of a great concern in the power engineering community. SSR causes torsional torque oscillations with ever-increasing magnitudes occurring in the machine shaft sections causing a premature fatigue life expenditure of the shaft metal. In this paper, dynamic braking switching strategy designed through fuzzy logic control theory and implemented via novel braking resistor model, namely chopper rectifier controlled braking resistor for tempering SSR torsional torque oscillations of a large turbo-generator. The proposed mitigation scheme has been tested on the IEEE second benchmark model for SSR studies. Comparative simulation study via MATLAB/Simulink-based modeling and simulation environment of the test model with and without the suggested mitigation regime should demonstrate its effectuality for mitigation of SSR torsional torque oscillations.
Received Nov 9, 2018 Revised Apr 1, 2019 Accepted May 2, 2019 Keywords: Chopper rectifier controlled braking resistor Fuzzy logic control Series compensation Subsynchronous resonance (SSR) Thyristor controlled braking resistor
Copyright © 2019 Institute of Advanced Engineering and Science. All rights reserved.
Corresponding Author: Mohamed Fayez, Electrical Engineering Department, University of Benha, Shoubra, Cairo 108, Egypt. Email: eng_mf69@yahoo.com
1.
INTRODUCTION Usually, the rotor of a steam turbine-generator set is considered as a single rotational mass with certain inertia constant in the analysis and investigation of many power system dynamic problems [1]. But, the rotor of a steam turbine-generator is a highly complicated mechanical structure. The rotor of a steam turbinegenerator set (i.e. shaft train) is composed of a string of dominant massive elements, each one has a considerable inertia constant and relatively large diameter, rigidly connected in tandem at the thinner shaft extensions [1]. The complete shaft train can surpass 50 meters of total length and weight several hundred tons [1]. The spring-mass model of the turbine-generator shaft is prerequisite for elucidating the torsional torque oscillations occurring at the shaft extensions [1]. SSR torsional torque oscillations is a problem of a great concern in the power engineering community [1, 2]. SSR is a transmission system-based problem that causes adverse devastating effects on the turbine-generator sets operating under this circumstance [2]. Therefore, SSR is considered as multidisciplinary power utility problem. In 1970, and again in 1971, a 909 MVA cross-compound turbine-generator at Mohave coal-fired power plant in southern Nevada, California, USA, had encountered collector ring damage in the high-pressure unit exciter shaft due to SSR torsional torque oscillations [2, 3]. The devastating consequences of Mohave accidents have fed off the industry’s enthusiasm about this phenomenon to find reliable mitigation countermeasures and underscored the urge for quickly implementing them [3]. SSR causes torsional torque oscillations with ever-increasing magnitudes occurring in the machine shaft sections causing a premature fatigue life expenditure of the shaft metal [1-5]. The shaft is then definitely destined to experience irrevocable Journal homepage: http://iaescore.com/online/index.php/IJAAS
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
165
low-cycle fatigue cracks or even a complete shaft fracture [1-5]. SSR arises in power plants, especially with steam-turbine prime movers, as an undesirable collateral result for transmission lines emanating from a power plant and equipped with series capacitors at certain conditions [1-5]. The fatigue damage of large turbinegenerator shaft is definitely undesirable and catastrophic event in the power industry [5, 6]. This is because the shaft rehabilitation process is very expensive and takes several months to be accomplished during which the unit is intuitively out of service [6]. Also, during the shutdown shaft rehabilitation period, the plant should purchase the electric energy required to energize the auxiliary equipments of the plant from the utility which brings more economic casualties to the plant owner beside the inability to sell the output electrical energy of the plant [6]. Series compensation has been widely utilized by the utility since 1952 for economically transmitting bulk power transactions over long transmission distances [7]. Series compensation has many overwhelming benefits. It aggrandizes the electric power transferring capability of long transmission corridor, enhances the voltage profile of the grid, and boosts the angular stability of the system and so forth [7]. At certain cases, series compensated power systems are vulnerable to have SSR condition in which the series compensated electric system interchanges a considerable amount of energy with a turbine-generator mechanical system at one more of the natural frequencies of the combined electromechanical system below the synchronous frequency of the electric system [4-6]. For the sake of secure operation of bulk power systems, large turbo-generators are certainly needed to be in service for as long as possible due to their active and reactive power supplying capabilities [5]. So, utilizing protection relays to protect large units from SSR conditions will not generally be considered except as a final sanctuary [5]. Therefore, SSR mitigation countermeasures should alleviate the devastating SSR torque oscillations and eliminate the need to trip