__MAIN_TEXT__

Page 1

INTERNATIONAL SCIENTIFIC AND VOCATIONAL JOURNAL (ISVOS JOURNAL)

Vol.1

Issue:1

Date: December 2017

Received:26.12.2017

Accepted:29.12.2017

Final version : 31.12.2017

International Scientific and Vocational Journal ISVOS Journal Strategies That Transform the Retail José G. Vargas-Hernández a*, María Luisa Anaya Rosas b aUniversity

Center for Economic and Managerial Sciences. University of Guadalajara.Periférico Norte 799 Edificio G-201-7, Núcleo Universitario Los Belenes CUCEAZapopan, Jalisco C.P. 45100; México bMaestría en Negocios y Estudios EconómicosCentro Universitario de Ciencias Económico Administrativas. Universidad de Guadalajara

Abstract The objective of the present investigation is to analyze the strategies that the companies implement and the changes that originate in the retail trade and the retail trade and in the commercial establishments in the developed countries. Based on a literary review, it identifies the framework of the current situation, encompassing all the social products and services that cause the new mechanisms of sale of products and services by the firms and the different ways of buying products and services by Los Consumers Concluding that this variation is mainly present due to the technological level, the culture of the purchase and the strategies of sale, that owns and develops the consumers and the companies. Keywords: “Strategies, retail, commercial establishment, purchase, sale”

1.

Introduction

Companies to maintain their position in the market and seek their expansion, need to create strategies, which consist of aligning or directing their internal resources to modify, adapt and survive environmental conditions (Vargas and Guillen 2014). In a globalized environment, where free trade agreements allow import and export, without the intervention of the government that establishes tariffs, prohibitions or limitations, it has developed a high degree of risk and uncertainty for all companies. Previously, the competition was only local, regional and national, but now the competition is international, in any sector or industry. On the other hand, rapid technological advancement raises and prompts new challenges and opportunities to companies around the world. In countries like Mexico that are in development, are imminent the changes that require the formulation and implementation of business strategies, that establishes mechanisms of adaptation to avoid the loss of competitiveness and the lag, ensuring their development and well-being. This process of globalization and technological progress has as its most obvious manifestation the elevation of the standards of quality of life of individuals, the modification of their way of life and work. Among the many changes that are being presented, the technological advance has developed a new way of selling products and services, through the internet with the already popular electronic commerce or e-commerce as it is also known. Therefore, retail or retail in commercial establishments as it is known in the English language has changed. Companies are increasingly betting on e-commerce, because it facilitates in time and form the relations with customers, and has allowed the design and creating new products, reducing costs and offering cheaper products with better quality and service. The retail trade is any purchase and sales operation directed to the final consumer that purchases the product for personal use or consumption without commercial purposes, also called retail or retail. Performed mainly in establishments or stores in the street (Paz, 2008). Paz also considers that the new demands of the market and competition has necessarily brought a renewal. The traditional small retailer will be reduced in number, but will be important to generate added value in providing services to its customers. Therefore, large companies that are at the forefront of innovation and technology studying the changes, trends, needs and opportunities of markets and customers, constantly seeking to meet consumer demand and increase their profits by modifying their resources and capacities according to the strategists formulated. *

Corresponding author. Tel.: +52(33) 37703340 ; fax: +37703300. E-mail address: josevargas@cucea.udg.mx, jgvh0811@yahoo.com,jvargas2006@gmail.com


2

In this paper, it is analyzed the cases of Bebe Store and Ralph Lauren companies that carried out strategies, which transformed their retail sales into commercial establishments.

2.

Background of the problem

Electronic commerce is any type of commercial transaction carried out through the Internet (Bachs, Lรณpez-Jurado and Yaguez, 2002). This new modality of commerce has modified the way of selling by the companies and to buy by the part of the consumers, neglecting the retail trade in the commercial establishments. The benefits that companies have when using electronic commerce are: A. Lower costs in ordering processes B. Internet as an effective distribution channel C. Reduction of errors in orders and in processing times D. Access to more customers E. There are no geographical barriers to customer access among other according to Bachs, Lรณpez-Jurado and Yaguez (2002). All companies want these benefits. Therefore, they investigate, formulate and implement strategies to include electronic commerce in their sales processes. In such a way, that the companies are leaving behind the retail trade in commercial establishments, diminishing the necessity of personnel to carry out the traditional sales. Retail trade in commercial establishments in developed countries such as the United States of America has undergone major changes, such as the closure of establishments of some commercial chains, caused by the commitment to electronic commerce for the sale and distribution of its products; and by the consumer's buying culture, preferring the internet as a buying and selling channel, living conditions are completely influenced by technology. In countries such as developing Mexico, retail trade in commercial establishments still generates more benefits than electronic commerce, caused by the culture of purchase, where consumers prefer to carry out the traditional purchase of their products and services, physically, to use the Internet.

3.

Delimitation of the problem

The Internet transforms the internal management of companies, therefore, they must prepare themselves for the new technological, political, cultural and economic challenges, that is, of the market and its agents (Bachs, Lopez-Jurado and Yaguez, 2002). As a result, retail trade in commercial establishments has changed, companies are increasingly betting on e-commerce, because it facilitates customer relations and timing, has allowed the design and creation of new products, reducing costs and offering cheaper products and with better quality and service. One of the most important changes that created e-commerce is to present the sales or commercial transactions, in the way of establishing relations between company and consumer is that the exchange is realized without the human intervention (Bachs, Lopez-Jurado and Yaguez, 2002). The retailer according to Paz (2008) is very important within the chain of distribution to the extent that it adds a value to the products it sells, in developing countries like Mexico it is characterized by presenting a trade that does not generate value by excess supply and are doomed to disappear. The changes are socially negative, because the chances of the disappearance of commercial establishments where the largest amount of retail trade is made today are very high. Currently, commercial transactions or sales are being carried out through the Internet. This process in developing countries such as the United States of America is very frequent, but not in developing countries such as Mexico, where consumers still do not fully trust trade electronically, which is developing slowly. Therefore, when closing the establishments, the staff requirement to perform the work of direct sales with the clients, maintenance of the premises, among others, decreases. From the economic point of view, they can be positive, that is to say, for companies to reduce operating costs is one of their best strategies, with retail pay rent of establishments, sales personnel, electricity, maintenance, inventories, among others. With


3

e-commerce these costs and operating expenses are greatly reduced. Consequently, companies continue to sell and their profit margin is higher.

4.

Justification

The reason that led to investigate the effects of the use of electronic commerce in retail or retail, as mentioned by Bachs, LĂłpez-Jurado and Yaguez (2002) is that technological advances impels and constantly causes changes in industrial and economic processes, and consequently in the way of life. One of the technological advances is information technologies. The use of this tool in companies has developed the automation of many business processes with a decrease in costs and an improvement in lead times. This development facilitates the development of business models that involves IT, transforming the retail trade, through telecommunications and information technologies, since they allow operations that can be performed without physical presence, therefore, the place where is carried out the market is no longer determining factor to assess its operation. It is important to analyze where it is and how vulnerable international trade is in all industries. Technology is presenting unprecedented changes, such as the closure of commercial establishments, overriding the need for personnel. Jack Ma, the founder of the Alibaba Group, whose main activity is the largest commercialization of products and services in China, through e-commerce, says a new "seismic movement" is occurring at the technological level, with potentially disastrous consequences. This will provoke social conflicts in the next thirty years, which will have a strong impact on all types of industries, the rise of artificial intelligence and a longer life expectancy will lead to an aging workforce that will fight for fewer jobs.

5.

Theoretical assumption

E-commerce is transforming retail or retail and developing business strategies focused on the closure of establishments and dismissal.

6. 6.1.

Theoretical conceptual framework Strategies

Mention Quintero (2003) that strategic planning is then a constant practice that relates the means or paths (strategies) to the purposes, purposes and results. To properly carry out this practice requires gathering and analyzing information about the internal situation of the organization and what is happening or could occur in its external environment. By identifying the options available to arrive at a desirable and probable outcome, strategic planning contributes in a rational way to improved decision making. It also says that strategic planning has also played an important role in formulating policies, in defining goals and indicators and in allocating resources to programs, projects or activities that are essential in the implementation of strategies or in achieving their goals, purpose, mission, vision, It is therefore also a learning process in which organizations, through successive trials of successes and mistakes, derive important lessons about how to perform and deal with both present and future, their problems, weaknesses, challenges, strengths, threats and opportunities. Therefore, the planning process allows to establish the direction and an environment conducive to an informed and innovative business management within a space delimited by the institutional characteristics and the dynamics of the environment. In addition, this managerial technology requires a systematic and careful adoption that aims at increasing productivity, quality and competitiveness. According to Lopez and Martin (2012) in his book Strategies for Entrepreneurship, to begin they present a very precise definition of strategy by G. Hamel and C.K. Prahalad, which is the following: Business strategy is to create competitive advantages in less time than competitors. The authors mention that the strategic formulations have not changed in their definition, what has changed has been the relative weight of each of them in business practice. Vargas and GuillĂŠn (2014) refer that companies are created and integrated by rational and social individuals, who by nature and instinct tend to seek, learn, create, change and develop their customs in their environment, to improve their quality of life, this raises its evolution; The same


4

happens with companies that are born and are transformed by the needs of individuals to survive, progress and face the adversities of the environment to survive. Therefore, for companies to survive, strategies are needed, defined as follows: it is the alignment or direction given to the internal resources of an organization to modify, lead, adapt and in the worst case to survive the conditions. They consider that the companies make strategic decisions to achieve the established objectives, to implement or to carry out the plans formulated by the executives of the top management, in consideration to the changes of the environment, that have caused the evolution of the processes of strategic transformation of the companies. 6.2.

Resources and Capabilities

In the theory of resources and capabilities from the point of view of Vargas and Guillén (2014) it focuses on the use of resources and capabilities of the company for the development of an effective strategy by companies. It is also mentioned by Peng (2012) in his work global strategy, argue that the main difference between firms lies in the kind of resources and capacities that counts how to accumulate and use them to succeed. On the other hand, a current industrial company defines it as a set of operational units, each with its own facilities and specific personnel, whose combination of resources and activities are coordinated, supervised and assigned by subordinates. Vargas and Guillén (2014) also mention that the development of new technologies and the opening of new markets led to economies of scale and scope and to lower transaction costs. Emphasizing that production and distribution facilities acquired to fully utilize economies of scale and scope lead to the growth of the company. Mahoney (2010) points out that this dynamic theory can become an approach to strategic management that contributes an important part to science in the evolution of the organization. They see the need for an empirical basis to establish both the nature of dynamic capabilities and the impact of dynamic capabilities on sustainable competitive advantage evident. The capabilities that can be especially useful in dynamic business environments are operational flexibility and strategic flexibility. One of the main activities that companies develop is the sale of their products or services. This action is of relevant importance, because it is the one that generates the income that the company needs to survive and where the profits are reflected, for carrying out its preponderant activity. Sánchez and Herrera (2016) believe that throughout history, companies have presented different ways of marketing products and / or services in the customer - supplier chain, in addition to considering their competitors. This has allowed the perspective of the management of resources and capacities, this evolving with the passage of time, caused because it is a factor of yields within the organizations. It is also important to note that in an organization, human resources can be a potential source of sustainable competitive advantage, according to Sánchez and Herrera (as cited in Wright, Dunford & Snell, 2001). They say that all companies stand out in the constitution of human capital, where they assume a shared responsibility to achieve the mission, vision and objectives of the company, through the provision of its resources and capabilities. Consequently, organizations set their actions, according to their commercial activity, to satisfy their clients, seeking to remain in the market. Companies faced with globalization have experienced very sudden and diverse changes that have generated new parameters of competitiveness in different domains, since globalization is conceptualized as the integration of economies (Brunner, 2000). This has led to the pursuit of strategies by companies that help shape a solid organizational performance, to stay and / or survive in this highly competitive global market. 6.3.

Electronic commerce

One of the commonly accepted definitions of electronic commerce is the use of computers interconnected between them to create and transform business relationships between companies and their customers (Bachs, López-Jurado and Yaguez, 2002). These applications are to provide solutions that improve the quality of goods and services, increase the speed of delivery and reduce operating costs. Silva (2009) mentions that in the first instance "e-commerce" or "e-commerce" had the meaning of "electronic purchase" or "online sale", with the passage of time and the constant E-commerce is the process of buying and selling goods and services electronically, through transactions via the Internet, networks and other digital technologies (Silva as cited in Laudon 2002). Another definition that Silva refers to (as cited in McLEOD, 2000) is e-commerce as the use of computers to facilitate all company operations. Most of the operations are internal: as in the areas of finance, human resources, information services, manufacturing and marketing. Then e-commerce is understood as the online sale, it is said that the purchase / sale of products and services is made, also considering the market, e-commerce can interact not only with customers but also with suppliers, investors, competitors and others that will lead to different forms of exchange of products and services; and also to consider other search options and purchase these.


5 On the other hand, Fernández, Sánchez, Jiménez and Hernández (as cited in the Ganga and Águila, 2006) point out that electronic commerce is an innovation that produces effects in organizations, even generating changes that improve organizational structure, management decisions, productivity, effectiveness and competitive advantages, as well as processes through its simplification. Fernández, Sánchez, Jiménez and Hernández (as quoted in Canals, 2001) also consider that innovation in online commerce generates new business opportunities, allowing diversification of traditional sales channels, reducing the cost of business activities and generating new products and services, allowing the company to reorganize its production structure in order to reach more customers. 6.4.

Retail or Retail

The concept of retail is a business management orientation that maintains that the key tasks of a retailer are: a) to determine the needs and desires of its target market and b) to direct the company toward the satisfaction of those needs and desires in a more Efficient than its competitors (Quintero, 2015). They also state that retail is classified according to the activity of the products sold, which corresponds to the economic activity. The classification of the general direction of the domestic trade and other commercial formats is characterized by the products they sell. Retail is a term in the English language of the retail trade, meaning any purchase and sales operation directed to the final consumer that purchases the product for personal use or consumption without commercial purposes, also called retail or retail, performed mainly in establishments or stores in the street (Paz, 2008). Cavalcante and Akemi (2015) refer that retail definit ions seem to characterize it as a mere place for product supply. On the other hand, it is a place where value, objective and subjective exchanges are carried out. On the one hand, the client receives physical products, services, convenience and various experiences and, on the other hand, pays with his time, money and energy. Therefore, the benefits that are obtained in the purchase must be greater than the costs or sacrifices related. A great potential for competitive gain for retailers is the creation of situations and incentives that increase the perceived benefits of the customer during the act of purchase. In retail, it is necessary to understand the buyer, who is the role played by a person in the buying process, who chooses where and what to buy, once they decide to go to a physical store, online or interact with other channel purchases. The buying behavior is different from the consumer behavior, since its motivations, the influence of the environment, the answers and the results obtained through this process are peculiar and influenced by the conflict of comparison and choice. They consider that consumer behavior and the buying process during leisure time have changed due to social, economic and cultural factors, generating new patterns of behavior and new demands for products and services (Molina, García and Gómez, 2011). They focus mainly on the economic aspect of the purchasing process; Where it satisfies the customer, as well as other elements related to the act of purchase (quality of service, variety and price of products, purchase preferences and spending of tourists in destination). They also include elements such as the experiential and intangible values of establishments or in shopping centers. However, these investigations do not establish the factors that could actually condition the behavior according to the perceptions obtained in commercial establishments. Buyers participating in a shopping experience are immersed in a number of activities, in addition to the purchase of the product or service.

7.

Contextual framework

A company must be dynamic in all its operations. This is required by belonging to a dynamic environment, in which it is unknown how the influence of external variables mainly and their consequences internally. Consequently, the uncertainty and economic instability of the environment and of the organizations is multiplied, automatically generating a difficult environment to create and develop competitive strategic maneuvers (Dess and Beard, 1984). This complicated environment at the global level, has led to the creation and establishment of business strategies with high social and economic impact. Some cases are presented below, in the United States a developed country. 7.1.

Bebe Store case

Bebe Store Women's clothing chain, has announced the closure of all its stores and inventory liquidation by the end of May 2017. The company mentioned that it is studying strategic alternatives for restructuring the business to focus on internet sales or e-commerce, after of having losses of approximately 20 million dollars. The Bebe Store adds to the long list of retail victims in the United States, taking the painful decision, as it recently did Ralph Lauren, to close all its physical sales points to avoid the deep crisis that the retail lives.


6

First in the month of March 2017 announced the closure of 12% of its stores, after seeing that the sales of the first quarter of the year fell and registered losses of 5 million dollars. It concluded and informed that closed the entire network of 175 stores in the United States in a developed country. 7.2.

Ralph Lauren case

High-end fashion retailer Ralph Lauren said on Tuesday, April 2, 2017 that it would cut jobs and shut down its Polo store on Fifth Avenue in New York, seeking to cut costs. The company declined to mention details, although it had already announced 50 store closures during the fiscal year ending March 31. Bomey (2017) CEO of the company said that these changes would save $ 140 million in annual expenses and cost $ 370 million in unique restructuring charges. The plans include an unspecified number of store closures, a reduction in the workforce and the closing of certain corporate operations, according to its public statement. The strategy to be implemented, which the company called the Way Forward Plan, includes the restructuring of its current digital operations and the switch to a more profitable and flexible e-commerce platform in an agreement with the software provider Salesforce, according to "Together, these actions are an important part of the Company's efforts to achieve its declared goal of returning to sustainable and profitable growth and investing in the future," said Ralph Lauren. The new strategy includes exploring "new retail concepts," such as the Ralph's Coffee brand, and "developing new store formats," the company said. The company has suffered from a heavy dependence on department store business, namely including its deal with Macy, and insufficient online business. Other headaches included swollen inventory, frequent brands, too many businesses in too many outlet stores, and a new generation of customers who act, think, communicate and dress differently than previous generations of the brand, wrote Jane Singer, in February for a retail analysis The Robin Report. The transformation of retail into commercial establishments, i.e. sales physically, is evolving very fast in developed countries as in the United States; Companies are formulating strategies to close their stores, thus creating uncertain scenarios for shopping malls, where hundreds of stores are set up for retail sale. 7.3.

