Increasing water pump efficiency using artificial intelligence

What do artificial intelligence and machine learning have to do with Melbourne’s drinking water? And, how can these innovative technologies improve electricity efficiency in water distribution?

Increasing water pump efficiency through artificial intelligence

The Winneke treatment plant at Christmas Hills is one of the largest water treatment sites for Melbourne’s drinking water. Processing around 350 million litres of water every day, it uses a system of pumps to distribute water to millions of homes and businesses across Melbourne.

Operation of these pumps uses a significant amount of electricity, but Melbourne citizens use a variable amount of water every day. To ensure there is sufficient water at all times, the plant needs to match this amount.

In order to operate these pumps in the most efficient manner while still ensuring customers receive the required flow, Melbourne Water is using a custom artificial intelligence (AI) program developed using the Python programming language. The AI program mines historical pump data to ‘learn’ the most efficient pump configuration at any given time.

The AI system is powerful enough to consider a wide range of factors in its decision-making. The local control system also has rules built into its code to ensure the AI system will always operate within set parameters. This important, fail-safe feature ensures production will continue if the AI system fails.

It is anticipated that the AI system will reduce energy costs at Winneke by 20 percent per year. Plans are underway to roll out the AI platform for other Melbourne Water pump stations.

Last updated:
20 May 2019