the units that are prone to resonant condition. There have been a considerable number of technical studies and research papers addressing the SSR problem and presenting many mitigation countermeasures with control algorithms, mitigation equipment sizes, and the most effective location in the network for these prospective mitigation candidates [6]. Several mitigation approaches have been developed for tempering of SSR oscillations such as, Static VAR Compensator (SVC), Thyristor-Controlled Series Capacitor (TCSC), Static Synchronous Series Compensator (SSSC), Static Compensator (STATCOM), Unified Power Flow Controller (UPFC), and excitation system control through using an appropriate supplementary stabilizing signal [8-10]. Among the multiple presently conceivable SSR mitigation candidates, the dynamic braking resistor (BR) is an extremely effective and relatively cheap mitigation candidate for SSR problems [10-15]. BR was initially implemented for augmenting transient stability of electric power systems [1]. BR functions as a pseudo electric load with the ability to consume the momentarily accelerating energy in power systems arising due to severe nearby faults, especially in power systems with hydro dominated power plants [1]. Therefore, by dissipating the extra generated power, BR should prohibit generator pole slipping conditions and thereby augmenting the transient stability of electric power systems [1]. Bonneville Power Administration (BPA) dynamic brake is one of the largest existing BR installations in realistic power systems with power dissipation ability of 1400 MW with 230 kV rated voltage [1]. Many works of literature have addressed the utilization of BR connected to the grid via a three-phase AC voltage controller, i.e. thyristor controlled braking resistor (TCBR), for mitigation of multi-modal shaft torsional oscillations [15-20]. TCBR per phase model is constructed as two back-to-back connected high-power thyristor valves with BR in series which means that three BRs were utilized for SSR torsional oscillations mitigation [16-20]. Very recently, a novel BR model, namely Chopper Rectifier Controlled Braking Resistor (CRCBR) model, is introduced to the academic community as an alternative innovative mean for implementing the concept of dynamic braking interventions for enhancing the transient stability of power systems [21-23]. In [21, 22], the effectiveness of CRCBR for transient stability enhancement in single machine infinite bus (SMIB) system and the Western System Coordinated Council (WSCC), 3-machine 9-bus system, respectively, have been tested. CRCBR controlled through a fuzzy logic controller (FLC) is proposed in this article for mitigation of SSR torsional oscillations. CRCBR is fundamentally a single BR bank linked to the three-phase terminals of the synchronous generator through a three-phase full-wave uncontrolled rectifier diode bridge and chopper switch (CS) [21, 22]. No similar work could be found in the literature to date regarding the utilization of CRCBR for mitigation of SSR torsional torque oscillations. The CRCBR is either switched ON or switched OFF based on the energization signal produced by the FLC. The generator mass speed deviation is implemented in this work as a local control input signal to the FLC for strategizing the CRCBR switching intervals. The influence of the proposed scheme in mitigating SSR torsional oscillations is tested in this work using the IEEE second benchmark model via MATLAB/Simulink-based modeling and simulation environment. Comparative simulation study of the test system with and without the suggested mitigation regime should demonstrate its effectuality for mitigation of SSR torsional torque oscillations. Subsynchronous resonance oscillations mitigation via fuzzy controlled novel braking … (M. Fayez)
166
ISSN: 2252-8814
The rest of this paper is organized as follows. In section 2, the well-known test system is briefly described. In section 3 the concept of utilizing the FLC to orchestrate the switching strategy of the CRCBR is introduced. In section 4, MATLAB/Simulink time domain simulation results are delineated with comments. In section 5, the key conclusions of this work are drawn. Finally, the references used in this work are listed. 2.
SYSTEM MODEL To scrutinize the effectiveness of the suggested mitigation regime in this paper, the well-known IEEE second benchmark model for computer simulation of SSR is adopted. Figure 1 depicts the single line diagram of the system under study together with CRCBR and the steam turbine-generator shaft detail [21-23]. The IEEE second benchmark model is basically a single synchronous generator (600 MVA/22kV/60 Hz/3600 rpm) connected to an infinite bus through generator step-up transformer (600 MVA/22kV/500kV) and two extrahigh voltage (EHV) transmission lines, one of which is equipped with a series capacitor of 55% seriescompensation ratio. The generator is driven by one single-flow high-pressure (HP) steam turbine and one double-flow low-pressure (LP) steam turbine connected in tandem. The model of CRCBR consists of one BR bank connected to the generator terminals via six-pulse full wave uncontrolled Rectifier Bridge and a chopper switch as shown in Figure 1 [21-23]. In this article, the Insulated Gate Bipolar Transistor (IGBT) is utilized as a chopper switch. The BR bank is assumed to be capable of dissipating 50 % the connected generator apparent power (i.e. 300 MW).