Malls

Traditional retailers are in a critical situation of high risk and uncertainty, as stores close and e-commerce giants like Amazon.com Inc. are on the rise. Analyzing this scenario arise a series of questions What will happen with the malls? How will they survive? Or are they going to disappear? As a consequence of the sharp strategies of the commercial chains in closing the stores. Bloomberg (2017) mentioned that earlier this year a REIT Symposium was held at New York University, which was attended by executives from different companies in the United States to address the issue of the retail or trade crisis retail and the uncertain scenario of shopping centers. In the symposium they said that the warnings of deaths should be discarded, at least for a category. Shopping malls are not going to go extinct, but are changing, sometimes by a healthy natural selection. It is common knowledge that store closures are painful, often depicting a Darwinian moment. In addition, they mentioned that shopping malls are not the same as open-air shopping centers and, more importantly, high-quality malls - or "A" class - are different from "B" and "C" . According to the consensus, high quality malls will perform better. Mathrani, CEO of shopping center operator GGP Inc. commented, "If you have the best real estate, you will thrive," quality is what matters, and says that the ideal mall today includes a department store, a grocery store, an Apple store, A Tesla agency and a business that started online, such as Warby Parker stores. Kenneth Bernstein, CEO of Acadia Realty Trust, said "Food is the new fad, and physical fitness is the new food." They also concluded that shoppers are now not as willing to travel as far as they once were for a day at the mall, so location and convenience are also important. "Outdated shopping malls will continue to exist, provided they are in the right location."

8.

Method

In the present investigation the qualitative method based on the literary revision is used, starting from analyzes realized of companies in the United States, in order to explain and describe the phenomenon under study.

9.

Conclusions and recommendations

After analyzing some strategies that are being carried out by large companies in the United States of America, it is considered that electronic commerce is a process for the commercialization of products and services at a global level, which was developed


7

thanks to the internet and constant technological progress. Therefore, it is an indispensable element or factor in the formulation of sales strategies in any company, which has significantly modified the retail in the commercial establishments, mainly reduces the need for personnel in the sales process. These new technological elements are transforming the way companies sell and how consumers buy, therefore, it must also be kept in mind the positive or negative effects that are being presented, which are the result of our evolution. Therefore, it is indispensable to be prepared to act in accordance with the new economic structures.

References Bachs, F., López-Jurado, M y Yaguez M. (2002). Internet, comercio electrónico y plan de negocio. España. Deusto planeta de Agostini Profesional y Formación S.L. Bloomberg, M. (8 de abril del 2017). Desaparecerán los centros comerciales. El Financiero. Recuperado de: http://www.elfinanciero.com.mx/empresas/desapareceran-los-centros-comerciales.html. Bomey, N. (4 de abril del 2017). Ralph Lauren cerrará almacenes, oficinas en esfuerzo de vuelta. USA Today. Recuperado de: https://www.usatoday.com/story/money/2017/04/04/ralph-lauren-restructuring/100016380/ Brunner, J. (2000). Globalización y el futuro de la educación: tendencias, desafíos y estrategias. Recuperado de: http://www. schwartzman.org.br/simon/delphi/pdf/brunner.pdf Cavalcante, B. y Akemi, A. (2015). The Value for the Consumer in Retail BBR - Brazilian Business Review, vol. 12, núm. pp. 46-65 FUCAPE Business School Vitória, Brasil. E-ISSN: 1807-734X. Dess, G. G., & Beard, D. W. (1984). Dimensions of organizational task environments. Administrative science quarterly, 5273. El Economista. (21 de abril del 2017). La última víctima del retail en EEUU: Bebe Store cierra todas sus tiendas y liquida el stock. Recuperado de: http://www.eleconomista.es/distribucion/noticias/8308753/04/17/La-ultima-victima-del-retail-en-EEUUBebe-Store-cierra-todas-sus-tiendas-y-liquida-el-stock.html Fernández, A., Sánchez, M., Jiménez, H., Hernández, R. (2015). La importancia de la Innovación en el Comercio Electrónico. Universidad Business Review, núm. 47, julio-septiembre, 2015, pp. 106-125 Portal Universia S.A. Madrid, España. ISSN: 16985117. López, C. y Matin, A. (2012). Estrategias empresariales. Segunda Edición. España. Ecoe Ediciones. Mahoney, J. (2012). Economic Foundations of Strategy. Thousand Oaks, CA: Sage. Molina, A., García, J. y Gómez, M. (2011). Elementos clave para el comercio minorista de un destino. Universia Business Review, núm. 29, 2011, pp. 80-99 Portal Universia S.A. Madrid, España. ISSN: 1698-5117. Moliner, M., Sánchez, J., Callarisa, L., y Rodríguez, R. (2008). La calidad de la relación: un concepto emergente. El caso de un establecimiento comercial. Cuadernos de Economía y Dirección de la Empresa. pp. 97-121 Asociación Científica de Economía y Dirección de Empresas Madrid, España. ISSN: 1138-5758. Paz, H. (2008). Canales de distribución: gestión comercial y logística. Buenos aires. Argentina. lectorum-Ugerman. Peng, M. W. (2012). Global Strategy. Cincinnati: Thomson South-Western Quintero, L. (2015). El sector retail, los puntos de venta y el comportamiento de compra de los consumidores de la base de la pirámide en la comuna 10 de la ciudad de Medellín. 2015. Revista Ciencias Estratégicas, vol. 23, núm. 33, enero-junio, 2015, pp. 109-118 Universidad Pontificia Bolivariana Medellín, Colombia. ISSN: 1794-8347. Ramírez, C. y Alférez, L. (2014). Modelo conceptual para determinar el impacto del merchandising visual en la toma de decisiones de compra en el punto de venta. Pensamiento & Gestión, núm. 36, pp. 1-27 Universidad del Norte Barranquilla, Colombia. ISSN: 1657-6276. Sánchez, J. (2003). Estrategia integral para pymes innovadoras. Revista Escuela de Administración de Negocios. Universidad EAN Bogotá, Colombia.


8 Sánchez, S. y Herrera, M. (2016). Los recursos humanos bajo el enfoque de la teoría de los recursos y capacidades. Revista Facultad de Ciencias Económicas: Investigación y Reflexión, vol. XXIV, pp. 133-146 Universidad Militar Nueva Granada Bogotá, Colombia. ISSN: 0121-6805. Silva, R. (2009). Beneficios del comercio electrónico perspectivas. Universidad Católica Boliviana San Pablo Cochabamba, Bolivia. ISSN: 1994-3733. Vargas Hernández J. G. y Guillen Mondragón I. J. (2005). Los procesos de transformación estratégica en relación con la evolución de las organizaciones, Estudios Gerenciales. No 94, Enero-marzo del 2005. Universidad ICESI. 65-80. ISSN: 0120 6648.


INTERNATIONAL SCIENTIFIC AND VOCATIONAL JOURNAL (ISVOS JOURNAL)

Vol.1

Issue:1

Date: December 2017

Received:26.12.2017

Accepted:29.12.2017

Final version : 31.12.2017

International Scientific and Vocational Journal ISVOS Journal IMPACT OF BUSINESS PERFORMANCE AND TQM ON THE SMEs OF MEXICO José G. Vargas-Hernández a1, Vanessa Yesenia b, Vázquez Lermab, a Research

Professor, Administration Department University Center for economic and Managerial Sciences, University of Guadalajara Periférico Norte 799 Edif. G201-7 Núcleo Universitario Los Belenes, Zapopan Jalisco b Universidad Autónoma De Sinaloa Zona Norte, Los Mochis Pról. Ángel Flores y Justicia Social s/n Col. Jiquilpan, 81220, Ahome, Sinaloa, México

Abstract The purpose of this research is to analyze the degree of effectiveness of TQM in SMEs in Mexico as a function of Business Performance, considering that Business Performance is related to Total Quality Management. This research is analytical and descriptive, since no statistical method is used, only the variables and the research problem are analyzed in detail and in a detailed manner, and descriptive because tables and tables can be found to help facilitate understanding of these important concepts. This research will analyze the impact of both variables, and with this, the entrepreneurs can make the necessary changes in their production line or in their employees to obtain better and greater positive results. Keywords: “Business Performance, Total Quality Management (TQM), SMEs, Effectiveness”

1.

Introduction

Nowadays the competitiveness in the market is very high. Every day companies seek to face this situation through the use of new techniques that increase their benefits and improve corporate performance (Alfalla Luque, 2012, pp. 64-88). It is necessary to realize that innovations drive the business and that they are considered as tools to maintain competitiveness. All innovation must contribute to the creation of added value for the client and for the company. The importance of this research is to find a method or tools that as part of daily use in organizations can have a presence in the market, generating new standards of competitiveness, and business performance. The hypothesis of this research is that Business Performance is related to Total Quality Management (TQM). With these data, it can be started to develop in a deeper way what each of these concepts consists of, considering that the objective is to analyze the degree of effectiveness of TQM in Mexican small and medium enterprises (SMEs) based on business performance. Among the main authors of this research are Cruz Álvarez and Feizollahi and Giménez Espín, considered experts in productivity and quality, and some recognized theories of the parents of the administration such as Frederick Taylor or Henry Fayol. This investigation is of analytical and descriptive type, because in it are the main concepts on the investigation, as well as the main authors and creators of the same, and descriptive. Also, in it are tables that help to explain in a simpler way for its correct understanding. In view of the results and recommendations of this scientific research, SMEs in Mexico should implement more and better these tools because with them they can provide better services to their clients and with this the company’s growth more. This research would not have been possible without the support of my family who gave their emotional support and financial resources. To my boyfriend who was always supporting me to attend this summer course, and of course the PhD. José G. Vargas Hernández who accepted me as his student at the University of Guadalajara, in the University Center of Economic and Administrative Sciences, in this summer of scientific research to enrich my scientific knowledge and finally to the Autonomous

1

Corresponding author. Tel.: +523337703340 E-mail address: jvargas2006@gmail.com


10

University of Sinaloa through the Academic Business Unit who provided me with a small financial support to make my summer scientific stay.

2.

Background of the problem

Mexican SMEs are constantly looking for better and effective methods to ensure their persistence in the environment in which they develop. Since the demands of the markets change and become increasingly demanding, that is why new standards of competitiveness have been generated that practically they force companies to be better and better at what they do. The current business situation requires incorporating tools that are part of the daily use of organizations and promote the success of them. The Management of the Total Quality is tied with the good performance of the production and the behavior of the client. Production performance can be achieved through good relationship with suppliers, benchmarking, quality improvement and continuous improvement of the process. The measurement of quality is the most important task, followed by comparative evaluation, continuous process improvement and relations with suppliers. In turn, the yield of production has a positive effect on the behavior of the clients, which results in a good corporate performance (Agus & Latifaah, 2000), (Feizollahi, 2013; Smith, 2014). The performance of the organizations is reflected by the increase in sales, market share and the presence of the brand (Cruz Ă lvarez, 2014, pp. 127-142). The evaluation of performance is a periodic determination of the operational effectiveness of an organization, part of the organization and its employees by objectives, standards and criteria established in advance (Goentoro, 2016, pp. 93-96).

3.

Delimitation of the problem

The present research is carried out to the SMEs of Mexico, to achieve the presence of a brand, reflect sales increase and achieve the level of competence. The focus of this research is the elements that are important to achieve business performance in SMEs in Mexico, as they are competitive performance, financial performance and service quality. The main limitation is that most of the SMEs in Mexico do not have the knowledge of the tools or processes that exist to be able to carry out the good functioning of the company or to apply them for daily use to obtain great results, such as being the avantgarde with products, achieving quality processes, and achieving a competitive standard that leads them to achieve business success. In other cases, it is known about these tools but managers do not have the knowledge of how to apply them to achieve the results expected. The current society is going through constant changes that require new institutions challenges and goals in their daily work. It is a revolution where quality, information and knowledge are shown as essential resources for the guarantee of a correct institutional performance. However, owning these resources is only a step to be in correspondence with the new demands of changing environments (PĂŠrez, 2007, pp. 71-76).

4.

A.

What is the degree of effectiveness of TQM in SMEs in Mexico based on business performance?

B.

What is the degree of effectiveness of TQM in SMEs in Mexico in terms of personnel management?

C.

What is the level of effectiveness of TQM in SMEs in Mexico in terms of continuous improvement?

D.

What is the level of effectiveness of TQM in SMEs in Mexico based on leadership?

Justification

The objective of this research is to analyze the degree of effectiveness that exists between Business Performance and Total Quality Management in SMEs in Mexico. Currently, companies are obliged to contribute new ideas, products or services to the market. If companies do not update their products, the products would become unattractive and the company will start to have serious problems. It is necessary to realize that innovations drive the business and that they are considered as tools to maintain competitiveness. All innovations must contribute to the creation of added value for the client and for the company. The importance of this research


11

is to find a method or tools that as part of daily use in organizations can have a presence in the market, generating new standards of competitiveness and business performance

5.

Research variables and hypotheses A. Dependent Variable (X): Business Performance B. Independent Variable (Y): Total Quality Management. Table 1. Description of research variables, dimensions and indicators

VARIABLE

Xo

DESCRIPTION It is the quantitative and qualitative result obtained by the company in a given period. It can have positive or negative effects, since they can be affected by its social and environmental performance (Oliveira, 2013: 131-167).

Business

DIMENSIONS Competitive performance Financial performance

Yo Total Quality Management (TQM)

Source: Own elaboration.

Improvement of processes and products

Increase

of

utilities

Quality in the service quality

Attracting new customers and retaining existing ones

Staff management

Participation, Training, Team Information and Analysis.

Performance

TQM is an integrated manufacturing system aimed at continuously improving and maintaining quality products and processes through business management, human resources, suppliers and customers in order to achieve and even exceed the expectations and needs of customers (Hackman and Wageman, 1995, Powell, 1995, Cua, McKnoe and Schroeder, 2001: 309-349).

INDICATORS

Continuous improvement Leadership

Commitment


12 Figure 1. Research construct

Source: Own elaboration A.

General hypothesis

Business Performance is related to Total Quality Management (TQM). B.

Specific hypotheses

Competitive Performance is related to Personnel Management. Financial Performance is related to Continuous Improvement. The quality in the service is related to the Leadership.

6.

Research objectives A.

General objective

Analyze the degree of effectiveness of TQM in SMEs in Mexico based on Business Performance. B.

Specific objectives

1) Analyze the degree of effectiveness of TQM in SMEs in Mexico in terms of Personnel Management. 2) Analyze the degree of effectiveness of TQM in SMEs in Mexico in terms of Continuous Improvement. 3) Analyze the degree of effectiveness of TQM in Mexican SMEs based on Leadership. Table 2. Congruence matrix

General question

Specific questions

General objective

1. What is the degree of effectiveness of TQM in SMEs in Mexico in terms of Personnel Management? What is the degree of effectiveness of TQM in SMEs in Mexico based on Business Performance?

2. What is the level of effectiveness of TQM in SMEs in Mexico in terms of Continuous Improvement? 3. What is the level of effectiveness of TQM in SMEs in Mexico based on Leadership?

7. 7.1.

Specific objective

1. Analyze the degree of effectiveness of TQM in Mexican SMEs in terms of Personnel Management. Analyze the degree of effectiveness of TQM in Mexican SMEs based on Business Performance

2. Analyze the degree of effectiveness of TQM in SMEs in Mexico in terms of Continuous Improvement. 3. Analyze the degree of effectiveness of TQM in Mexican SMEs based on Leadership.

Conceptual and theoretical framework Conceptual framework

In order to define the variables of the research, the first step is to define the basic concepts that are necessary to better understand the research, first define the variable (x), according to classical authors, then authors of state of the art.


13

7.1.1. Business performance Business performance is the quantitative and qualitative result obtained by the company in a given period. Example of a qualitative result is the evaluation of the performance of the organizations through the opinion of the clients, and in quant itative terms it is the evaluation of the profitability. Then business performance can have positive or negative effects, since they can be affected by their social and environmental performance (Lopes de Oliveira, 2013, pp. 131-167). Business performance is measured based on three dimensions: 1) economic, 2) social and 3) environmental (Gรณmez, 2013, pp. 1-35). Business Performance proposes to measure business performance through three blocks: market measures, measures based on accounting and measures based on the perception of managers, whatever their classification, these have positive or negative influence on business performance (Orlitzky , 2003, pp. 403-441). 7.1.2. Total Quality Management (TQM) To learn more about the dependent variable in the first order, the various concepts cited by classical authors are presented, and secondly, the concepts cited by authors of the state of art are listed. Total Quality management defines it as a collection of certain activities related to quality: a)

Quality becomes part of the plan of all senior management.

b)

Quality goals are incorporated into the business plan.

c) The expanded goals derive from benchmarking: the emphasis is on the consumer and the competition; there are goals for the annual improvement of quality. d)

The goals of deploy to the action levels.

e)

Training is carried out at all levels.

f)

The measurement is made in each area.

g)

Managers regularly analyze progress with respect to goals.

h)

The superior performance is recognized.

i)

The reward system is rethought (Juran, S.f., s.p.).

The total quality management is defined as the management body is totally committed to quality: The client's requirements are understood and assumed exactly. Total: Every member of the organization is involved, including the client and the provider, when this possible (Ishikawa, 1990). Total Quality Management (TQM), is a very discussed technique in the business world and is considered as a business practice that has been identified as a type of innovation, mainly organizational innovation that helps improve corporate performance (Bernardo, 2014, pp. 132-142; Feizollahi, 2013, page 1879; Yong Lam, 2014, pp. 106-111). Total Quality Management (TQM), is a business philosophy that has become popular internationally in many business areas (Kanji, 1990). All the tools and techniques presented are only means to achieve the objectives of continuous improvement through Total Quality Management (Khalid, 2011, p.p.). Among the main benefits of TQM can be found the improvement of quality, employee participation, teamwork, better working relationships, customer satisfaction, employee satisfaction, productivity, communication and participation in the market, leadership management, strategic planning, process management, financial performance and market performance (Ahmad, 2012; Khalid, 2011; Yong Lam, 2014). TQM is sustained in the commitment of the entire organization with customer satisfaction, continuous improvement of products and processes, teamwork and assignment of responsibilities (Agus & Latifaah, 2000). It is understood that TQM favors the adjustment of business activity with all the relevant agents for the organization (Oakland, 2000) so that it improves the results and the competitive position of organizations in complex and dynamic environments such as the current ones (Samson, 1999). 7.2.

