Sustainability Matters Apr/May 2022

Page 17

smart water

How much should the water industry trust

artificial intelligence? Chaim Kolominskas, Manager EVS Water, Envirosuite

Deterministic and AI learning approaches should be used together as the industry continues along its journey of digitalisation.

to run tests, or run offline simulations or

I

studies. This is slow and impractical. Leaders in the sector are looking to alter-

dentification and interpretation

augmented with sensors that collect data to

native options such as digital twin technology

of patterns and relationships

support control of the process.

as a way of creating a virtual representation

in the vast amounts of data

As engineers and plant managers, we

of their operations, where changes can be

collected by water utilities has

know and trust deterministic models. These

simulated — and implications understood

the potential to drive better

models have proven reliable over the years

— quickly and with no impact to business-

operating decisions and drive reductions in

to represent and simulate processes such

as-usual operations. Automated deterministic

cost; however, engineers and decision-makers

as chemical reactions in drinking water

models, connected in near real time, are ideal

are sceptical of using artificial intelligence

or industrial wastewater treatment. If the

for this purpose, but they are reactive and

(AI) or machine learning (ML) as standalone

inputs into a process are known, the model

cannot deliver useful projections on what

approaches.

can produce a set of outputs that we use

you should do to avoid an issue.

to make decisions. It’s accepted that best

Hybrid approaches that combine deter-

as ‘black boxes’ that produce non-transparent,

practice models can represent a process

ministic and AI approaches can overcome

unexplainable outcomes and require constant

accurately enough to be used to understand

the limitations of both methods. Deterministic

oversight and supervision. Large amounts of

the implications of changes to the process.

modelling can be used to represent total

clean, historical and granular data are needed

However, these approaches are manual and

process performance, grounded in accepted

to be accurate. More importantly, experienced

traditionally require substantial technical

science of the industry with AI used to drive

operators and practitioners in the industry see

expertise to drive. Deterministic models are

accurate forecasts and recommendations for

the lack of connection to the industry itself as

also by their very nature not conducive to

short-term actions to improve performance.

a weakness. An ML model on its own does

forecasting.

Keeping personnel involved to oversee and im-

Algorithms are often presented and sold

plement recommendations remains important

not care about the industry that it is applied

The water industry needs to change

to and makes no connection to the accepted

to meet increasingly stringent operational,

physics, chemistry and biology of the process.

regulatory and environmental requirements,

At Envirosuite, we see the value in both

How should the water industry realise the

as well as the needs of surrounding com-

deterministic and AI learning approaches and

benefits that AI can deliver, while managing

munities. Understanding the implications of

believe that they should be used together as

the risks that come with the approach?

while trust is built in the system.

what doing something different would look

the industry continues along its journey of digi-

Digitalisation is driving change in the water

like is, however, a key challenge still fac-

talisation. As approaches evolve and improve,

and wastewater treatment sectors, just as it

ing water utilities. It is currently difficult to

we look forward to the further opportunities

is in other industrial and corporate environ-

measure or understand the applicability of

that AI could deliver to the water industry,

ments. Water and wastewater treatment plants

alternative operating scenarios. Operators

as long as we keep the humans in the loop.

collect more data than ever before. Control

and engineers must either live test and see

Envirosuite Operations Pty Ltd

and treatment equipment is increasingly

what happens, take part of the plant offline

https://envirosuite.com/platforms/water

This issue is sponsored by — Schneider Electric — se.com/au/getreadyformore 17


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.
Sustainability Matters Apr/May 2022 by Westwick-Farrow Media - Issuu