TLP-Gen . G
THP -LP HP
LP
600 MVA-60 Hz 22 kV-500 kV
B22
Turbine-Generator Set
B500
Line A
Cs = 55%
B500
Infinite Bus
G Line B 600 MVA 22kV 3600 rpm
Δω_G
Rectifier Bridge A
FLC
B
Firing Circuit
F1 Generator Terminals A
Chopper Switch
300 MW
B C
Chopper Switch BR
C
Gate signal
Figure 1. The IEEE second benchmark model with a turbine generator shaft multi-mass model incorporating chopper rectifier controlled braking resistor device [21-23] The brake is switched ON or OFF via the chopper switch to mitigate the shaft torsional oscillations by reducing the net available energy for acceleration and hence speed deviation of the generator mass due to severe disturbances such as system close in faults [1]. The proposed scheme utilizes a local control signal represented by the generator mass speed deviation to help the FLC determine whether the CRCBR should be switched ON or switched OFF. The system is completely modeled in MATLAB/Simulink environment. The electrical parameters for the synchronous generator are obtained from [23]. The under-study steam turbine configuration is appropriately illustrated in [23]. It consists of HP section and single LP section. The turbine contributing torque fractions for the turbine sections HP, and LP are 50%, and 50%, respectively [23]. The detailed turbine-generator shaft mechanical data is obtained from [23]. The electrical parameters of the system in per-unit, with 100 MVA as a common base power, are also obtained from [23]. 3.
FUZZY LOGIC CONTRLLER DESIGN The fuzzy logic is differing from the crisp logic in Boolean theory which utilizes only two logical levels (0 or 1) in that it uses unbounded logical levels (from 0 to 1) to deal with issues that have uncertainties or obscure situations [24-26]. FLCs are demonstrated to be more robust and their performances possess a lesser sensitivity to the parametric differences than the traditional controllers [24-26]. Recently, fuzzy-logic control has gained a reputation in effectively solving many power system dynamic problems [24-26]. FLC is a nonlinear controller with a rule-based system, in other words, it is an artificial-intelligence (AI) based system Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 164 – 170
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
167
which relied on the use of expert knowledge [24-26]. This expert knowledge is commonly acquired through conducting comprehensive mathematical modeling, analysis, and development of control algorithms for the power systems [24-26]. A set of fuzzy linguistic rules, usually distinguished by “IF-THEN” statements, representing a control decision-making mechanism responsible for regularizing the effectuations for system stimulants [24-26]. The basic FLC configuration consists of the following four prime stages: fuzzification, knowledge base, inference engine, and defuzzification [24-26]. In the fuzzification stage, the crisp input values are mapped into fuzzy variables by using normalized membership functions and occasionally input weighting factors [24-26]. The fuzzy-logic inference engine appropriately deduces the suitable control action based on the attainable rule base [24]. Two major inference engine types are Takagi-Sugeno (TS), and Mamdani [27]. The fuzzy control action is converted to proper crisp value through the defuzzification process using normalized membership functions and occasionally output weighting factors [24-26]. TS type is more efficient than the Mamdani type for dynamic systems with fast changing dynamics because it works well with linear, adaptive and optimization techniques and it is well suited to mathematical analysis [27]. Therefore, TS inference mechanism is utilized in the simulation study. The overlapping of the membership functions makes the mapping smoother and enhances the controller performance [24]. It has proven that the damping level obtained from using the generator mass speed deviation as the controller input is better than the damping level obtained from using other turbine sections speed deviation signals [17, 20]. Also, the generator mass speed signal is not so difficult to capture like HP, and LP speed signals since the entire steam turbine is rigorously sealed with the outer cases via steam seal system. So, the generator mass speed deviation in per-unit, Δɷ, is selected as input to the FLC while the output signal is constant with either 0 or1. The output signal of FLC is then sent to the IGBT gating circuit which produces the gate signals for the chopper switch based on the controller decision output. The Gaussian and Sigmoid inputs membership functions are shown in Figure 2 in which three linguistic variables, NB (Negative Big), Zo (Zero) and PB (Positive Big), are defining the fuzziness of the controller input. The membership functions parameters are determined by trial and error according to the generator mass speed deviation swing range based on an extensive simulation study. The output type is constant having either 0 or 1 values (0 for both ZO and NB, 1 for PO). Where 0 indicates that the IGBT should be in OFF state and 1 indicates that the IGBT should be in ON state.