Theoretical framework

In the first order, the theories that talk about Business Performance are listed, and the importance they have for organizations.


14 Table 3. Review of business performance theories

Theory

Principles of scientific administration

Classical Theory of Administration

The theory x and the theory.

The theory of resource dependence

Author

Principles

Frederick W. Taylor

The methods of science to the problems of administration to obtain high industrial efficiency. The main scientific methods applicable to the problems of administration are observation and measurement (Taylor, 1973).

Henry Fayol

It was very important to both sell and produce, to finance themselves and to secure the assets of a company. In short, the organization and its components were considered as a large interdependent system, as internal customers. Fayol created favorable scenarios for administrative efficiency and, therefore, for the generation of profits for the company (Fayol, 1976).

Douglas Murray Mcgregor

Douglas Murray Mcgregor Theory X: It is based "on the old model of threats and the presumption of mediocrity of the masses, it is assumed that individuals have a natural tendency to leisure and that work is a form of punishment" (McGregor, 1960, p. 134-144). Theory Y: "Considers that their subordinates find in their employment a source of satisfaction and that they will always strive to achieve the best results for the organization, thus, companies must release the skills of their workers in favor of said results" (McGregor , 1960, pp. 134-144).

Aldrich y Pfeffer

Also, both the ability to acquire the necessary resources and the efficiency in the use of the resources of an organization are important tools of judgment for the effectiveness of the organization and operation. However, measuring organizational performance is explicitly or implicitly related to what the organization achieves, its goals or objectives (Aldrich, 1976, pp. 79-105).

Source: Own elaboration. In second order, the theories that talk about total Quality management (TQM), and the principles of each of them are listed. Table 4. Review of theories of total quality management

Theory

Author

Principles Quality starts with education and ends with education. Those data that do not have scattered information (variability) are false. The first step towards quality is to know the needs of customers. The ideal state of quality control occurs when inspection is no longer necessary.

Theory of the quality of Kaoru Ishikawa

Kaoru Ishikawa

Eliminate the root cause and not the symptoms. Quality control is the responsibility of all workers in all divisions. Do not confuse the means with the objectives. Put quality first and focus your eyes on long-term profits. Marketing is the entry and exit of quality.


15

Senior management should not show anger when their subordinates present the facts. 95% of the problems of a company can be solved with simple analysis tools "(Ishikawa, 1990). General theory and practices of quality management of Philip B. Crosby

14 points to boost the management of quality and the seven diseases of William Edwards Deming

Theory of quality Joseph M. Juran

Philip B. Crosby

Consequently, inspection, experimentation, supervision and other nonpreventive techniques have no place in this process. Statistical levels of compliance with specific standards induce staff to fail. Crosby argues that there is absolutely no reason to make mistakes or defects in any product or service (Crosby, 1987).

William Edwards Deming

Quality does not mean luxury. Quality is a degree of uniformity and predictable reliability, low cost and adapted to the market. In other words, quality is everything that the consumer needs and longs for. Since the consumer's needs and wishes are always changing, the way to define quality with reference to the consumer is to constantly redefine the requirements (Deming, 1989, p.p.).

Joseph M. Juran

A form of quality is income oriented, and consists of those characteristics of the product that satisfy consumer needs and, as a consequence, produce income. In this sense, a better quality usually costs more. A second form of quality would be cost oriented and would consist of the absence of faults and deficiencies. In this sense, a better quality generally costs less "(Juran, 1990, p.p.).

Source: Own elaboration 7.3.

Empirical review of the literature Table 5. Empirical review of Business Performance according to some authors

Authors (Years)

Tittle of the research

Context

Method and instrument used

Results and findings

(Hochsztain, 2015)

Success factors of an enterprise: An exploratory study based on Data Mining techniques.

A case study is presented based on data from a survey of participating entrepreneurs

A study based on the data of a survey.

They show that the two more relevant elements to anticipate the success of an enterprise are to have financing and, previously, the employment situation of the entrepreneur is an independent worker.

Structural equations,

The results achieved make a contribution to the state of affairs, characterized by the lack of a critical synthesis of theoretical contributions and controversial empirical results.

of the program, applying classification. (Camisรณn Zornoza, 2007)

Competitive strategies and business performance: Comparative study of Robinson's models Pearce and miles & Snow in the Spanish hotel sector.

Source: Own elaboration.

Hotels of Spain

Exploratory factor analysis and ANOVA.


16

Table 6. Empirical review of total quality management according to some authors.

Authors (Year)

(JesĂşs Perdomo, 2011)

(Samat, 2004)

Tittle of research

Context

Human Resources Management Focused on Total Quality and Innovation.

The population was taken from the Mercantile Registry for the Bogota Region and was constituted by 357 companies, of which a response rate of 28.29% was obtained, that is, 101 questionnaires validated.

The relationship between total quality management (TQM) practices, service quality, market orientation, and organizational performance

Managers of 175 service organizations in the northern region of Malaysia (Kedah, Perak, Penang and Perlis), and only 101 were returned.

Method or instrument used

Results and findings

Questionnaires and ANOVA

The study found equivalent and complementary results to those obtained from previous work. In the first place, he confirmed that the practices associated with the dimension of teamwork are directly associated with the best results of innovation. And second, it was not evident that the dimensions of training and motivation were those that contribute the most significantly to the results of innovation, contrary to that found in other investigations.

Questionnaire using the SERVQUAL model

They show that only empowerment of employees, information and communication, continuous improvement focus had a significant effect on the quality of service, employee empowerment and customer focus had an orientation effect. Both the quality of the service and the orientation to the market were organizational performance, however; The relationship between TQM practices and organizational performance did not mediate.

Source: Own elaboration

8.

Contextual framework

There are 4 million 15 thousand business units of which 99.8% are SMEs that are responsible for generating 52% of the Gross Domestic Product (GDP) and 72% of jobs" (INEGI, 2015, page s.p.). The following table shows the number of companies by size and the share of each, highlighting that small and medium-sized companies have a very low percentage. Table 7. Number of companies by size

Source: Own elaboration based on data from INEGI (2015). Next in Figure 2 we can see the distribution of SMEs according to the main limitations of why they do not want their businesses to grow.


17

Figure 3 shows the distribution of the number of companies according to the actions they implemented before problems presented in the production process, highlighting that the greatest limitation is continuous improvement, that is why the TQM tool must be implemented, improve the processes and services of SMEs. Figure 2. Distribution of SMEs according to the reason why they do not want to grow.

Distribution of SMEs according to the main reason why they do not want their businesses to growth 2015 Paga mas impuestos 10% 7%

19%

Falta de informacón de técnicas Falta de conocimiento de implementación de técnicas satisfecho con su empresa

21%

43%

Source: Prepared by the authors, based on data collected by INEGI (2015). Figure 3. Distribution of the number of companies according to the actions they implemented before problems presented in the production process, 2015.

Distribution of the number of companies according to the actions they omplemented before problems in the production process, 2015

19 44

35

Sin acciones en solución Mejora continua Poco conocimiento para la aplicación de herramientas Source: Prepared by the authors, based on data collected by INEGI (2015).


18

9.

Research methods

9.1.

Types of research

This research is analytical and descriptive, because it establishes the comparison of the variables and groups of control studies and includes the main concepts about the research, as well as the main authors and creators of the same, and descriptive because in The research includes tables and tables and pie charts that help explain the variables in a simpler way for the correct understanding of these concepts. 9.2.

Research design Table 8. Description of the variables, dimensions, indicators, instrument operationalization of variables and statistical analysis of research. Variables

X0 Business Performance

Y0 Total Quality Management (TQM)

Description

It is the quantitative and qualitative result obtained by the company in a given period. It can have positive or negative effects, since they can be affected by its social and environmental performance (Oliveira, 2013: 131-167).

TQM is an integrated manufacturing system aimed at continuously improving and maintaining quality products and processes through business management, human resources, suppliers and customers in order to achieve and even exceed the expectations and needs of customers (Hackman and Wageman, 1995, Powell, 1995, Cua, McKnoe and Schroeder, 2001: 309-349).

Dimensions

Indicators

Competitive performance

Quality Improvement of Processes and Products

Financial Performance

Increase of Utilities

Quality in the Service

Attracting new customers and retaining existing ones.

Staff management

Participation and development team

Continuous improvement

Information and analysis

Leadership

Commitment

Instruments

Descriptive and bibliographical analysis

Operationalization of variables

Information is collected from recognized authors and databases of scientific journals, to contrast with Total Quality Management.

Descriptive and bibliographical analysis

Information is collected from recognized authors and databases of scientific journals,to contrast with Business Performance.

Statistical analysis

Tables figures Circular graphs

Tables figures Circular graphs

Source: Own elaboration. 9.3.

Research instrument

In this research we have used databases such as Scopus, EBSCO, INEGI and scientific journals such as Redalyc. Authors such as Cruz Álvarez (2014), Feizollahi (2013), Giménez Espín (2014), alfalla Luque (20129, Agus (2011, 2010), Ahmad (2012), Yong Lam (2014), Bernardo (2014), Camisón (20079, were analyzed to find the impact and relationship between the variables. 9.4.

Data analysis

This research is analytical and descriptive, because it contains the main concepts about the research, as well as the main authors and creators of the same, and descriptive because in the research there are tables that help to explain in a certain way and simpler for the correct understanding of these concepts. 9.5.

Limitations

This investigation has as main constraints the time, there was very little time to develop in a more profound and concrete way the research as the second limitation is the economic resource because it was limited, there was no total access to the information, and the sites There are very few reliable websites to obtain scientific articles that give sustenance to the research.


19

10. Analysis of results 10.1. Testing hypotheses and results The hypothesis of this research is the relationship that business performance has with total quality management. The result was that they have a direct relationship, since successful companies implement these techniques and tools that contribute to better competitiveness and business success, the variables of this research go hand in hand as TQM seeks the continuous improvement of products and service and therefore the performance in the market and thus achieve business success. 10.2. Contrasting hypothesis This research corroborates what Yong Lam (2014) finding consider that TQM is the main tool for business performance, so it must have solid foundations that, together with the other elements, help and complement business performance (pages 106 -111). TQM is established as an essential tool that contributes to business performance to formulate a model of success in SMEs in Mexico. 10.3. Findings The finding in this research is that TQM is a very important technique in international organizations, but it does not have the same impact in small and medium enterprises, due to the lack of knowledge about this technique and the implementation of this tool in SMEs. It should be considered cultural and institutional factors. 10.4. Contributions This research provides knowledge to entrepreneurs of SMEs that today the business environment is very complex and competitive that is why large organizations recommend SMEs to invest in TQM systems since investment helps to transform into a market oriented 10.5. Implications This research can serve entrepreneurs of SMEs in Mexico, so that they are guided by the techniques and tools that exist to be successful and competitive in the market that is so demanding that there is a day to day, the importance of each One of them in their companies, knowing how to implement it will achieve great results, and with it a great satisfaction. 10.6. Future research This work is focused on entrepreneurs of SMEs who are unaware of the techniques and how their correct implementation should be, facilitating this knowledge through this scientific research. Providing the most relevant features so that they are of great use and these can be reflected in a positive way in their companies. 10.7. Research limitations This research presents limitations such as the time since it was too little to be able to delve more deeply into the aforementioned variables, however, enough results were obtained to verify the hypothesis of this investigation, in turn, the information collection was lacking since the databases were in the library of the university and this had a limited schedule, and finally on the websites were consulted scientific journals but in them the information was limited.

11. Conclusions and recommendations Within this research it is important to point out that the Business Performance has a direct relationship with Total Quality Management being this way that SMEs in Mexico should better train their employees so that they can implement these two important tools in a correct way and make the daily use and these can provide significant results, since companies nowadays must innovate and be competitive, and always look for a more efficient way to carry out their operational processes by implementing new strategies to obtain better changes.


20

The majority of SMEs do not implement these tools for several reasons, but the ones found in this research were the lack of information on the existence of said tools, the lack of knowledge on how to implement them. The recommendations offered by this research is that entrepreneurs of SMEs in Mexico, are given the opportunity to learn a little more about these tools that today is a very important technique in global organizations, some of the benefits of implementing these Tools are: better relationship with suppliers, improvement of quality and continuous improvement of processes, to achieve a good performance of production and its relationship with customer behavior, achieving business success.

References Agus, A. Hassan, Z. (2011). Enhancing production performance and customer performance through Total Quality Management (TQM): Strategies for competitive advantage. Procedia. Social and Behavioral Sciences, 24,, 1650-1662. Agus, A. K., Latifaah, S. (2000).The structural impact of total quality management on financial performance relative to competitors through customer satisfaction: a study of Malaysian manufacturing companies. Total Quality Management, 814-819. Ahmad, M. F. (2012). Relationship of TQM and Business Performance with Mediators of SPC, Lean Production and TPM. Procedia . Social and Behavioral Sciences, , 186-191. Aldrich, H. Pfeffer, J. (1976). "Enviorment and organizations". Aunual Review of Sociology, 2., 79-105. Alfalla Luque, R. M. (2012). Is worker commitment necessary for achieving competitive advantage and customer satisfaction when companies use HRM and TQM practices? . Universia Business Review, 36, 64-88. Bernardo, M. (2014). Integration of management systems as an innovation : a proposal for a new model. Journal of Cleaner Production, 132-142. Camisón Zornoza, C. G. (2007). Estrategias competitivas y desempeño empresarial: estudio comparativo de los modelos de robinson & pearce y miles & snow en el sector hotelero español. Investigaciones Europeas de Dirección y Economía de la Empresa, 161-182. Crosby, P. B. (1987). La calidad no cuesta, el arte de asegurar la calidad. Mexico: Continental. Cruz Álvarez, J. G. (2014). Total Quality Customer Satisfaction Model. . CBU International Conference Proceedings,1-4. Cruz Álvarez., B. J. (2014). Aproximación Teórica para el Diseño de un Modelo Integral de Satisfacción de Cliente. INGENIARE, Universidad Libre-Barranquilla, 127-142. Deming, W. E. (1989). Calidad, productividad y competitividad. La salida de la crisis. Madrid: Diaz de Santos. Fayol, H. (1976). Administracion industrial y general. México: Herrero Hermano Sucs S.a,. Feizollahi, S. S. (2013). The investigation of relationship between organization strategy, total quality management (TQM) and organization performance. . Advances in Environmental Biology, 7(8),, 1879-1885. Giménez Espín, J. A. (2014). La gestión de calidad: importancia de la cultura organizativa para el desarrollo de variables intangibles. Revista Europea de Dirección Y Economía de La Empresa, 23(3),, 115-126. Goentoro, E. (2016). Strategy implementation as intervening for company's resources and regulations in order to form business performance [studies in unit businesses of pt pertamina (PERSERO)]. CLEAR International Journal Of Research In Commerce & Management, 93-96. Gómez, P. U. (2013). La relación entre responsabilidad social empresarial y desempeño financiero. Un estudio transversal en los países de la Unión Europea,. Revista de Contabilidad y Tributación, 1-35. Hochsztain, M. M. (2015). Factores de exito de un emprendimiento: Un estudio exploratorio con base en tecnicas de Data Mining. Espiritu Emprendedor, 31-40. INEGI. (2015). Encuesta Nacional sobre Productividad y Competitividad de las Micro, Pequeñas y Medianas Empresas (ENAPROCE). Obtenido de Instituto nacional de estadistica geografia e informtica : http://www.inegi.org.mx/est/contenidos/proyectos/encuestas/establecimientos/otras/enaproce/default_t.aspx


21 Ishikawa, K. (1990). ¿ Que es el control total de calidad? Barcelona: Norma. Jesús Perdomo, O. H. (2011). La Gestión de Recursos Humanos Enfocada en la Calidad Total y la Innovación. Vniversitas Economicas. Juran, J. M. (1990). Juran y el liderazgo para la calidad. Manual para ejecutivos. Madrid: Dias de Santos. Juran., J. M. (s.f.). Filosofia de la calidad. s.p. Kanji, G. (1990). Total Quality Management: the Second Industrial Revolution»,. Total Quality Management,, 3-11. Khalid, S. I. (2011). TQM Implementation in Textile Manufacturing Industry to Success: Review and Case Study. International Business Research. Lopes de Oliveira, F. M. (2013). El desempeño económico financiero y la responsabilidad social corporativa. Petrobrás versus Repsol. Revista Contaduría y Administración, 131-167. McGregor, D. M. (1960). The Human Side of Enterprice en Management Review . American Management Association. Oakland, J. (2000). Total Quality Management-Text with Cases,. Butterworth-Heinemann, Londres. Orlitzky, M. S. (2003). Corporate Social and Financial Performance: A Meta-Analysis. Organization Studies,, 403-441. Pérez, M. C. (2007). La gestión de la calidad total en el centro para la promoción del Comercio Exterior de Cuba. Ciencias De La Información, 38(3), 71-76. Samat, N. b. (2004). The relationship between total quality management (tqm) practices, service quality, market orientation, and organizational performance. Business Administration. Samson, D. y. (1999). «The Relationship between Total Quality Management Practices and Operational Performance», . Journal of Operations Management, 393-409. Smith, R. A. (2014). Quantifying Quality Management System Performance in order to improve business performance. . South African Journal of Industrial Engineering, 25(2), 75-95. Taylor, F. W. (1973). Principios de la Administración Científica. Buenos Aires: Editorial Ateneo. Yong Lam, S. L. (2014). A literature review and proposed framework: TQM, market orientation and performance of service organizations. . International Journal of Academic Research, 106-111.


