Fewer failures, more water. How can AI improve the operation of waterworks?

We lose as much as 30% of water due to leaks in the water supply and sewage system. Artificial intelligence can significantly reduce these losses, and this is just the beginning of the benefits that AI brings to the water supply industry. This has been confirmed by researchers from the Łukasiewicz – Poznań Institute of Technology.

Poland has some of the lowest renewable drinking water resources in Europe. It is also among the countries where these resources are shrinking the fastest. This situation is influenced, among other things, by climate change, which brings heat waves and droughts. That is why we should manage water in a sustainable manner.

This is not always possible, because Polish water supply networks, especially municipal ones, are often old and in poor technical condition. As the Supreme Audit Office showed a few years ago, in more than half of water and sewage companies, losses amounted to over 30 per cent of the volume of water pumped into the network.

Is it possible to minimise these losses without costly investments? Tools based on artificial intelligence come to the rescue here.

‘Properly trained and fed with data, artificial intelligence can help us recognise a pre-failure condition early enough, predict when a failure is likely to occur, and suggest corrective actions so that the consequences for residents and the water and sewage company are as minor as possible,’ explains Włodzimierz Woźniak from the Łukasiewicz – Poznań Institute of Technology.

When will there be a failure?

This question can be answered thanks to AI, and with this knowledge, failures can be prevented. One area of application for artificial intelligence is predictive maintenance. This involves an AI model analysing information from various sensors, cameras, monitoring devices, as well as environmental and historical data. By analysing such a huge collection, it is able to detect incidents and relationships between events that humans cannot see. This allows it to indicate where a failure has already occurred or may occur (and at what time).

Small leaks from water supply networks often go unnoticed. Properly trained artificial intelligence can use satellite images to detect changes in the terrain that suggest that the ground near the pipeline is more moist. This suggests that a pipe has sprung a leak and is leaking in that location.

AI can also detect leaks by analysing the chemical composition of sewage, weather and geological data, and monitoring information.

What is the demand for water?

There are periods when the demand for water increases, and when the water supply network is unable to meet it, the water pressure drops and does not reach all users. Artificial intelligence allows you to respond to their needs on an ongoing basis.

Remote water meter readings are helpful here, as they provide the water and sewage company with real-time information on water consumption. By analysing this data, AI allows the power of the pumps to be controlled appropriately, thus ensuring the correct water pressure in the network.

What is the best way to treat wastewater?

Artificial intelligence can support and control water treatment and wastewater treatment processes. Let’s illustrate this with the example of wastewater aeration. In order to select the right parameters for this process, a lot of data needs to be analysed (e.g. sewage composition, bacterial activity, pump efficiency and weather conditions). An AI model is perfect for this, allowing you to achieve the best results while saving energy and reducing pump consumption.

Less water loss

‘The introduction of solutions using automation and artificial intelligence undoubtedly entails costs for many companies. However, these investments bring savings, ensure continuity of water supply and reduce water loss,’ concludes Włodzimierz Woźniak. He gives the example of the Water Supply and Sewage Company in Tarnowskie Góry, which, after introducing a water management support system, reduced water losses from around 30% to 10-11%.