Figure 2. FLC input membership functions in per-unit The proposed control scheme is very straightforward and simple since it has only three control rules where the BR is inserted if the generator speed deviation exceeds a certain value (the acceleration state) and removed elsewhere (steady state and deceleration state). There are three premise membership functions depicted in Figure 2, one for each rule, and the conclusions are singleton so the fuzzy control rules are as follows, If the input (Δɷ) is NB then the output is 0, If the input (Δɷ) is Zo then the output is 0 and If the input (Δɷ) is PB then the output is 1. 4.
SIMULATION RESULTS To further demonstrate the effectiveness of the proposed mitigation regime, time domain simulation study via Simulink model is carried out. Three-phase to ground (3LG) self-healing fault is applied at line B, very close to generator high voltage bus at fault point F1, as shown in Figure 1. The fault is applied at 0.25 second from the simulation time of 5 seconds. The fault is self-cleared after 0.0169 seconds from the inception of the disturbance. The torsional torque profiles of turbine-generator shaft system in per-unit (p.u.) and the speed responses of turbine-generator different masses with and without the fuzzy-controlled CRCBR are listed as a family of curves shown in Figure 3 and Figure 4. Subsynchronous resonance oscillations mitigation via fuzzy controlled novel braking … (M. Fayez)
168
ISSN: 2252-8814
(a)
(b) Figure 3. Torque responses with and without fuzzy controlled CRCBR due to the three-phase self-healing fault at the generator HV bus in p.u., (a) LP turbine-generator shaft torsional torque, (b) HP-LP Shaft torsional torque
(a)
(b)
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 164 – 170
Int. J. of Adv. in Appl. Sci.
ISSN: 2252-8814
169
(c) Figure 4. Relative turbine-generator mass speed responses due to the three-phase self-healing fault at the generator HV bus in p.u., (a) Generator mass speed deviation, (b) LP turbine mass speed deviation, (c) HP turbine mass speed deviation. Finally, in Figure 5 it is shown the comparative responses of the mechanical rotor angle deviation for the synchronous generator under the given fault condition. Due to applying the proposed technique the oscillations in the rotor angle is settled after 1 second from the perturbation initiation.
Figure 5. Mechanical rotor angle deviation in [rad] As evidenced from the simulation results portrayed in Figure 3, Figure 4 and Figure 5, in the base case simulation results without the proposed scheme, it is clearly seen that both speed and torque responses seem to be sustained with ever-increasing amplitudes in the time frame of the simulation which indicates the unstable nature of the responses. The driving mechanical torque transmitted between the shaft's masses have experienced devastating oscillations with extremely unacceptable amplitudes. These torque oscillations will, for sure, end up causing a premature shaft fatigue life expenditure and irrevocable shaft cracks or even a shaft fracture with the unpleasant outcomes. Due to applying the fuzzy controlled CRCBR, it is obviously shown that the shaft torsional oscillations of a turbine-generator set and the speed responses have experienced an excellent supplementary damping, which further enables the utilizing of series compensation safely near the steam power plants without any mechanical jeopardy. 5.