INTERNATIONAL SCIENTIFIC AND VOCATIONAL JOURNAL (ISVOS JOURNAL)

Vol.1

Issue:1

Date: December 2017

Received:20.12.2017

Accepted:30.12.2017

Final version : 31.12.2017

International Scientific and Vocational Journal ISVOS Journal Short-Term Load Forecasting Model Using Flower Pollination Algorithm Volkan ATEŞa , Necaattin BARIŞÇIb a Electrical b Computer

and Electronics Engineering Kırıkkale University, TURKEY Engineering Faculty of Technology Gazi University, TURKEY

Abstract Electricity is natural but not a storable resource and has a vital role in modern life. Balancing between consumption and production of the electricity is highly important for power plants and production facilities. Researches show that electricity load consumption characteristic is highly related to exogenous factors such as weather condition, day type (weekdays, weekends and holidays etc.), seasonal effects, economic and politic changes (crisis, elections etc.). In this study, we propose a short-term load forecasting models using artificial intelligence based optimization technique. Proposed 5 different empirical models were optimized using flower pollination algorithm (FPA). Training and testing phase of the proposed models held with historical load and weather temperature dataset for the years between 2011-2014. Forecasting accuracy of the models was measured with Mean Absolute Percentage Error (MAPE) and monthly minimum approximately %1,79 for February 2013. Results showed that proposed load forecasting model is very competent for short-term load forecasting. Keywords: “Short-Term Load Forecasting; Nature-Inspired Optimization; Flower Pollination Algorithm; Artificial Intelligence”

1.

INTRODUCTION

Electric energy is an energy source produced by using various sources and also having a wide variety of consumption fields. Electric energy consumption has a large share in total energy consumption because it can easily be converted into basic energy types such as heat, light, motion. The magnitude of the industrial sector volume, which can be expressed as a measure of development, is directly proportional to the consumption of electric energy. Load consumption forecasting models are essential and vital for planning and production scheduling of the production facilities. These forecasting models and studies help to make important decisions on purchasing, generating electric power, load switching, and infrastructure development. The forecasting studies are also extremely important for suppliers and other participants in production, transmission, distribution and energy markets [1]. Load demand forecasting is very important for planning and production of power plants and production facilities. Load demand forecasts are generally divided into three main divisions: short-term, medium-term and long-term depending on the time periods [2]. Medium and long-term demand forecasts basically are related to what the electricity consumption is based on in total years, what might be in the future. The establishment of new power generation plants and the planning of large-scale energy investments are planned according to this demand. Short-term based approaches are based on optimum usage of existing resources and profiling the consumption characteristics. Basically, short-term methods are used for security and survival of the main system. When literature is examined, it is seen that, there are several types of models and methods tried out on load consumption and energy demand forecasting. These studies are divided into two main sections as classical statistical based approaches and artificial intelligence based approaches. These models and methods are classified as time series based (univariate) models which are modelled as a function of historical load data and expert systems which are modelled as systems of exogenous factors especially weather and social variables [3]. Traditional statistical models are based on relation on historical load changes. Some of these univariate studies are autoregressive models [4, 5], dynamic linear, non-linear models [6, 7], non-parametric regressive model [8], structural model [9], curve-fitting procedural model [10]. These models have lower forecasting errors for the routine periods of consumption. Literature shows that statistic based models are not sufficient for non-linear periods of consumption. Thus, researchers leaned to nature inspired and artificial intelligence based non-linear models. Some of these studies are fuzzy logic and artificial neural


23

networks based forecasting models [11-14]. S. Hassan et al. (2016) proposed fuzzy type-2 based load forecasting model. They also used extreme learning machine (ELM) to optimize and finding optimal fuzzy parameters [15]. D.K. Chatuverdi et al. (2015) proposed a model by using generalized neural networks (GNN) [16]. Song Li et al. (2015) used hybrid load forecasting model. They used ELM and modified artificial bee colony (MABC) to optimize input weights of ELM [17]. In Turkey, there are varying types of studies on short-term load forecasting. E. Yukseltan et al. (2017) proposed a linear model using climatic and econometric dataset. They trained and tested forecasting model for the period 2012-2014 and the weekly, daily horizon was estimated [18]. H.H. Cevik et al. (2015) proposed fuzzy and adaptive neuro-fuzzy models to estimate short-term load consumption of Turkey. They used historical weather information and seasonal changes in their study [19]. I Esener et al. (2006) proposed an artificial intelligence based model for short-term load forecasting. They used signal processing techniques and ANN with historical weather condition changes [20].

2.

MATERIAL AND METHODS

In this paper, we present a hybrid short-term load forecasting model by using structural mathematical models with nature inspired optimization technic. We used historical load consumption and weather temperature changes data set for the period 2011-2014. We used five different mathematical equation for modelling load consumption and flower pollination algorithm (FPA) was used for finding optimal parameters in equations. Forecasting accuracy of the mathematical models was measured using mean absolute percentage error (MAPE). 2.1.

Dataset Preparation

Electricity load consumption has linear and non-linear characteristic. Literature shows that there are several types of input variables used to estimate and define load consumption characteristic. These variables are varying to location, model type and other exogenous factors. While having long-term forecasting, globally scaled factors are used such as economic growth depth and population changing for years, in short-term load forecasting, most of the researchers used hourly changing data such as air temperature, raining period, insolation period etc. In this study, we used 4 different input variables. These variables are;    

Last Day Consumption (LDC) Last Week Consumption (LWC) Weekly Consumption Trend (LCAL) Weekly Temperature Trend (TEFF)

Last day consumption (LDC) – Last week consumption (LWC) Historical changes of the electricity load consumption are very significant indicator for future demand estimations. Electricity load usually follows a routine path. Actual load consumption values belonging to 2013 is seen in Figure 1. Last day consumption for the time Lt-24 and last week consumption Lw-7 may help to estimate future load demand.

Fig 1. Actual hourly electricity load consumption of 2013

Weekly Consumption Trend (LCAL)


24

Data obtained with the observations of the amount of consumption for a certain period, may give significant information about the future situation of electricity consumption. These observations are not sufficient information enough for a certain forecasting lonely but may be helpful for an accurate estimation indeed. In this study, we tried to understand consumption characteristic using least square trend analysis for 7 days consumption period and used these results as an input for the mathematical models. Load trend analysis was calculated with using equations 1, 2 and 3. Load Trend đ??ż = đ?‘Ž + đ?‘?đ?‘Ľđ?‘–

(1)

Slope đ?‘?=

∑đ?‘›đ?‘–=1 đ?‘Ľđ?‘– đ?‘Śđ?‘– − đ?‘›đ?‘ĽĚ… đ?‘ŚĚ… ∑đ?‘›đ?‘–=1 đ?‘Ľđ?‘– 2 − đ?‘›đ?‘ĽĚ… 2

(2)

Load Intercept đ?‘Ž = đ?‘ŚĚ… − đ?‘?đ?‘ĽĚ…

(3)

Sample load trend data using these equations is seen in Table 1. Table 1. 25.12.2011 - 31.12.2011 Load consumption data

Time Period (Distance) 1 2 3 4 5 6 7

Date 25.12.2011 26.12.2011 27.12.2011 28.12.2011 29.12.2011 30.12.2011 31.12.2011 1.1.2012

Load Consumption (MWh) 26175 24386 26412 26493 26345 26463 26083 26595

x2

x2y

1 4 9 16 25 36 49

26175 97544 237708 423888 658625 952668 1278067

Weekly Temperature Trend (LCAL) Literature research shows that, air temperature changes are highly related to hourly based load consumption characteristic. Thus, we used historical temperature data set collected from six different location in Turkey between 2011 and 2014 as an input for proposed forecasting models. The weighted average value of temperature collected from different cities of Turkey calculated with using coefficients. these coefficients were defined by using economic development values according to report published by Republic of Turkey Ministry of Science, Industry and Technology in 2013. The effect of each locations is seen in Table 2. Table 2. Coefficient values of locations

2.2.

17030 (City-1)

17130 (City-2)

17135 (City-3)

17220 (City-4)

17351 (City-5)

17603 (City-6)

0.02

0.15

0.02

0.10

0.04

0.67

Mathematical Models

In this study, five different linear and non-linear empirical models were developed for load consumption forecasting. Mathematical models were mostly used empirical models in literature. In this study, five different linear and non-linear empirical models were developed for load consumption forecasting. Mathematical equations were empirical models mostly used by researchers in the literature. These are linear, power, exponential, semi-quadratic and quadratic models. Linear model Lm =

w1Ldc + w2Lwc + w3Lcal + w4Teff + w5

(4)

Km =

đ?‘¤1 (đ??żđ?‘‘đ?‘? )đ?‘¤2 + đ?‘¤3 (đ??żđ?‘¤đ?‘? )đ?‘¤4 + đ?‘¤5 (đ??żđ?‘?đ?‘Žđ?‘™ )đ?‘¤6 + đ?‘¤7 (đ?‘‡đ?‘’đ?‘“đ?‘“ )

Power model �8

(5)


25

Exp. model Em =

đ?‘¤1 đ?‘’ đ?‘¤2đ??żđ?‘‘đ?‘? + đ?‘¤3 đ?‘’ đ?‘¤4đ??żđ?‘¤đ?‘? + đ?‘¤5 đ?‘’ đ?‘¤6đ??żđ?‘?đ?‘Žđ?‘™ + đ?‘¤7 đ?‘’ đ?‘¤8đ?‘‡đ?‘’đ?‘“đ?‘“ + đ?‘¤9

Qm =

đ?‘¤1 (đ??żđ?‘‘đ?‘? )2 + đ?‘¤2 (đ??żđ?‘¤đ?‘? )2 + đ?‘¤3 (đ??żđ?‘?đ?‘Žđ?‘™ )2 + đ?‘¤4 (đ?‘‡đ?‘’đ?‘“đ?‘“ ) +

(6)

Qua. model 2

đ?‘¤5 (đ??żđ?‘‘đ?‘? )(đ??żđ?‘¤đ?‘? ) + đ?‘¤6 (đ??żđ?‘‘đ?‘? )(đ??żđ?‘?đ?‘Žđ?‘™ ) + đ?‘¤7 (đ??żđ?‘‘đ?‘? )(đ?‘‡đ?‘’đ?‘“đ?‘“ ) +

(7)

đ?‘¤8 (đ??żđ?‘¤đ?‘? )(đ??żđ?‘?đ?‘Žđ?‘™ ) + đ?‘¤9 (đ??żđ?‘¤đ?‘? )(đ?‘‡đ?‘’đ?‘“đ?‘“ ) + đ?‘¤10 (đ??żđ?‘?đ?‘Žđ?‘™ )(đ?‘‡đ?‘’đ?‘“đ?‘“ ) + đ?‘¤11 Sem-Qua. model Sm =

đ?‘¤1 (đ??żđ?‘‘đ?‘? ) + đ?‘¤2 (đ??żđ?‘¤đ?‘? ) + đ?‘¤3 (đ??żđ?‘?đ?‘Žđ?‘™ ) + đ?‘¤4 (đ?‘‡đ?‘’đ?‘“đ?‘“ ) +đ?‘¤5 √(đ??żđ?‘‘đ?‘? )(đ??żđ?‘¤đ?‘? ) + đ?‘¤6 √(đ??żđ?‘‘đ?‘? )(đ??żđ?‘?đ?‘Žđ?‘™ ) +đ?‘¤7 √(đ??żđ?‘‘đ?‘? )(đ?‘‡đ?‘’đ?‘“đ?‘“ ) + đ?‘¤8 √(đ??żđ?‘¤đ?‘? )(đ??żđ?‘?đ?‘Žđ?‘™ )

(8)

+đ?‘¤9 √(đ??żđ?‘¤đ?‘? )(đ?‘‡đ?‘’đ?‘“đ?‘“ ) + đ?‘¤10 √(đ??żđ?‘?đ?‘Žđ?‘™ )(đ?‘‡đ?‘’đ?‘“đ?‘“ ) + đ?‘¤11 Artificial intelligence based flower pollination algorithm was used to improve forecasting accuracy of the empirical models. Accuracy of the empirical models was calculated by using mean absolute percentage error (MAPE) and mean square error (MSE) methods. MAPE đ?‘›

1 đ??´đ??żđ?‘Ą − đ??šđ??żđ?‘Ą ∑ (| | đ?‘Ľ100) đ?‘› đ??´đ??żđ?‘Ą

(9)

đ?‘Ą=1

MSE đ?‘›

1 ∑(đ??´đ??żđ?‘Ą − đ??šđ??żđ?‘Ą )2 đ?‘›

(10)

đ?‘Ą=1

2.3.

Flower Pollination Algorithm

Nature inspired optimization algorithms have become very popular in the last 20 years. it can be seen in the literature, these algorithms are frequently used in NP hard problems especially in the solution of complex problems in engineering and industrial field [21]. Flower pollination algorithm (FPA) is a nature inspired optimization algorithm developed by Xie-She Yang in 2012. The algorithm is inspired by pollination process of the flowery plants. Pollination process is having a vital role for the flowery plants. Pollination can take two major forms: abiotic and biotic. About 90% of flowering plants belong to the biotic pollinat ion group. In biotic pollination process, pollens are transferred by natural transporters such as insects, flies or other animals. The abiotic form of pollination is realized with natural effects such as wind and diffusion. Grass is one of the good example of abiotic pollination process. From the aspect of view of biological evolution, the main purpose of flower pollination is the survival of the fittest and the optimal reproduction. This may be expressed like an engineering optimization process of flowery plant species. Optimization process of the FPA is based on four rules. These rules are: 1. 2. 3. 4.

Biotic and cross-pollination can be considered processes of global pollination, and pollen-carrying pollinators move in a way that obeys LĂŠvy flights. For local pollination, abiotic pollination and self-pollination are used. Pollinators such as insects can develop flower constancy, which is equivalent to areproduction probability that is proportional to the similarity of two flowers involved. The interaction or switching of local pollination and global pollination can be controlled by a switch probability p ∈ [0, 1], slightly biased toward local pollination

These rules are converted into proper mathematical equations. In biotic process of the pollination, flower pollen gametes are transferred and carried by pollinator animals such as insects and used for global search in algorithm. Therefore rule 1 and rule 3 are represented mathematically as below equation 11. đ?‘Ľ đ?‘Ą+1 = đ?‘Ľđ?‘–đ?‘Ą + đ?›ž đ??ż(đ?œ†)(đ?‘”∗ − đ?‘Ľđ?‘–đ?‘Ą )

(11)


26 đ?‘Ľđ?‘–đ?‘Ą is the pollen i, solution vector đ?‘Ľđ?‘– at iteration t. đ?‘”∗ is the best current solution among all solutions. đ?›ž is the scaling factor and used for controlling the step size for each iteration. đ??ż(đ?œ†) is a LĂŠvy-flights based step size parameter and corresponds the strength of the pollination. Levy-flights can be used for mimic of global pollination which is realized by bio-natural agents mentioned before. The local pollination, both Rule 2 and Rule 3 is represented in equation 12. đ?‘Ľ đ?‘Ą+1 = đ?‘Ľđ?‘–đ?‘Ą + đ?œ–(đ?‘Ľđ?‘—đ?‘Ą − đ?‘Ľđ?‘˜đ?‘Ą )

(12)

đ?’™đ?’•đ?’‹ − đ?’™đ?’•đ?’Œ are pollens from different flowers of same species and đ??? is The working process of FPA may be represented with pseudo code as below.

drawn from uniform distribution [0-1].

1. 2. 3.

Initialize a population of n flowers/pollen gametes with random solutions Find the best solution đ?&#x2018;&#x201D;â&#x2C6;&#x2014; in the initial population. Define a switch probability p â&#x2C6;&#x2C6; [0, 1] while (t <MaxGeneration) for i = 1 : n (all n flowers in the population) if rand < p, Draw a (d-dimensional) step vector L from a Levy distribution Global pollination via đ?&#x2018;Ľ đ?&#x2018;Ą+1 = đ?&#x2018;Ľđ?&#x2018;&#x2013;đ?&#x2018;Ą + đ?&#x203A;ž đ??ż(đ?&#x153;&#x2020;)(đ?&#x2018;&#x201D;â&#x2C6;&#x2014; â&#x2C6;&#x2019; đ?&#x2018;Ľđ?&#x2018;&#x2013;đ?&#x2018;Ą ) else Draw from a uniform distribution in [0,1] Do local pollination via đ?&#x2019;&#x2122;đ?&#x2019;&#x2022;+đ?&#x;? = đ?&#x2019;&#x2122;đ?&#x2019;&#x2022;đ?&#x2019;&#x160; + đ???(đ?&#x2019;&#x2122;đ?&#x2019;&#x2022;đ?&#x2019;&#x2039; â&#x2C6;&#x2019; đ?&#x2019;&#x2122;đ?&#x2019;&#x2022;đ?&#x2019;&#x152; ) end if Evaluate new solutions. If new solutions are better, update them in the population end for Find the current best solution g end while 4. Output the best solution found â&#x2C6;&#x2014;

3.

RESULTS AND DISCUSSION

In this study, we trained and tested 5 different empirical models with historical load consumption and temperature data set. Flower pollination optimization algorithm was used to optimize mathematical models and having accurate load forecasting values. Parameters and pre-defined constraints of the optimization method FPA are given in Table 3. Table 3. Parameters and constraints for FPA. Population Limit (Np) Probability Density Beta Step Size Maximum Iteration Limit

30 0.80 3/2 0.01 2000

Training and testing of the proposed models was used for monthly horizons. Proposed models trained with data of 2012 and 2013 then tested with data of 2014. Training and testing results for the period 2013 and 2014 are seen in Table 4. Table 4. Training and testing mape results for FPA. February

February

April

April

March 2014

2013

2014

2013

2014

March 2013

Linear

1,871

1,816

2,015

2,146

3,436

2,932

Exponential

11,257

11,919

10,156

11,148

10,206

11,141

Power

2,302

2,576

2,938

2,865

3,573

2,949

Quadratic

9,29

10,33

9,241

10,042

9,634

9,924

Semi-Quadratic

1,794

2,068

1,796

1,953

2,535

2,279


27

As it seen in Table 3, semi-quadratic model generally gives the better forecasting values than the other empirical models. Obtained results of the proposed models for the 12-hour period of 1 February 2013 is given in Table 4 and comparative results for the period between 1-5 February 2013 is seen in Figure 2. Table 5. Training and testing mape results for FPA

Real (MWh) 27114

SemQuad. 27654 27814 26865 26918 26822

Error Linear 1,991

Error exp. 2,582

Error power 0,917

Error quad. 0,722

Error semi-quad 1,075

25480

26002 27816 25290 25375 25175

2,048

9,167

0,746

0,411

1,196

24374

24958 27821 24374 24486 24292

2,395

14,142

0,000

0,459

0,335

23783

24416 27815 23632 23981 23637

2,661

16,952

0,633

0,834

0,613

23574

24327 27810 23463 23884 23579

3,195

17,968

0,473

1,315

0,023

24040

24758 27809 23860 24218 23971

2,986

15,679

0,751

0,742

0,285

24376

25434 27813 24538 24806 24631

4,341

14,101

0,666

1,764

1,046

25122

26630 27823 25734 25874 25959

6,001

10,750

2,437

2,993

3,330

29277

30599 27908 29921 30286 30472

4,514

4,676

2,199

3,447

4,082

32054

33055 27970 32597 33218 33000

3,124

12,742

1,695

3,633

2,951

33032

33793 28047 33552 34104 33682

2,303

15,091

1,574

3,245

1,967

33805

34152 28037 33975 34437 33925

1,028

17,064

0,502

1,869

0,354

Linear

Exp.

Power Quad.

Fig 2. Comparative results of the proposed empirical models for the period between 1-5 February 2013

4.

CONCLUSION

Electricity load consumption estimation is having a vital role for official and non-official environments which are responsible for generation and distribution of the resources. From the aspect the view of literature, short-term load forecasting is becoming very popular research area recent years. There are several types of studies and varying of models have been developed for aiming accurate load forecasting. These models and techniques are basically divided into two main categories: classical statistical based models and artificial intelligence or nature inspired models. In this study, we proposed a hybrid forecasting model which composed classical structural based empirical models and natureinspired optimization techniques. We used flower pollination algorithm for optimizing empirical models. Proposed forecasting models were trained and tested by using historical load consumption and weather temperature data. Obtained results show that, proposed empirical models give promising results on hourly based short-term load forecasting. Proposed semi-quadratic model is


28

having more accurate forecasting values than other models. Best result was obtained for February 2013 with %1,79 forecasting accuracy.

5.

ACKNOWLEDGMENT

This research was supported by Scientific Research Projects Coordination Unit (B.A.P.K.B.) of Kırıkkale University [Project No: 2016/099, 2017].

6.

REFERENCES

[1] Feinberg E.A. and Genethliou D., “Chapter 12 Load forecasting”, Applied Mathematics for Power Systems, pp.269-282. http://www.ams.sunysb.edu/~feinberg/public/lf.pdf [2] Fan S. and Hyndman R.J., “Short-Term Load Forecasting Based on a Semi-Parametric Additive Model,” IEEE Trans. Power Systems, vol.27, no.1, pp.134-141, 2012 [3] Hippert H.S., Pedreira C.E. and Souza R.C. “Neural Networks for Short-Term Load Forecasting: A Review and Evaluation”, IEEE Trans. Power Systems, vol.16, no.1 pp. 44-55, 2001 [4] Mbamalu G.A.N. and El-Hawary M.E., “Load forecasting via suboptimal seasonal autoregressive models and iteratively reweighted least squares estimation,” IEEE Trans. Power Systems, vol.8, no.1, pp. 343–348, 1993. [5] Yang H.T. and Huang C.M., “A new short-term load forecasting approach using self-organizing fuzzy ARMAX models,” IEEE Trans. Power Systems, vol.13, no.1, pp. 217–225, 1998. [6] Douglas A.P., Breipohl A.M., Lee F.N. and Adapa R.,“The impact of temperature forecast uncertainty on bayesian load forecasting,” IEEE Trans. Power Systems, vol.13, no.4, pp. 1507–1513, 1998. [7] Sadownik R. and E.P. Barbosa, “Short-term forecasting of industrial electricity consumption in Brazil,” J. Forecast., vol.18, pp. 215–224, 1999. [8] Charytoniuk W., Chen M.S. and Van Olinda P., “Nonparametric regression based short-term load forecasting,” IEEE Trans. Power Systems, vol.13, no.3, pp. 725–730, 1998. [9] Harvey A. and Koopman S.J., “Forecasting hourly electricity demand using time-varying splines,” J. American Stat. Assoc., vol.88, no.424, pp. 1228–1236, 1993. [10] Taylor J.W. and Majithia S., “Using combined forecasts with changing weights for electricity demand profiling”, J. Oper. Res. Soc., vol.51, no.1, pp. 72–82, 2000. [11] Mamlook R., Badran O. and Abdulhadi E., “A fuzzy inference model for short-term load forecasting”, Energy Policy vol. 37, no.4, pp.1239–1248, 2009 [12] Pandian S.C., Duraiswamy K, Rajan C.C.R and Kanagaraj N, “Fuzzy approach for short term load forecasting”, Electr Power Syst Res vol.76 no.(6–7), pp.541–548, 2006 [13] Aggarwal S, Kumar M, Saini LM and Kumar A, “Short-term load forecasting in deregulated electricity markets using fuzzy approach”, J Eng. Technol, vol.1, no.1, pp.24–31, 2011 [14] Ying L.C., Pan M.C., “Using adaptive network based fuzzy inference system to forecast regional electricity loads”, Energy Convers Manag., vol.49, no.2, pp.205–211, 2008 [15] Hassan S., Khosravi A., Jaafar J. and Khanesar M. A., “A systematic design of interval type-2 fuzzy logic system using extreme learning machine for electricity load demand forecasting”, Electrical Power and Energy Systems, vol.82, pp.1–10, 2016 [16] Chaturvedi D.K., S.O. M. “Short-Term Load Forecasting Using Fuzzy Logic and Wavelet Transform Integrated Generalized Neural Network”, Electrical Power and Energy Systems, vol.67, pp.230-237, 2015


29 [17] S. Li, P. L. G. “Short-term load forecasting by wavelet transform and evolutionary extreme learning machine”, Electric Power System Research, vol.122, pp. 96-103, 2015 [18] Yukseltan E., Yucekaya A. and Bilge A.H., “Forecasting electricity demand for Turkey: Modeling periodic variations and demand segregation”, Applied Energy, vol.193, pp. 287–296, 2017 [19] Çevik H.H. and Çunkaş M., “Short-term load forecasting using fuzzy logic and ANFIS”, Neural Computing and Applications, vol. 26, no.6, pp.1355–1367, 2015 [20] Esener I., Yuksel T. and Kurban M., “Short-term load forecasting without meteorological data using AI-based structures”, Turkish Journal of Electrical Engineering & Computer Sciences, vol.23, pp.370-380, 2015 [21] Yang X.-S., Nature-Inspired Optimization Algorithms, Londra: Elsevier, 2014.


INTERNATIONAL SCIENTIFIC AND VOCATIONAL JOURNAL (ISVOS JOURNAL)

Vol.1

Issue:1

Date: December 2017

Received:20.12.2017

Accepted:30.12.2017

Final version : 31.12.2017

International Scientific and Vocational Journal ISVOS Journal INTERLOCK OPTIMIZATION OF AN ACCELERATOR USING GENETIC ALGORITHM İbrahim Burak Koç a, Anas Al Janadi b, Volkan Ateş b1 aInstute

of Accelerator Technologies, Ankara University Golbasi Campus, Golbasi Ankara TURKEY of Electrical and Electronics Engineering, Kirikkale University, Kirikkale TURKEY

bDepartment

Abstract Accelerators are systems where high-tech experiments are conducted today and contain high-tech constructions. Construction and operation of accelerators require multidisciplinary studies. Each accelerator structure has its own characteristics as well as similar features of accelerator structures. Control systems come to the forefront as one of the most important structures that make up accelerators. Since control systems have critical importance for accelerators, in such systems when a problem occurs, there is a danger of environmental and human safety as well as machine system. For that reason interlock systems are being developed in different structures. In the literature, FPGAs and PLCs in such interlock systems have been shown to be suitable for use in accelerators [1,7]. In this work, we describe an interlock system that evaluates the operation and protection modes of devices used in an electron accelerator. In order to ensure that this system can operate at minimum cost and maximum safety, the defined system is divided into 3 subsystems. The error messages from the control devices in the accelerator control systems are the input to the interlock system. The purpose of the interlock system that evaluates error messages is to ensure that the accelerator closes safely. The purpose of this study is to specify which of the 3 interlock subsystems which are defined for minimum cost and maximum security should be connected to the fault outputs from the control devices. As an evaluation criterion, 6 features are defined for the control devices and each control device is weighted according to the importance of the task. In the solution of the problem, genetic algorithms were used for assigning 74 controller outputs to 3 interlock subsystems. Thanks to the Genetic Algorithm used in the study, 94.3% success rate was obtained in terms of cost and safe system. Keywords: “Optimization, Genetic Algorithm, Accelerator”

1.

INTRODUCTION

In accelerator systems, high-tech devices are used for the control of high-energy electrons. Some of these devices are control devices for obtaining laser at the output of the accelerator, and some of them are intended to ensure that the system does not pose a risk to parameters that are important for human and environmental health as well as for healthy operation. Such advanced technology and control of high risk systems should be assessed both accurately and efficiently in terms of machine, human and environmental health as well as cost. Control of electron electron beams in accelerators depends on many physical parameters. Control of these parameters is done through electronic systems or interfaces. In a structure where complex control systems and electronic systems exist, in the event that the control of the electron high energy beams can not be achieved, there are damages to the system, the environment and human health. This means that faults or failures in the system should be detected very quickly and should be avoided without harming the system, environment or human health. Each electronic device used in the control of the electron packs in the accelerators has a predetermined operating range. These devices generate error signals when they start working outside of the specified range. This error signal enters a central system, which activates a shutdown procedure in such a way that it does not harm human and machine health. This central system is called the interlock system. The error output of each device is called interlock output. That is, the devices enter the interlock outputs as inputs to the interlock system and thus report the error to the interlock system. The interlock system activates different 1

Corresponding author. Tel.: +0-000-000-0000 ; fax: +0-000-000-0000 . E-mail address: volkanfire@hotmail.com


31

shutdown procedures by evaluating the error messages received at the input. The shutdown procedures vary according to the severity of the error signals entering the interlock system. The priority of the error signals, which are input to the interlock system, is important in proportion to the damage it will give. It may be permissible to react more slowly to some error signals when it is necessary to react more quickly to some error signals. There is a need for a fast interlock system where rapid reaction is required and the system must be shut down quickly. Furthermore, the interlock system should have a very low probability of failure in continuous operation. However, connecting all the interlock outputs in the system to the fast and extreme reliable interlock subsystem increases the cost too much. For this it is the case that some of the interlock outputs have to be connected to systems that run slower and have sufficient reliability and therefore are less costly. In this context, it is envisaged that the interlock system to be used in the accelerator will come to 3 subsystems. These are FPGA, PLC and software interlock systems.

Figure 1. General structure of interlock system

The FPGA interlock subsystem is suitable for operations requiring very high speed shutdown procedures. The shutdown delay is around 1ms and the failure rate is extremely low, so it is highly reliable. However, it is the highest cost system. The PLC interlock subsystem can be selected for operations requiring a high-speed shutdown procedure. The shutdown delay is around 50mS, which depends on the size of the application. Although the failure rate is low, it is higher when compared to FPGAs. Cost is high but is lower than FPGAs. It offers the optimum reliability and cost ratio within Interlock subsystems. Finally, the Software interlock subsystem has the lowest cost but the highest probability of error. The shutdown delay is around 500ms. Six evaluation criteria for 74 different error sources were introduced in order to predict which interlock subsystem the error outputs of the devices would enter. Occurrence probability: Likelihood of occurrence of the error. Risk: The cost of the fault when error occurs. Source speed: The speed of the source as the control system to be sent to interlock. In the other words the time elapsed between the occurrence of the error and the error output of the relevant control system device. Human health: Impacts on the human health if the accident occurs. In some cases, it is necessary for the interlock devices' error outputs to be configurable as an input to the interlock system. For example, some error sources can be turned off during the development phase, or the protection intervals can be extended, providing flexibility during development. Such an authorization is only for high-level expert users. Such authorizations are; Editable-treshold: The limit values for the error source can be set.


32

Editable-enable/disable: Completely ignorability of the error coming from the error source. There are 74 control error outputs that can cause an error in the accelerator application. For each control error output, the data set for the interlock system was created by normalizing and weighting the above 6 properties in the interval of 0-1. The data set was evaluated by expert personnel and the knowledge base was determined. As an example of the systems in the accelerator, the simple structure of the Beam loss monitor system (BLM) is given in Figure 2. In Figure 2a, Electron beams passing through the cylindrical vacuum beam line, induce current on a wire detector located outside the beam line. According to the induced current value, it is learned how close the electron beams approach the wall of the beam line. In Figure 2a. and 2b, the minimum distance allowed to approach the wall of the beam line is shown by the broken line. As shown in Figure 2b, as the electron packages approach the allowable distance, the induced current on the wire increases. When the current value induced on the wire outside the beam line exceeds the permissible current value, the BLM device outputs the interlock output (error output). Here the BLM system is an interlock source. The Blm device is an interlock device. The interlock device generates the interlock signal (error signal) according to the previously determined parameters in the interlock system. The BLM system is considered as interlock source number 6 in Table 1.

Figure 2. Basic layout of the Beam Loss Monitoring (BLM) system as example of an interlock source

There are many devices and parameters that allow control and adjustment of electron packs. If one of these devices goes out of control or if the parameters are selected incorrectly, the error probability increases. For this reason, the “Occurrence Probability” is estimated to be 0,6. As a result of the high-energy electron packs striking the beam line, the vacuum line will be drilled and the vacuum will be broken. Vacuum distortion will affect many systems operating under vacuum in the electron field. Therefore, the risk factor is weighted as 0,7. BLM system is a system that needs to react quickly as the “Risk” factor is high. The response time for system security should be around 50ms. When normalized between 0 and 1, the “Source Speed” value is 0.02 as seen from Table 1. The maximum approach distance to the wall of the beamline must be adjustable or completely switchable in order to provide flexibility during the development phase if the expert staff uses the machine. This is shown in Table 1 by weighting the editable options (enable and threshold) to 1.


33

Although the BLM interlock system is risky for the machine, the human health effect is weighted as 0,3 so that there will be no human in the environment during the operation. This system needs to be connected to the FPGA from interlock subsystems because the risk ratio is high and it needs to react quickly. The other 73 interlock sources are similarly weighted. Some of these values are given in Table 1. Table 1. Some of the interlock sources and their weighted features

No

Interlock Source

Occurrence Probability

Risk

Source speed (1/ms)

Editable Enable

Treshold

Human health

Knowlage Base

1

HV Spark

0,8

0,2

0,002

0

0

0,05

C

2

Beam Control volt down

0,2

0,6

0,02

0

0

0,1

B

3

Voltage divider down

0,1

0,4

0,002

0

0

0,1

C

4

HVPS down

0,2

0,3

0,002

0

0

0,3

C

5

Viewscreen wrong position

0,4

0,6

0,02

0

0

0,3

A

6

BLM

0,6

0,7

0,02

1

1

0,3

A

7

Aperture temparature too High

0,8

0,3

0,02

0

0

0,2

B

8

BPM

0,7

0,7

0,02

0

1

0,4

A

..

……

..

..

..

..

..

..

..

..

……

..

..

..

..

..

..

..

74

Beamline shutter valve open

0,4

0,2

0,1

0

0

0,9

A

What is expected from the optimization algorithm is to determine which interlock subsystem will fit into 74 interlock sources with 6 different characteristics. In doing that, it should choose the minimum cost possible and the required speed and maximum security level.

2.

MATERIAL AND METHOD

Solving such a data classification problem, different methods could be used. These are Feature Selection Methods, Probabilistic Methods, DecisionTrees, Rule-Based Methods, Instance-Based Learning, SVM Classifiers, Neural Networks, Genetic Algorithms and etc [2]. Genetic Algorithm is one of the most populer hueristic algorithms for classification problems. Several other people working in the 1950s and the 1960s developed evolution−inspired algorithms for optimization and machine learning [3]. Genetic algorithms (GAs) were invented by John Holland in the 1960s and were developed by Holland and his students and colleagues at the University of Michigan in the 1960s and the 1970s [3].


34

The genetic algorithm is an iterative and stochastic process that takes the individuals (populations) in a set into account. Each individual represents a potential solution for the given problem. At the beginning the population is randomly selected. Each individual in the population is given a value as a measure of the goodness by the help of of the fitness function This value is the information the algorithm uses to find the best solution for the given probing [6]. Selection, crossover and mutation operators are used in the genetic algorithm. Replacement is done with the generation of new individuals. This algorithm proceeds with the help of simple operators, creating better generations each time. To determine the better individuals in the population, the fitness value of each individual is calculated. The obtained fitness values are used to rank the individuals in the problem being solved. These values are then combined to form the fitness function. The population of the individuals obtained for the best fitness function is assigned as the solution of the problem [6]. Fitness Function: The fitness function is used to determine how close the solution to the given problem is. The fitness function in the genetic algorithm varies according to the problem given. Extensive literature is available on the characteristics of the fitness function to expedite the search using GAs [8, 9, 10]. Often, in implementing a GA, some additional measures are taken into consideration. Among various such mechanisms, linear scaling, sigma truncation and power scaling are the most popular [8, 10]. Detailed information about the fitness function used in this study will be given in the following sections. The operators used in the genetic algorithm can be summarized as follows: Selection (Reproduction): Reproduction is a process in which individuals are copied according to their fitness function. It means that individuals with the best fit will have a higher contribution to the next generation. When the selection process is performed, firstly the individuals above the median fitness value are collected in the matching pool. Individuals gathered in the pool can be selected using various methods. These methods can be Roulette wheel selection, random selection, rank selection, tournament selection, boltzmann selection, stochastic universal sampling [6]. Crossover Probability: In order to increase the potential of the existing gene pool, the crossover operator is used to generate better-quality sequences than the previous generation. Random values are selected from the new population obtained as a result of reproduction. These values are subject to crossover processing. By using the crossover operator, individuals with higher compliance values can be obtained from the parents. Mutation Probability: The crossover operator searches for the current gene potential. However, if the population does not contain the information necessary to solve the given problem, a good result can not be obtained. For this reason, an operator with the ability to generate new sequences from the existing gene pool is required. Mutation operator increases diversity by producing new series. Mutation prevents the problem from converging to any local optimum. Elitism Probability: Elitism is used to protect good individuals in the population and to improve the performance of the genetic algorithm. The best individuals in the current generation are selected at a certain rate and transferred to the next generation without being exposed to other genetic operators. In this way, the working speed of the genetic algorithm is increased and the possibility of disappearance of the individuals which are not wanted to be lost during the mutation is eliminated. On the other hand using the Elitism operator high-rate causes the genetic algorithm to converge to the local maximum instead of the global maximum.