CONCLUSION This paper authenticates the effectiveness of CRCBR controlled via FLC to damp SSR shaft torsional oscillations. From the simulation results, the speed and the torque profiles of the machine manifest a significantly good supplementary damping which allows the torsional oscillations and the speed oscillations to die out quickly. Utilizing only one BR unit, instead of three units in case of TCBR, might encourage the utilities to use it for damping the shaft torsional oscillations arising from SSR conditions and compromise between the benefits of implementing series compensation and the concerns of the machine shaft damage. The CRCBR is considered as a viable mean for boosting the security of the power system from the operational perspective by extending the fatigue life of the units in the power system to their maximum possible potential by neutralizing any SSR condition. Finally, implementing the proposed scheme should capacitate conventional series compensation on transmission lines emanating from thermal power plants safely and soundly without jeopardizing the shaft mechanical integrity of the turbine-generator set. Subsynchronous resonance oscillations mitigation via fuzzy controlled novel braking … (M. Fayez)
170
ISSN: 2252-8814
REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27]
M. Fayez Ahmed, et al., “Torsional Oscillations Mitigation for Interconnected Power System via Novel Fuzzy Control Based Braking Resistor Model,” Conseil International des Grands Réseaux Électriques (CIGRÉ), Paris, France, August 26-31, pp. 1–9, 2018. EPRI Technical Report 1011679, “Steam Turbine-Generator Torsional Vibration Interaction with The Electrical Network,” Electric Power Research Institute (EPRI), 2005. Geoff Klempner and Isidor Kerszenbaum, Handbook of Large Turbo-Generator Operation and Maintenance. John Wiley & Sons, 2008. Duncan N. Walker, Torsional Vibration of Turbomachinery. McGraw-Hill, 2003. P. M. Anderson, Power system protection. John Wiley & Sons, 1999. Leonard L. Grigsby, Power system stability and control, Third edition. CRC Press, 2012. Mircea Eremia, Chen-Ching Liu, and Abdel-Aty Edris, Advanced solutions in power systems HVDC, FACTS and artificial intelligence. John Wiley & Sons, 2016. R. Mohan Mathur and Rajiv K. Varma, Thyristor-based FACTS controllers for electrical transmission systems. John Wiley & Sons, 2002. Shruthi Ramachandra, et al., ”Dynamic bifurcation analysis and mitigation of SSR in SMIB system,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 15, no. 3, 2019. S. Helmy, et al., “Mitigating subsynchronous resonance torques using dynamic braking resistor,” International Middle East Power Systems Conference (MEPCON), Egypt, 2010, pp. 416–421. Li Wang, and Ching-Huei Lee, “Application of dynamic resistance braking on stabilizing torsional oscillations,” in International Conference on Computers, Communications and Automation, Beijing, China, 1993, pp. 145–148. R. M. Hamouda, et al., “Damping Torsional Oscillations in Large Turbo-Generators Using Thyristor Controlled Braking Resistors,” in 2008 Australasian Universities Power Engineering Conference (AUPEC'08), Australasia, 2008, pp. 1–6. A. H. Rahim and H. M. Al-Maghraby, “Dynamic Braking Resistor for Control of Subsynchronous Resonant Modes,” IEEE Power Engineering Society Summer Meeting, Seattle, WA, USA, 2000, pp. 1930–1935. A. H. M. A. Rahim and H. M. AI-Maghraby, “Control of SSR Oscillations Through A Quasi-Optimum Dynamic Braking Strategy,” International Journal of Power and Energy, vol. 23, pp. 90–95, 2003. A. H. M. A. Rahim, “A Minimum-Time Based Fuzzy Logic Dynamic Braking Resistor Control for Sub-Synchronous Resonance,” Electrical Power and Energy Systems, vol. 26, pp. 191–198, 2004. O. Wasynczuk, “Damping Shaft Torsional Oscillations Using A Dynamically Controlled Resistor Bank,” IEEE Transactions on Power Apparatus and Systems, vol. 100, pp. 3340–3349, 1981. EPRI Technical Report 103902, “Dynamic Brake Control to Reduce Turbine Shaft Transient Torque,” Electric Power Research Institute (EPRI), 1994. M. H. Ali, et al., “Minimization of Shaft Torsional Oscillations by Fuzzy Controlled Braking Resistor Considering Communication Delay,” WSEAS Transactions on Power Systems, vol. 3, pp. 174–179, 2008. M. H. Ali, et al., “Coordination of Fuzzy Controlled Braking Resistor and Optimal Reclosing for Damping ShaftTorsional Oscillations of Synchronous Generator,” in International Conference on Electrical Machines and Systems (ICEMS), Seoul, South Korea, 2007, pp. 1259–1264. M. H. Ali, et al., “A fuzzy Logic Controlled Braking Resistor Scheme for Damping Shaft Torsional Oscillations,” IEEJ Transactions on Power and Energy, vol. 124, pp. 207–214, 2004. R. Saluja and M. H. Ali, “Novel Braking Resistor Models for Transient Stability Enhancement in Power Grid System,” IEEE PES Innovative Smart Grid Technologies (ISGT) Conference, Washington DC, USA, 2013, pp. 1–6. R. Saluja, S. Ghosh, and M. H. Ali, “Transient stability enhancement of multi-machine power system by novel braking resistor models,” Proceedings of the IEEE Southeast Conference, Florida, USA, 2013, pp. 1 – 6. IEEE Committee Report, “Second Benchmark Model for Computer Simulation of Subsynchronous Resonance,” IEEE Transactions on Power Apparatus and Systems, vol. l04, pp. 1057–1066, 1985. M. Fayez Ahmed, et al., “Torsional Oscillations Mitigation via Novel Fuzzy Control Based Braking Resistor Model,” International Electrical Engineering Journal (IEEJ), vol. 7, pp. 2173–2181, 2016. Gilfred Allen Madrigal, et al., “Fuzzy logic-based maximum power point tracking solar battery charge controller with backup stand-by AC generator,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 16, no. 1, pp. 136-146, 2019. Salam Waley Jeaeb, et al., “Advanced optimal by PSO-PI for DC motor,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 16, no, 1, pp. 165-175, 2019. S. Mohagheghi, et al., “Modified Takagi-Sugeno Fuzzy Logic-Based Controllers for A Static Compensator in A Multi-Machine Power System,” in Industry Applications Conference, Seattle, USA, 2004, pp. 2637–2642.