3.

IMPLEMENTATION

In this study we used an application of genetic algorithm which K.C. Tan and friends proposed in data mining by evolving a set of comprehensive decision with high classification accuracy [4]. They called it Genetic Programming classifier(GPc) and we will say it TAN Algorithm in this study. Tan and his friends have extended the idea of boolinezed attributes [11] in their study and have used the Michigan approach to encode the rules, where each individual encodes only one rule. This approach provides a simpler structure, easier development and better comprehension [4]. Along with this, they introduced a new mapping technique to make a more effective fitness


35

evaluation. They also benefited from the covering algorithm, which preserves the rules that are good for developing classification rules and ignores bad rules. The flowchart of the TAN algorithm is shown in Figure 3. The initial population starts with ramped half and half method [5]. In order to protect the individuals with the best evaluation value in the population and transfer them to the next generation, the elitism which is determined by the user has been applied. Genetic operators have been applied to individuals who are not elite individuals. Tournament selection method and tournament size 2 were used as the selection method. Small selection of the tournament size provides better convergence by reducing the pressure on the tournament selection [12]. Individuals are subjected to crossover and mutation and then combined with elite individuals from the previous generation to obtain a new population. In TAN's work, a concept map was introduced to improve the rules. In each GP cycle, a concept table is created over the training set and stored for future reference. This reduces the need to rescan the training set and makes fitness evaluation more efficient [4]. Individuals in the population enter the covering algorithm. The aim of this algorithm is to improve multi-rule structure by supporting diversity. Multiple rules support the same situations in the dataset, allowing for earlier convergence. Often, only a few of these rules are useful and many are unnecessary. The token competation technique was proposed by Wong and Leung (2000) and is used to eliminate unnecessary rules in the rule set. The pseudo code for token competation of TAN Algorithm is given in Figure 4.

Figure 3. Flow diagram of the TAN algorithm used


36

Figure 4. Pseudo code for Token Competation of TAN Algorithm

Since the fitness evaluation in the genetic algorithm varies according to the problem given, it is beneficial to open a paranthesis for fitness evaluation at this point. Fitness function: The fitness function is used to improve the quality of each rule or individual. The fitness function can be expressed as: đ?&#x2018;Ąđ?&#x2018;?

đ?&#x2018;Ąđ?&#x2018;&#x203A;

fitness = đ?&#x2018;Ąđ?&#x2018;?+w1đ?&#x2018;&#x201C;đ?&#x2018;&#x203A; đ?&#x2018;Ľ đ?&#x2018;Ąđ?&#x2018;&#x203A;+w2đ?&#x2018;&#x201C;đ?&#x2018;?

(1)

True positive (tp) and true negative (tn) are correct classifications while, false positive (fp) and false negative (fn) are incorrect classifications [4]. To explain the terms of the classification in this work: Table 2. Confusion Matrix

Predicted Class

Actual Class

a

b

tp

fn

a

fp

tn

b

The w1 and w2 weighting values are used to improve the prediction accuracy in the classifier tests in the Tan algorithm [4]. These values are predefined by the user. As can be seen from Eq. (1), the fitness function has the highest value at low values of w1 and high values of w2. However, excessive reduction or increase of these values may result in various over-fittings such as an increase in the number of rules [4]. Therefore 0,2<w1<1 and 1<w2<20 must be selected [4]. Tan and his colleagues used w1=w2=1 value in their work [4]. The first part of the fitness function is known as sensitivity and the second part is known as specificity [4]. In this study, the TAN algorithm is achieved by running the program in the WAKA (Waikato Environment for Knowledge Analysis) [13] program. The algorithm used is compared with different classification algorithms and the results are evaluated. It is seen that the algorithm used provides the optimum result in terms of reliability and cost within other algorithms. Firstly, the interlock data set given in Table 1 has been adapted to the WEKA program. Later, the annex of the TAN algorithm was installed in the WEKA program. In the TAN algorithm used, the crossover probability value was chosen as 0.9, the mutation value as 0.9 and the elitism value as 0.1. In addition, interlock data set is used as training data. The outputs of the executed algorithm are given in Figure 5. and Table 3. The applied rules are shown in Figure 5 and detailed accuracy analysis according to the classes is given in Table 3.


37

Figure 5. Output screen for TAN Algorithm to be applied to rule base in WEKA program

As shown in Figure 5, 6 rules are used for 6 properties. Another thing to note is that the algorithm has arrived in 3.59 seconds as a result. === Detailed Accuracy By Class === TP Rate

FP Rate

Precision

Recall

F-Measure

1,000

0,056

0,870

1,000

0,930

0,767

0,000

1,000

0,767

1,000

0,080

0,857

1,000

MCC

ROC Area

PRC Area

0,906

0,972

0,870

0,868

0,813

0,883

0,861

0,923

0,888

0,960

0,857

Class a b c Weighted Avg.

0,905

0,041

0,918

0,905

0,903

0,863

0,932

0,862

Table 3. The values of TAN Algorithm in detail analysis in WEKA program

Comparison with different algorithms: Comparisons were made with 5 different algorithms to determine how well the TAN algorithm used was appropriate for classification in this study. The results of the comparison are given in Table 4.


38 Table 4. Comparison table for different algorithms

Rules TAN 67 Correctly Classified (90.54%) Instances Kappa statistic Mean absolute error Root mean squared error Relative absolute error Root relative squared error Total Number of Instances

Meta MultiClass Classifier 67 (90.540%)

Naive Bayes 46 (62.162%)

Meta Bagging

Meta Attribute Bayes net search Selected Classifier local K2 67 64 (90.540%) (86.486%)

66 (89.189%)

0.8582

0.4039

0.8559

0.8388

0.8587

0.7991

0.0631

0.283

0.136

0.1088

0.0989

0.1631

0.2511

0.4208

0.2267

0.2325

0.2224

0.2702

14.381 %

64.533 %

31.011 %

24.816 %

22.559 %

37.200 %

53.644 %

89.890 %

48.430 %

49.658 %

47.509 %

57.727 %

74

74

74

74

74

74

When the comparision table (Table 4.) is examined, it is seen that 3 algorithms are in the foreground in terms of percent accuracy. One of these is the TAN algorithm used in this study, while the others are the MetaMultiClass and MetaAttrubuteSelected classification algorithms. At this point, confusion matrices need to be examined more closely. Confusio n matrices of algorithms used for comparison is given in Table 5. Table 5. Confusion matrix comparision table

a 2 0 3 0

Meta MultiClass Classifier

Naive Bayes

Rules TAN

Meta Bagging

Meta Attribute Selected Classifier

b

c

a

b

c

a

b

c

a

b

c

a

b

c

0 2

0

7

7 2

6

18

2 2

0

20

0 2

0

20

0 2

0

4 2

0

5 1

2

1

6

2

6

22

0

24

0

3 0

4

5 1

0

0

4

0

7 2

2 0

3 0

Bayes net search local K2 a 2 0

b

c

0

0

a=a

4 2

b=b

1

6

20

24

0

0

4

c=c

Twenty-four of the 74 interlock sources used for training were identified as “a” (FPGA subsystem), 30 as “b” (PLC subsystem), and 24 as “c” (software subsystem). What is expected from the algorithms is to capture these values in the result of classification. In general terms, transitions from “a” to “c” (Classified as Software while it needs to be classified as FPGA), transitions from “a” to “b” (Classified as PLC when it needs to be classified as FPGA) and transitions from “b” to “c” (Classified as software when it needs to be classified as PLC ) is regarded as unacceptable and faulty classification because it reduces safety. On the other hand, in the opposite case, the transition from “c” to “a” (classified as FPGA when it needs to be classified as software), “c” to “b” transition (classified as PLC when it needs to be classified as software) and transition to “b” to “a” (classified as FPGA when it needs to be classified as PLC) acts to increase security. However, acceptance of such transitions is conditional on the fact that the number of transitions is not too great, as there is an increase in security as well as an increase in cost. When we examine the confusion matrix of the MetaMultiClassClassifier algorithm, there are two instances where “b” should be “a” (Connected to the PLC while the FPGA needs to be connected). This is unacceptable because it reduces safety. There are two states connected to “a” when it is necessary to connect to “b”. Such a situation would mean an increase in safety and an increase in cost at the same time, but it is acceptable if such a transition is low by the means of number. There are two cases connected to “b” when it is needs to be connected to “c” in a similar way. This also means increased safety and cost, but it is acceptable because it is a small amount. The other case is a transition from “b” to “c”. This transition also reduces safety and is unacceptable because it is connected to the Software when it needs to be connected to the PLC. As a result, this algorithm was not found to be successful due to the fact that it contains two unacceptable situations in terms of security and additionally two conditions that would increase the cost. When we examine the confusion matrix of the MetaAttributeSelectedClassifier algorithm, there is 1 case where c is needed to be b. This is unacceptable and is considered an error. On the other hand, there are 6 situations in which a must be. Although this is an acceptable value since it is in the direction of increasing safety, it is an imbalance that, as mentioned earlier, it will increase


39

the cost excessively if it goes above a certain amount (The fact that 6 of the 7 faulty classifications are in the direction of increasing the cost causes to depart from the optimum solution). When we look at the confusion matrix of the TAN algorithm, there are 4 cases where c should be b. This is unacceptable and is taken as the error value. On the other hand, there are 3 cases which are predicted as “a” needs to be “b”. These three cases can be accepted as they are in the direction of increasing safety. When examined in terms of cost, it can be said that the best solution is better approximated by looking at the ratio of these 3 values in the 7 values compared to the MetaAttributeSelectedClassifier algorithm In this case, it is seen that the TAN algorithm is better than other algorithms in terms of security and cost. Although classification accuracy is given as 90.54% in Table 4, classification accuracy is 94.3% when 3 cases which increase safety are ignored. The classification graph of the used TAN algorithm is given in Figure 6. Correct classifications and misclassifications are seen together in the graph.

Figure 6. Classification erros of TAN Algorithm

Each of the seven misclassified cases can be viewed from the Weka status information screen. The information of three instances which are classified as 'a' needed to be classified as 'b' are shown in Figure 7. In Figure 8. information of 4 instances which are classified as 'c' needed to be classified as 'b' is shown.

Figure 7. Classification information for instances that can not be classified correctly (instances classified as 'a' when it should be classified as 'b')


40

Figure 8. Classification information for instances that can not be classified correctly (instances classified as 'c' when it should be classified as 'b')

4.

CONCLUSION

In this work, an interlock system with three subsystems to evaluate the operation and protection modes of the devices used in the accelerator is defined. 74 systems with interlock error output used in the accelerator were identified and 6 different features were determined for these 74 systems. For each system, these 6 features are weighted 0-1 according to importance, and interlock dataset is created. Through the generated interlock dataset, a genetic algorithm is used to determine which interlock source (system) will enter into which interlock subsystem. As the genetic algorithm, an algorithm developed by TAN and his friends and defined as TAN algorithm is used in this study. This algorithm has run in WEKA program and classification has been made. The interlock dataset was used for the training of the algorithm and as a result, 94.3% success rate was obtained in terms of minimum cost and maximum security. In order to measure the success of the work done, 5 different algorithms were compared. It has been seen that the TAN algorithm used in this study is more successful in evaluating these algorithms based on confusion matrices. The work done in this paper can be a guide for the cost and security planning of interlock systems for newly installed accelerators. Considering the different needs of each accelerator, the systems to be used can be varied and weighted in different ways according to their safety need. The number of systems can be increased also. Optimized results can be obtained by conducting detailed analysis on the confusion matrix by running the algorithm. An already installed accelerator can be configured with a minimum cost interlock security system by increasing the number of systems and weighting each system as needed. It is hoped that this study will also be a resource for those who will work towards the development of new and more succesive algorithms for the given problem.


41

5.

REFERENCES

[1] M. Kago, T. Matsushita, N. Nariyama, C. Saji, R. Tanaka, A. Yamashita, Y. Asano, T. Fukui, T. Itoga, System Design of Accelerator Safety Interlock for the XFEL/SPRING-8, Proceedings of IPAC’10, Kyoto, Japan [2] Charu C. Aggarwal, Data Mining and Knowledge Discovery Series, CRC Press, 2015 [3] Melanie Mitchell, An Introduction to Genetic Algorithms, MIT Press, 1999 [4] Tan KC, Tay A, Lee TH, Heng CM. Mining multiple comprehensible classification rules using genetic programming. In: IEEE Congress on Evolutionary Computation, Honolulu, HI, 2002. p. 1302–7. [5] J. R. Koza. Genetic Programming: on the Programming of Computers by Means of Natural Selection. Cambridge, MA: MIT Press, 1992 [6] Sivanandam S. N. Deepa S. N. Introduction to Genetic Algorithms Springer-Verlag, Berlin, Heidelberg, 2008 [7] R. Schmidt Machine Protection and Interlock Systems for Circular Machines—Example for LHC CERN, Geneva, Switzerland arXiv:1608.03087v1 [physics.acc-ph] 10 Aug 2016 [8] Sankar Kumar Pal Classification and learning using genetic algorithms_ applications in bioinformatics and web intelligence-Springer (2007) [9] D. E. Goldberg. Genetic Algorithms in Search, Optimization and Machine Learning. Addison-Wesley, New York, 1989. [10] Z. Michalewicz. Genetic Algorithms + Data Structures = Evolution Programs. Springer, Berlin Heidelberg New York, 1992. [11] C. C Bojarczuk, H.S. Lopes, and A.A. Freitas. Genetic Programming for Knowledge Discovery in Chest-Pain Diagnosis. IEEE Engineering in Medicine and Biology, July/August, pp. 38-44,2000 [12] Miller, Brad; Goldberg, David (1995). "Genetic Algorithms, Tournament Selection, and the Effects of Noise". Complex Systems. 9: 193–212 [13] I. H. Witten and E. Frank. Data Mining: Practical Machine Learning Tools and Techniques with Java Implementations. CA: Morgan Kaufmann Publishers, 1999.


INTERNATIONAL SCIENTIFIC AND VOCATIONAL JOURNAL (ISVOS JOURNAL)

Vol.1

Issue:1

Date: December 2017

Received:20.12.2017

Accepted:28.12.2017

Final version : 31.12.2017

International Scientific and Vocational Journal ISVOS Journal The Sunshine Duration Error Rates were calculated with Gradient-Descent Algorithm and Genetic Algorithm for Use of Solar Energy Yalçın KAPLAN a,1 aİstanbul

Kemerburgaz Üniversitesi, Fen Bilimleri Enstitüsü, 34217 İstanbul

Abstract In this study, which is a source of renewable energy required to take advantage of solar energy to the maximum duration of sunshine was estimated. In the study, values were used of the city of Amasya. Artificial neural networks (ANN) backpropagation gradient-descent(GD) learn algorithm and genetic algorithm(GA) were used. Three hidden layer network model was designed with two inputs for ANN and GA. Between 2000 and 2010 values were used as input data monthly sunshine duration and humidity values. Output data was obtained monthly sunshine duration of 2010. The values obtained were compared with the actual values and the root mean square error (RMSE) was calculated. Result of the study, GA was used to calculate the values that are needed for solar energy. Keywords: “Artificial neural network, solar energy, Sunshine”

Giriş

1.

Nüfusun artışı ile birlikte enerji ihtiyacının artması toplumların en büyük sorunu haline gelmiştir. Ülkelerin ve şehirlerin hızla büyümesi artan nüfusu beraberinde getirmiş, sanayileşmenin de artmasıyla ihtiyaç duyulan enerji hızla artmıştır. İnsanların yaşam standartlarının artması ihtiyaç duyulan sanayileşme olgusunu da arttırmış ve artan sanayi ile birlikte çevre sorunlarını beraberinde getirmiştir, insanlar doğal düzeni korumak ve yeni nesillere temiz bir yaşam bırakmak gayreti içerisindedir[1]. Ülkemiz 17 Şubat 2009’da Kyoto Protokolü imzalaması ile protokolde geçen emisyon değerlerine ulaşmak için etmenleri tekrar değerlendirmek ve protokolde geçen uyum şartlarını sağlamak zorundadır[2]. Bütün ülkeler gelecekte ortaya çıkabilecek enerji sorunları için önlemlerini alıp plan ve programlarını yapmaktadır. Ülkemiz ise iktisadi ve sosyal bakımdan gelişen bir ülke olmasından dolayı her geçen gün biraz daha fazla enerjiye ihtiyaç duymaktadır. Ülkemiz enerji kaynakları bakımından çok zengin topraklara sahip değildir. Bu sebepten dolayı fosil enerji kaynaklarını kullanarak enerji üretmek ülke ekonomisine ciddi yük oluşturmaktadır[3]. Ayrıca sadece ülkemizde değil Dünya’da da fosil tabanlı enerji kaynaklarının sınırlı ve pahalı olması yine özellikle kirletici özellikleri ile çevreyi kirletmesi nedeniyle sera gazı salınımı gittikçe artmaktadır. Bu nedenle iklim kuşakları değişmektedir. Tüm bu menfi durumları ortadan kaldırmak ve ülkemizi enerji kaynağı açısından dış ülkelere olan bağımlılığını en aza indirmek için yenilenebilir enerji kaynaklarını kullanmak gerekmektedir[4]. Türkiye’de ki enerji açısından politik yapılara bakacak olursak enerji politikasında duyulan ihtiyacın karşılanmasına yönelik atılan adımlar olmuş ve bu zaman süresince enerji kaynaklarının verimliliğine çok fazla önem verilmemiştir. Son zamanlarda yapılan çalışmalarla duyulan ihtiyaçtan ziyade enerji kaynaklarının verimliliği de önemli ölçüde dikkate alınmış ve bu değişimin en büyük göstergesi ise 2007 yılında yürürlüğe giren enerji verimliliği kanunudur. Türkiye birincil enerji tüketimine bakacak olursak; doğalgaz %24.1, petrol %33.3, Kömür %28.1, Nükleer %4.4, hidrolik %6.9, diğer yenilenebilir %3.2olduğunu görürüz. Tüketimin %85.5’lük kısmı fosil yakıtlı kaynaklardan oluşmaktadır[5].

1

Corresponding author. Tel.: +0-000-000-0000 ; fax: +0-000-000-0000 . E-mail address: ylcnkpln@mynet.com


43 Bütün bu veriler dikkate alındığında ülkemizin çok hızlı bir şekilde yenilenebilir enerji kaynaklarını kullanması gerekmektedir. Ülkemiz yenilenebilir enerji kaynağı açısından zengin bir ülkedir. Temiz bir çevre ve ucuz bir enerji kaynağı için tüm çalışmaları ve planları bu doğrultuda yapmalı ve gereken adımları hızlı bir şekilde atmalıyız[6]. Bu çalışmada yenilenebilir enerji kaynağı olan güneş enerjisinin en verimli bir şekilde kullanmak için gerekli olan güneşlenme süresi tahmini yapılmıştır. Bu çalışmadaki amaç güneşlenme süresinin en iyi sonuçlara göre tahminini yapıp planları ve programları bu doğrultuda yapmaktır. Literatür taramasında ortaya çıkan bu yöntemlerden bazıları; YSA[7], Regresyon analizi[8], Bulanık mantık[9], Genetik algoritma[10], Zabara method[11] gibi metotlardır. Keskiner ve arkadaşları, YSA ve regresyon yöntemi ile sıcaklık tahmini yapmışlardır[12], Doğdu, yaptığı çalışma ile jeotermometre ile sıcaklık tahmini yapmak için yazılım geliştirmiştir[13]. Minaz, bulanık mantık ile basınç, sıcaklık ve rüzgâr hızı tahmini yapmıştır[14]. Li ve arkadaşları, güneşlenme süresini kullanarak güneş radyasyonu tahmini yapmışlardır[15]. Şahin, Türkiye’de bulunan yirmi ilin sıcaklık tahminini YSA kullanarak yapmıştır[16]. Venkadesh ve arkadaşları, genetik algoritma ve YSA kullanarak sıcaklık tahmini yapmışlardır[17]. Güneşlenme süresi tahminini yapılırken metot olarak yapay sinir ağları(YSA) ve genetik algoritma kullanılmıştır. Ayrıca YSA’da geri beslemeli ağ ve ağırlık hesabında algoritma olarak ise Gradient-Descent öğrenme algoritması kullanılarak iki girişli ve 3 katmanlı bir model tasarlanmıştır. Bulunan değerler ile hata oranları tespit edilmiş ve karşılaştırılması yapılmıştır.

2.

Güneş Enerjisi

Güneş enerjisi yenilenebilir enerji kaynakları içerisinde her geçen gün yıldızı parlayan bir enerji kaynağıdır. Güneş enerjisi bakımından ülkemiz son derece zengin bir ülkedir. Güneş enerjisinden maksimum şekilde yararlanmak için, güneşlenme süresinin en fazla olduğu alanı belirlemek ve planlamayı buna göre yapmak gerekmektedir. Şekil 1’de Türkiye güneş enerjisi radyasyon atlası görülmektedir. Şekli incelediğimizde Türkiye’nin büyük bölümünün güneş enerjisinden yararlandığı görülmektedir.Toplam güneş radyasyonunu ele alırsak, mavi kısımlar 1400-1550 KWh/m2-yıl, sarı kısımlar 1550-1750 KWh/m2-yıl, kırmızı renkle gözüken kızımlar ise 1750-2000KWh/m2-yıl temsil etmektedir[18].

Şekil 1.Türkiye Güneş enerjisi potansiyelatlası[18]

Ayrıca bir ay içerisinde ki bir günlük toplam güneşlenme süresi Şekil 2’de gösterilmiştir[18].


44

Şekil 2.Türkiye’nin ay içerisinde ki bir günlük güneşlenme süresi (saat)[18]

3. 3.1.

Metot Yapay Sinir Ağları (YSA)

Yapay sinir ağları, biyolojik sinir sisteminden esinlenerek ortaya çıkarılan matematiksel modelleme şeklidir. YSA’ ları insanların algılama yetisi gibi çalışır. İnsanların beyinlerini taban alarak çalışan bu modeller halen insanların sinir yapıs ına göre çok ilkel kalmaktadır[19].

Şekil 3.Biyolojik sinir hücresi[19]

Şekil 3’de basit bir biyolojik sinir hücresi görülmektedir. Basit bir sinir hücresinde çekirdek, soma, akson tepeciği, dendrit, akson ve sonlandırıcı düğümler bulunmaktadır. 3.2.

YSA’ ların genel özellikleri

YSA’ların genel olarak baskın bir karakteristik sergilemesine rağmen yine de birkaç genel özelliği bulunmaktadır. Birinci özelliği paralel olması, ikinci özelliği genelleme yapabilmesi, üçüncü özelliği doğrusal olması, bir diğer özelliği ise gerçekleştirilebilir olması. 3.3.

YSA’ ların uygulama alanları

YSA’ lar günümüzde birçok alanda kullanılmaktadır. Bunlar arasında modelleme, sınıflandırma, genelleme bulunmaktadır. YSA’ların uygulama alanlarına bakacak olursak karmaşık çözümü olmayan ve ya gerçek bir karmaşık hale sahip sorunların çözümünde başarılı olduğu görülmektedir. Bu durumda YSA’ lar aşağıdaki durumları meydana getirir[19].


45 Tablo 1.YSA’ların kullanıldığı alanlar[19]

Sınıflandırma Kümeleme Tahmin yapmak Optimizasyon Kontrol ilişkileri Sinyal işleme Veri sıkıştırma 3.4.

YSA hücresi

YSA hücresinde giriş verileri (Xn), ağırlıklar (Wn), toplama fonksiyonu (Toplam), eşik değeri (b), aktivasyon fonksiyonu (f), ve çıkış ise (sonuç) olarak adlandırılmıştır. Şekil 4’ da gösterilmiştir[20].

Şekil 4.YSA hücresi[20]

Dışarıdan gelen veriler hücre girişinden içeriye sokulur, ağırlıklarla giriş verileri çarpılarak toplam fonksiyon elde edilir. Toplam fonksiyonuna sabit değer eklenerek aktivasyon fonksiyonu giriş verisi elde edilir, aktivasyon fonksiyonunda işlenen veri çıkış verisi olarak çıkışa aktarılır. Denklem 1’de ayrıntılı formülü gösterilmiştir[20]. n

Sonuç  f ( x n wn  b)

(1)

i 1

3.5.

Sigmoid fonksiyonu

Sigmoid fonksiyonu YSA fonksiyonları arasında genelleme yapmak için kullanılan fonksiyondur. Türevi alınır ve 0 ile 1 arasında değerler alır. Sigmoid fonksiyon formülü Denklem 2’de gösterilmiştir[19].

f  3.6.

1 1  e  sonuç

(2)

Tanjant hiperbolik fonksiyonu

Bu fonksiyon YSA içerisinde sonuç almaya yarayan fonksiyondur. Doğrusal olmayan fonksiyondur ve -1 ile +1 arasında değer alır. Bu fonksiyonda Denklem 3’teki formül kullanılır[19].

f 

e sonuç  e  sonuç e sonuç  e  sonuç

(3)


46

3.7.

Gradient-Descent algoritması

Gradient-Descent algoritması YSA modellerinde sıklıkla kullanılan bir algoritma çeşididir. Denklem 4,5,6,7 formülasyonu olarak kullanılmaktadır[19].

j (k ) 

1 NT e ( k )e N ( k ) 2

(4)

e N (k ) = Ağın hata oranını verir.

WijL (k )  WijL (k  1)  eT (k )

y (k )  ijL (k )

(5)

Çıkış katmanı içerisinde;

 iL (k )  e(k ) T

df i L ( x iL (k )) dxiL (k )

(6)

Başka katmanlar içerisinde;

 iL (k ) 

df i L ( x iL (k )) dxiL (k )

Wi L 1 (k ) L 1 (k )

(7)

Denklemlerinin sonucu verdiği gözlenmektedir[21]. 3.8.

Genetik Algoritma (GA)

Genetik algoritma, çözümü çok zor olan soruların çözüme kavuşturulması için ortaya çıkarılan popülasyon tabanlı sezgiye dayanan bir algoritmadır[22]. Genetik algoritma, tabiatta olağan şekilde bulunan evrimsel süreci göz önünde bulundurarak birebir aynı süreci gerçekleştirmeye çalışır. GA amaç fonksiyonuna göre çalışır amaç fonksiyonunu en ideal hale getirmek için uğraşır. Problemlerin kıstasları kromozom ve genlerle temsil edilir. Sorunların bireylerin gösterimi probleme göre değişir. Sorunun çözümünü temsil eden bireydir ve GA’ nın başarısını gösterir. GA’da başlangıç popülasyonu genellikle random seçilir ve her bir sonuç kromozomlarla gösterilir. İleriki iterasyonda çaprazlama ve mutasyon kullanılır. Bir popülasyon içerisindeki bir çözüm diğer çözümle karşılaştırılır ve bir değer verilir. Bu değer uyum değeridir ve popülasyonun yaşamasına imkan verir[23]. 3.9.

GA’nın kullanıldığı alanlar

Genetik algoritmalar günümüzde birçok alanda kullanılmaktadır. Bunları sıralayacak olursak; tıp, istatistik, işletme, muhasebe, işaret işleme, sinyal işleme, lineer ve lineer olamayan kontrol sistemleri, robotik uygulamaları, örüntü tanıma, ses tanıma, planlama, yön bulma,yapay zeka, yöneylem problemleri, optimizasyon, ağ tasarımı vb. alanlarda başarıyla uygulanmaktadır[23]. Tablo 2.Genetik algoritmaların adımları[24]. İ=0 Kaç nesil üretileceğini belirle; (MaksNesil) Başlangıç popülasyonunu oluştur;(Popi) Popi ‘yi değerlendir; İ<MaksNesil olduğu müddetçe aşağıdaki adımları tekrarla; Çaprazlama ve mutasyonu kullanarak Popi+1 ‘i


47 oluşturPopi+1 i değerlendir İ=i+1 Sonlandır İdeal çözümü ver

Tablo 2’ de genetik algoritmaların adımları gösterilmiştir[23].

4.

Normalizasyon

Çalışmada kullanılan değerler sayısal değerler olduğundan işlem kolaylığı ve yer kaplaması açısından sıkıntıya düşmemek adına, gerçek sayıları 0,1 ile 0,9 arasına oranlanır. Bu işleme normalizasyon denir. Normalizasyon yapıldıktan sonra en büyük sayı 0,9 en küçük sayı ise 0,1 olmaktadır. Geriye kalan sayılar ise 0,1 ile 0,9 arasında değerler almaktadır. Denklem 8’de normalizasyon formülü bulunmaktadır[19].

X ( n)  ( 4.1.

X  X min ) * 0.8  0.1 X max  X min

(8)

Ortalama karekök hatası(OKH)

Ortalama karekök hatası, elde edilen tahmini değerler ile gerçek değerleri kıyaslama yaparak elde edilir. Formülü Denklem 9’da gösterilmiştir[19].

OKH 

1 N

N n 1

( Rn  Fn ) 2

(9)

Rn = Gerçek değer Fn = Tahmin değeri

5.

Uygulama

Yapılan bu çalışmada Amasya ilinin güneşlenme süresi tahmini yapılmıştır. Bu tahmin gerçekleştirilirken güneşlenme süresine etkisi olduğu bilinen nem değerleri de giriş verisi olarak kullanılmıştır. Yapay sinir ağları geri besleme modelinde ağırlık optimizasyonunda gradient-descent algoritması kullanılarak bir model tasarlanmıştır. Ayrıca genetik algoritma kullanılarak bir model daha tasarlanmıştır. İlk olarak giriş verileri geçmiş güneşlenme süresi ve güneşlenme süresine etkisi bilinen nem verileridir. Oluşturulan genetik algoritma modelinde 100 kromozomdan oluşan bir popülasyon oluşturulmuş, çaprazlanma oranı 0.8, mutasyon olasılığı ise 0.05 kullanılmıştır. İkinci metot olarak YSA modelini anlatacak olursak; iki adet giriş verisi kullanılmıştır. Bunlar güneşlenme süresi ve nem değerleridir. Amasya ilinin 2000-2010 yılları arası aylık değerleri kullanılmıştır. 2000-2008 yılları arası eğitim için kullanılırken, 2009 değerleri test verisi olarak kullanılmıştır. Eğitim verileri test verileri olarak kullanılmamıştır. 3 katmanlı geri beslemeli bir ağ oluşturulmuş ilk ve ara katmanda logaritmik sigmoid fonksiyonu, son katmanda ise tanjant sigmoid fonksiyonu kullanılmıştır. Genetik algoritma kullanılarak oluşturulan model de ise giriş verileri olarak güneşlenme süresi ve nem verileri kullanılmıştır. Yine Amasya ilinin 2000-2010 yılı aylık güneşlenme süresi, nem verileri ile bir model oluşturulmuştur. Oluşturulan modelde 2010 yılının aylık güneşlenme süresi tahmini yapılmıştır.


48

Güneşlenme süresi ve nem verileri ile GD algoritma tahmin sonuçları

1 GERÇEK TAHMİN

0.9

Normalize güneşlenme süresi verileri

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0

1

2

3

4

5

6

7

8

9

10

11

12

13

Aylar

Şekil 5.YSA gradient-descent algoritma sonuçları Güneşlenme süresi ve nem verileri ile Genetik algoritma tahmin sonuçları

1 GERÇEK TAHMİN

0.9

Normalize güneşlenme süresi verileri

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0

1

2

3

4

5

6

7

8

9

10

11

12

13

Aylar

Şekil 6. Genetik algoritma sonuçları

Şekil 5’de gösterilen grafik YSA sonuçları ile gerçek sonuçların karşılaştırılmasını göstermektedir. Uygulama MATLAB kullanılarak yapılmıştır. Şekil 6’da gösterilen grafik ise genetik algoritma kullanılarak yapılmıştır. Bu uygulamada MATLAB kullanılarak yapılmıştır.

6.

Sonuç

Yapılan çalışmada elde edilen sonucu irdeleyecek olursak, iki model oluşturulmuştur. Bu modeller YSA ve genetik algoritma modelleridir. Modellere tahminler yaptırılmış ve 2010 yılının aylık güneşlenme süreleri normalizasyon değerleri bulunmuştur. Algoritmaların değerleri kendi aralarında gerçek değerle karşılaştırılarak OKH tespit edilmiştir. Elde edilen değerler Tablo 3’te gösterilmiştir. Tablo 3.Algoritmaların OKH değerleri

Uygulama yöntemleri

Ortalama karekök hata oranı

Gradient-Descent Algoritma

0,0154

Genetik Algoritma

0,0113


49 Yapılan çalışmada OKH değerlerini incelediğimizde güneşlenme süresi tahmin işleminde genetik algoritmanın son derece başarılı olduğu gözlemlenmiştir. Ayrıca YSA gradient- descent algoritmasının da son derece makul sonuçlar verdiği gözlemlenmiştir. Ancak hata oranlarına bakıldığında genetik algoritmanın daha başarılı olduğu görülmüştür.

7.