Int. J. of Adv. in Appl. Sci. Vol. 8, No. 2, June 2019: 164 – 170
Institute of Advanced Engineering and Science Indonesia
: D2, Griya Ngoto Asri, Bangunharjo, Sewon, Yogyakarta 55187, Indonesia
Malaysia
: 51 Jalan TU 17, Taman Tasik Utama, 75450 Malacca, Malaysia
COPYRIGHT TRANSFER FORM (Please compile this form, sign and send by e-mail) Please complete and sign this form and send it back to us with the final version of your manuscript. It is required to obtain a written confirmation from authors in order to acquire copyrights for papers published in the International Journal of Advances in Applied
Sciences (IJAAS) Full Name and Title Organisation Address and postal code City Country Telephone/Fax E-mail Paper Title Authors
Copyright Transfer Statement The copyright to this article is transferred to Institute of Advanced Engineering and Science (IAES) if and when the article is accepted for publication. The undersigned hereby transfers any and all rights in and to the paper including without limitation all copyrights to IAES. The undersigned hereby represents and warrants that the paper is original and that he/she is the author of the paper, except for material that is clearly identified as to its original source, with permission notices from the copyright owners where required. The undersigned represents that he/she has the power and authority to make and execute this assignment. We declare that: 1. This paper has not been published in the same form elsewhere. 2. It will not be submitted anywhere else for publication prior to acceptance/rejection by this Journal. 3. A copyright permission is obtained for materials published elsewhere and which require this permission for reproduction. Furthermore, I/We hereby transfer the unlimited rights of publication of the above mentioned paper in whole to IAES. The copyright transfer covers the exclusive right to reproduce and distribute the article, including reprints, translations, photographic reproductions, microform, electronic form (offline, online) or any other reproductions of similar nature. The corresponding author signs for and accepts responsibility for releasing this material on behalf of any and all co-authors. This agreement is to be signed by at least one of the authors who have obtained the assent of the co-author(s) where applicable. After submission of this agreement signed by the corresponding author, changes of authorship or in the order of the authors listed will not be accepted.
Retained Rights/Terms and Conditions 1. 2.
3.
Authors retain all proprietary rights in any process, procedure, or article of manufacture described in the Work. Authors may reproduce or authorize others to reproduce the Work or derivative works for the author’s personal use or for company use, provided that the source and the IAES copyright notice are indicated, the copies are not used in any way that implies IAES endorsement of a product or service of any employer, and the copies themselves are not offered for sale. Although authors are permitted to re-use all or portions of the Work in other works, this does not include granting third-party requests for reprinting, republishing, or other types of re-use.
Yours Sincerely,
Corresponding Author‘s Full Name and Signature Date: ……./……./…………
IAES Journals, http://iaescore.com/journals International Journal of Advances in Applied Sciences (IJAAS), email: iaes.editor@gmail.com, editor@iaesjournal.com
Guide for Authors International Journal of Advances in Applied Sciences (IJAAS) is a peer-reviewed and open access journal dedicated to publish significant research findings in the field of applied and theoretical sciences. The journal is designed to serve researchers, developers, professionals, graduate students and others interested in state-ofthe art research activities in applied science areas, which cover topics including: chemistry, physics, materials, nanoscience and nanotechnology, mathematics, statistics, geology and earth sciences. Papers are invited from anywhere in the world, and so authors are asked to ensure that sufficient context is provided for all readers to appreciate their contribution.