Kaynaklar

[1]Variyenli H.İ., Menlik T., Özkaya M.G., ”Isı enerjisi destekli bir kompresörün buhar sıkıştırmalı soğutma sistemindeki performansının deneysel incelenmesi” Gazi Üniv. Müh. Mim. Fak. Der. Cilt 26, No:1,1-8, 2011. [2] Özdemir V., “Türkiye’nin karbonizasyon indeksinin temel enerji göstergelerine bağlı olarak yapay sinir ağları ile tanımlanması”Gazi Üniv. Müh. Mim. Fak. Der. Cilt 26, No:1,9-15, 2011. [3] Ucar A. Balo F., “Evaluation of wind energy potential and electricity generation at six locations in Turkey” Applied Energy 86(1):1864-1872,2009. [4] Dünyada yenilenebilir enerji, Nisan 2017, sayı:197 , Seta Raporu [5] “Türkiye’nin enerji görünümü”,Nisan 2016. [6] Lüy, M. Saray, U., Wind Speed Estimation For Missing Wind Data With Three Different Backpropagation Algorithms. Energy Education Science and Technology Part A, Cilt 30(1), 45-54, 2012 [7] Kayri M, Kayri İ, Gencoglu M.T, “The performance comparison of Multiple Linear Regression, Random Forest and Artificial Neural Network by using photovoltaic and atmospheric data” 14th International Conference on Engineering of Modern Electric Systems (EMES), 1-2 June 2017. [8]Meenal R, Selvakumar I.A,“Estimation of global solar radiation using sunshine duration and temperature inChennai” International Conference on Emerging Trends in Engineering, Technology and Science (ICETETS) 24-26 Feb 2016. [9] Hussain S, Alili A.A, “Soft Computing Approach for Solar Radiation Prediction Over Abu Dhabi, UAE: A Comparative Analysis” IEEE International Conference on Smart Energy Grid Engineering (SEGE), 17-19 Aug,p1-6, 2015. [10] Al-Hajj R., Assi A.,Batch F.,“An Evolutionary Computing Approach for Estimating Global Solar Radiation”. 5th International Conference on Renewable Energy Research and Applications. Birmingham UK. 20-23 Nov. 2016 [11] Ismailou A,Han M,“Estimation of Solar Radiation and PV Output Power in CAMEROON” APPEEC’11 Procedings of the 2011 Asia-Pasific Power and Energy Engineering Conference,Washington USA, p1-4, 2011 [12] Keskiner A.D., İbrikçi T., ve Çetin M., “Yapay sinir ağlarıyla coğrafi bilgi sistemi ortamında olasılıklı sıcaklık tahmini ve karşılaştırılması”, Ankara Üniversitesi Ziraat Fakültesi Tarım Bilimleri Dergisi-Journal of Agricultural sciences 17.3,2012. [13] Doğdu M.Ş., “Jeotermal suların rezervuar sıcaklığının tahmininde kullanılan jeotermometre hesaplamaları için bilgisayar programı”, Jeoloji Mühendisliği Dergisi 30(1), 2006. [14] Bilecik İlinin Farklı Yöntemler Kullanılarak Basınç, Sıcaklık ve Rüzgâr Hızı Tahmini [15] Li H., Ma W., Lian Y., Wang X., Zhao L., “Global solar radiation estimation with sunshine duration in Tibet, China”, Renewable Energy, Cilt 36, 3141-3145, 2011. [16] Şahin M., “Modelling of air temperature using remote sensing and artificial neural network in Turkey”, advances in Space Research Cilt 50,973-985, 2012. [17] Venkadesh S., Hoogenboom G., Pottera W., McClendon R., “A genetic algorithm to refine input data selection for air temperature prediction using artificial neural networks”, Applied Soft Computing 13, 2253-2260, 2013. [18]http://www.eie.gov.tr/MyCalculator/Default.aspx (Erişim tarihi: 01.10.2017) [19]Saray U.,“Rüzgar potansiyelinin yapay sinir ağlarıyla analizi ve uygulaması”, Kırıkkale Üniversitesi Yüksek Lisans Tezi, Kırıkkale, 2012.


50

[20] Structural learning in artificial neural networks using sparse optimization [21] Training qubit neural network with hybrid genetic algorithm and gradient descent for indirect adaptive controller design [22] D. Goldberg, Genetic Algorithms in Optimization, Search and Machine Learning, Addison Wesley, 1989. [23] Yigit V., “Genetik Algoritma ile Türkiye Net Elektrik Enerjisi Tüketiminin 2020 Yılına Kadar Tahmini”, International Journal of Engineering Research and Development, Vol.3, No.2, June 2011.


INTERNATIONAL SCIENTIFIC AND VOCATIONAL JOURNAL (ISVOS JOURNAL)

Vol.1

Issue:1

Date: December 2017

Received:26.12.2017

Accepted:29.12.2017

Final version : 31.12.2017

International Scientific and Vocational Journal ISVOS Journal Improved Compound Multiphase Waveforms with Additional Amplitude Modulation (periodic mode) for Marine Radars V. Koshevyy a, O. Pashenko a,1 a National

University Odessa Maritime Academy, Odessa, Ukraine

Abstract This paper has presented the basis of a compound multiphase waveform design with additional amplitude modulation, capable of controlling a waveform pick-factor, suitable for use with marine radar. The waveform shows good Doppler tolerance, with the low side-lobes performance maintained over the central zone of an ambiguity function. A clear waveform design procedure has been presented that does not require the implementation of numerical optimization procedures. It has been shown that a compromise between side-lobes suppression and the value of pick-factor can be found. These waveforms allow achieving better results as compared to compound multiphase waveforms without additional amplitude modulation under mismatched weighting filtering. In this article we considered periodic mode of Radar. Keywords: “composite signal, ambiguity function, cross-ambiguity function, amplitude modulation, periodic mode”

The compound multiphase signal consists of multiplication of two sequences; each of them is a signal with a quadratic variation of phases. Expressions for complex amplitudes for the base (snB) and external (snV) signal and final multiphase compound signal’s sequence (sn) are follows [1,2]: 2       1  N  µ   snB  exp  j    2  n  N B E  n B 0B     N B      4  

    n    E    N B1            1  N  µ sVn  exp  j    2  E  n  N E V V 0V   N B1     4 N V               sn  snV  sn B

(1)      

2

   , n  0, N  1   

(2)

(3)

where αʹ=αТ02; βʹ=β(Т0NB)2; α, β, µ0B, µ0V, NB1– parameters of phase modulation; T0 – duration of one pulse; NB T0- period of the base sequence; NV T0- period of the external sequence; NT0 = NBNV T0- period of the sequence; E[x] - integer part of x. If NB is even so µ0B=0, if NB is no even so µ0B =1. µ0V is formed by the same way. As opposite to previous prepared article (we considered aperiodic mode of Radar) [3], in this article we considered periodic mode, which is important for building continuous wave radar. For illustration three types of periodic compound multiphase signals were considered. The length of sequence is N=324 (NB=18, NV=18), µ0B=µ0V=0 with parameters: 1. 2.

3.

1

α'=-1/NB, β' =1/NB2. The signal with such parameters has a maximum ratio of the free zone (FZ) area around the central peak (CP) to the CP topographic section area (on the zero level) [2]. α'=-1/NB, β' =1/NB. The side-lobes level is increased in the FZ region for these parameters in comparative to the case 1). But the Autocorrelation Function (ACF) has the lowest side-lobes level for its entire length. It should also be noted that FZ decreases around CP of Ambiguity Function (AF) [4]. α'=-1/NB, β'=0.

Corresponding author. Tel.: +0-000-000-0000 ; fax: +0-000-000-0000 . E-mail address: vmkoshevyy@gmail.com


52 At the same time the periodic compound multiphase signals with the length of sequence N=306 (NB=18, NV=17), α'=1/NB, β' =2/NV, µ0B =0, µ0V =1 were considered.. Signals with an additional amplitude modulation (AM) can be described by the following expression:

snam  sn * vn vn 

vnb E  n / N  N B B

v

v

 n  E  1  NV 

(4)

, n  0  N  1,

where snam - compound multiphase signal with an additional AM; sn - complex envelope of the signal; vnb, vnv - weighting coefficients for the base and external sequences. The expression of the AF for the periodic compound multiphase signal with an additional AM has the form:

 ss  k , l  

N 1

snam *  s(am nk )  e

i

2 ln 4N

(5)

n 0

where k - discrete values of time delay with the selected step T0; l - discrete values of frequency. It’s convenient to use the pick-factor parameter for describing behaviour of AM modulation [5]. It is defined as the ratio: 2



snam_ max N

N 1 n 0

snam

(6)

2

The law of the changing weighting function sin is considered in the paper. This function allows controlling the behaviour of AF. Its advantage is possibility to control the signal peak-factor and the side-lobes level of AF.   y n vn B  sin  ( B  ) , n  0  N B 1 N  1 N  z B B B   vnV

(7)

  y n  sin  ( V  )  , n  0  NV  1 N  1 N  z V V V  

NB 1 , 2 2 y N  ( NV  1) N 1 2 y N  ( N B  1) ; zB , zV - variables for the basic and external sequences are zB  B B , zV  V V . 1  yV  V N B  1  2 yB NV  1  2 yV 2 where y B , yV - the positive numbers for the base and external sequences, respectively, varies in the range 1  yB 

The formula (8) describes other one of the possible weighting functions for the signals (3):   y n vn B  ( sin  ( B  ) )2 , n  0  N B  1 N  1 N  z B B B    y n vnV  ( sin  ( V   NV  1 NV  zV

(8)

 ) ) 2 , n  0  NV  1 

The structure of AF of multiphase compound signals with an additional AM is compared with Cross-ambiguity function (CAF) compound multiphase signals without AM, but with mismatched treatment using weighing functions sin (7) and sin2 (8). On the left upper part of the Figure the body of CAF in the central peak area is represented. Lower the sections of CAF along axe of range for different discrete values of Doppler frequency (l=0, l = 1, l= 2, l = 3, l=4) in FZ (central peak area) are represented. On the right upper part of the Figure the law of the phase modulation of the composite multiphase signal is illustrated. Lower the Cross correlation function (section of CAF by the plane l=0) on all it’s length is presented. And more lower the sections of CAF along Doppler axe b y the planes (k=0; k=1; k=2; k=3).


53

a) b) Fig. 1. a) The body of the CAF of the periodic multiphase compound signal without an additional AM, N=324 (NB=18, NV=18) α '=1/NB, β'=1/NB, µ0V = µ0B = 0, NB1 = NB, with the weighting function sin (yB = yV = 1), phases of the signal, sections of the body of the CAF; b) the body of the AF periodic multiphase compound signal with an additional AM, N=324 (NB=18, NV=18) α '=-1/NB, β'=1/NB, µ0V = µ0B = 0, NB1 = NB, with the weighting coefficients sin (yB = yV = 1), phases of the signal, sections of the body of the AF.

As a result (Fig.1,2,3), side-lobes level is significantly reduced after using the periodic compound multiphase signals with an additional AM. CP area isn’t expanded significantly. By increasing the yB and yV, the value of peak-factor is decreasing. By changing the parameters yB and yV we can control the peak-factor of the signal. Table 1 shows these values: Table 1. The peak-factor dependence from parameters yB=yV for periodic multiphase compound signal with an additional AM, N=324 (NB=18, NV=18) α '=-1/NB, β'=1/NB, µ0V = µ0B = 0, NB1 = NB, (matched processing) with the weighting coefficients sin

yB=yV

with the weighting coefficients sin

1 2 3 5 6 7 8 9

3.59 2.83 2.24 1.5 1.28 1.13 1.05 1.01

with the weighting coefficients sin2 ξ 6.38 4.97 3.77 2,09 1,6 1,28 1.09 1.01


54

a)

b)

Fig. 2. a) The body of the CAF of the periodic multiphase compound signal without an additional AM, N=324 (NB=18, NV=18) α '=-1/NB, β'=1/NB, µ0V = µ0B = 0, NB1 = NB, with the weighting function sin (yB=yV = 2), phases of the signal, sections of the body of the CAF; b) the body of the AF periodic multiphase compound signal with an additional AM, N=324 (NB=18, NV=18) α '=-1/NB, β'=1/NB, µ0V = µ0B = 0, NB1 = NB, with the weighting coefficients sin (yB=yV =2), phases of the signal, sections of the body of the AF The calculations were made for three types of the signals, as indicated above: α'=-1/NB, β' =1/NB2, N=324 (NB=18, NV=18). α'=-1/NB, β' =1/NB, N=324 (NB=18, NV=18), α'=-1/NB, β' =0, N=324 (NB=18, NV=18). The three kinds of signals with different sets of parameters were compared. The compound multiphase signal with an additional AM N=324 (NB=18, NV=18), α '=-1/NB, β'=1/NB, µ0V= μ0B = 0, NB1 = NB, (matched processing) with the weighting function sin shows the highest ratio of the CP to the side-lobes level. The signal with β'=1/NB2 shows almost the same. For example, we demonstrated the cases of signals with parameters α'=-1/NB, β'=1/NB, N=324 (NB=18, NV=18), yV=yB =1,2 (Fig.1,2). They have the highest side-lobe level in the FZ region.


55

a)

b)

Fig. 3. a) The body of the CAF of the periodic multiphase compound signal without an additional AM, N=324 (NB=18, NV=18) α '=-1/NB, β'=1/NB, µ0V= µ0B = 0, NB1 = NB, with the weighting function sin2 (yB=yV = 2) for (8), phases of the signal, sections of the body of the CAF; b) the body of the AF periodic multiphase compound signal with an additional AM, N=324 (NB=18, NV=18) α '=1/NB, β'=1/NB, µ0V = µ0B = 0, NB1 = NB, with the weighting coefficients sin2 (yB=yV =2) for (8), phases of the signal, sections of the body of the AF

Thus, the lowest side-lobes level was given by signals with peak-factor ξ = 3,59 and ξ = 2,83 for weighting function sin, while yV=yB= 1,2. Peak-factor for signals with lowest side-lobes level were ξ =6,38 and ξ =4,97 (for weighting function sin2). For comparison AF body for rectangular shape of compound signals is presented on Figure 4.

Fig. 4. The body of the AF of the periodic multiphase compound signal, N=324 (NB=18, NV=18), α '=-1/NB, β'=1/NB, µ0V= µ0B = 0, NB1 = NB, phases of the signal, sections of the body of the AF

Undoubtedly, the compound signal on Fig. 4 has the highest side-lobes level in the FZ region. This is a big drawback of this type processing. The signals with different pulses’ quantity of base and external sequence (NB=18, NV=17) are considered. Parameters of signal are α'=1/NB, β'=2/ NV, µ0V =1, μ0B= 0, NB1 = NB.


56

a)

b)

Fig. 5. a) The body of the CAF of the periodic multiphase compound signal without an additional AM, N=306 (NB=18, NV=17) α'=1/NB, β'=2/NV, µ0V =1, µ0B = 0, NB1 = NB, with the weighting function sin (yB=yV =1), phases of the signal, sections of the body of the CAF; b) the body of the AF periodic multiphase compound signal with an additional AM, N=306 (NB=18, NV=17) α'=1/NB, β'=2/NV, µ0V =1, µ0B = 0, NB1 = NB, with the weighting coefficients sin (yB=yV =1), phases of the signal, sections of the body of the AF

a)

b)

Fig. 6. a) The body of the CAF of the periodic multiphase compound signal without an additional AM, N=306 (NB=18, NV=17) α'=1/NB, β'=2/NV, µ0V=1, μ0B = 0, NB1 = NB, with the weighting function sin2 (yB=yV = 3) for (8), phases of the signal, sections of the body of the CAF; b) the body of the AF periodic multiphase compound signal with an additional AM, N=306 (NB=18, NV=17) α'=1/NB, β'=2/NV, µ0V=1, μ0B= 0, NB1 = NB, with the weighting coefficients sin2 (yB=yV =3) for (8), phases of the signal, sections of the body of the AF

The results of research for these cases are presented in Table 2:


57 Table 2. The peak-factor dependence from parameters yB=yV for periodic multiphase compound signal with an additional AM, N=306 (NB=18, NV=17) α'=1/NB, β'=2/NV, µ0V=1, μ0B = 0, NB1 = NB, (matched processing) with the weighting coefficients sin and sin2

yB=yV

with the weighting coefficients sin

1 2 3 5 6 7 8 9

3.58 2.80 2.21 1.46 1.25 1.11 1.03 1.00

with the weighting coefficients sin2 ξ 6.36 4.92 3.69 2.01 1.53 1.24 1.07 1.01

As we can see, if we reduce quantity of pulses in the signal the pick-factor is reduced. The side-lobes level in the FZ region does not hesitate in large limits if we use N=306 or N=324. But periodic multiphase compound signal with an additional AM with N=306 have lower side-lobes level on all sections of the body of the AF (for l=0), without high side peak. To summarize, periodic multiphase compound signal with an additional AM have the lowest side-lobes level. A matched processing was used, so we can say this type of signal does not have signal to noise losses. These results can be used not only for the compound multiphase signals, but also for equivalent compound LFM signals [6,7,8].

Fig. 7. The body of the AF of the periodic multiphase compound signal, N=306 (NB=18, NV=17) α'=1/NB, β'=2/NV, µ0V=1, μ0B= 0, NB1 = NB, phases of the signal, sections of the body of the AF

To summarize, periodic multiphase compound signals with an additional AM have the lowest side-lobes level in FZ and have good Correlation functions with low sensitivity to the Doppler shifts. A matched processing was used, so this type of signal does not have signal to noise ratio losses. These results can be used not only for the compound multiphase signals, but also for equivalent compound LFM signals [6,7,8].


58

References [1] V.M. Koshevyy. Synthesis of Waveform-Filter pairs under Additional Constrains with Group-Complementary Properties IEEE, Radar Conference 2015, May 2015, Arlington,VA (USA), pp 0616-0621. [2] V.M. Koshevyy, Synthesis compound multiphase signals, Izvestiya VUZ. Radioelectronika (Radioelectronics and Communication Systems), vol. 31, N8, 1988, pp. 56-58. [3] V. Koshevyy & O. Pashenko, Improved Compound Multiphase Waveforms with Additional Amplitude Modulation (periodic mode) for Marine Radars, Marin Navigation and Safety of Sea Transportation. Activities in Navigation, 2017 [to be published]. [4] V.M. Koshevyy, V.I. Kuprovskyy. Investigation of properties of compound multiphase signals.Izvestiya VUZ. Radioelectronika (Radioelectronics and Communication Systems), N8, 1991, pp. 63-66. [5] Ch.E. Cook, M. Bernfeld, Radar Signals. An Introduction to Theory and Application, Artech House, Inc., Boston 1993. [6] B.L. Lewis, F.F. Kretschmer, Linear frequency modulation derived polyphase pulse compression codes, IEEE Trans. on Aerospace and Electronic Systems Vol. AES-18, N5, Sept. 1982, pp. 637-641. [7] N. Levanon, E. Mozeson, Radar signals, J. Wiley, NJ, 2004. [8] V. Koshevyy & O. Pashenko, Signal Processing Optimization in the FMCW Navigational Radars, Marin Navigation and Safety of Sea Transportation. Activities in Navigation. (edited) Adam Weintrit. CRC Press. 2015 pp. 55-60

Profile for isvos journal

Vol 1 issue 1 - isvosjournal  

The International Scientific and Vocational Studies(ISVOS) Journal is an peer-reviewed internationally respected journal which is published...

Vol 1 issue 1 - isvosjournal  

The International Scientific and Vocational Studies(ISVOS) Journal is an peer-reviewed internationally respected journal which is published...

Advertisement