The Types of papers The Types of papers we publish. The types of papers that may be considered for inclusion are: 1) Original research; 2) Short communications and; 3) Review papers, which include meta-analysis and systematic review
How to submit your manuscript All manuscripts should be submitted online at http://iaescore.com/journals/index.php/IJAAS
General Guidelines 1) Use the IJAAS guide (http://iaescore.com/journals/ourfiles/IJAAS_guideforauthors_2019.docx) as template. 2) Ensure that each new paragraph is clearly indicated. Present tables and figure legends on separate pages at the end of the manuscript. 3) Number all pages consecutively. Manuscripts should also be spellchecked by the facility available in most good word-processing packages. 4) Extensive use of italics and emboldening within the text should be avoided. 5) Papers should be clear, precise and logical and should not normally exceed 3,000 words. 6) The Abstract should be informative and completely self-explanatory, provide a clear statement of the problem, the proposed approach or solution, and point out major findings and conclusions. The Abstract should be 150 to 250 words in length. The abstract should be written in the past tense. 7) The keyword list provides the opportunity to add keywords, used by the indexing and abstracting services, in addition to those already present in the title. Judicious use of keywords may increase the ease with which interested parties can locate our article. 8) The Introduction should provide a clear background, a clear statement of the problem, the relevant literature on the subject, the proposed approach or solution, and the new value of research which it is innovation. It should be understandable to colleagues from a broad range of scientific disciplines. 9) Explaining research chronological, including research design and research procedure. The description of the course of research should be supported references, so the explanation can be accepted scientifically. 10) Tables and Figures are presented center. 11) In the results and discussion section should be explained the results and at the same time is given the comprehensive discussion. 12) A good conclusion should provide a statement that what is expected, as stated in the "Introduction" section can ultimately result in "Results and Discussion" section, so there is compatibility. Moreover, it can also be added the prospect of the development of research results and application prospects of further studies into the next (based on the results and discussion). 13) References should be cited in text by numbering system (in Vancouver style), [1], [2] and so on. Only references cited in text should be listed at the end of the paper. One author should be designated as corresponding author and provide the following information: • E-mail address • Full postal address • Telephone and fax numbers Please note that any papers which fail to meet our requirements will be returned to the author for amendment. Only papers which are submitted in the correct style will be considered by the Editors.
International Journal of Advances in Applied Sciences (IJAAS) Institute of Advanced Engineering and Science (IAES) e-mail: ijaas@iaesjournal.com
IJAAS Journal Order Form Volume
Number
Amount
Price (Rp)
Price (USD)
Total
Name
:
Company
:
Address
:
City / State
:
Zip
:
Telephone/Fax
:
: ........................, ..........................
Signature: ..................................
Order form for subscription should be sent to the editorial office by fax or email
Payment by Bank Transfer Bank Account name (please be exact)/Beneficiary: LINA HANDAYANI Bank Name: CIMB NIAGA Bank Branch Office: Kusumanegara City: Yogyakarta Country :Indonesia Bank Account # : 5080104447117 Bank Swift : BNIAIDJAXXX >>> Please find the appropriate price in the price list on next page >>>
The price list for domestic and foreign subscribers
Volume
Number
Year
1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4 5 5 5 5 6 6 6 6 7 7 7 7 8 8 8 8
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
2012 2012 2012 2012 2013 2013 2013 2013 2014 2014 2014 2014 2015 2015 2015 2015 2016 2016 2016 2016 2017 2017 2017 2017 2018 2018 2018 2018 2019 2019 2019 2019
Price (IDR) for domestic subscribers 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 290.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 319.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00 349.000,00
Price (USD) for foreign subscribers 36 36 36 36 36 36 36 36 40 40 40 40 40 40 40 40 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44
The price included the printing, handling, packaging and postal delivery fees of the hardcopy to the address of the authors or subscribers (with Registered Mail). For foreign subscribers, an additional fee is charged if you would your order is mailed via Express Mail Service (EMS): - $25 for ASIA Continent - $35 for AUSTRALIA Continent - $35 for AFRICA Continent - $39 for AMERICA Continent - $39 for EUROPE Continent (No additional fee for delivering your order by Registered Mail).
