Implement AI in water treatment

New technologies are democratizing in all areas and water treatment also benefits. The amount of data collected from the water treatment network is only growing. So much so that artificial intelligence (AI) is starting to be used to improve the efficiency of water distribution and sanitation.


The Helios group operates public water treatment infrastructure in Quebec. The company’s mandate covers the maintenance and upkeep of the network as well as the production of drinking water and wastewater treatment.

Quentin Deroo, its Technology Director, tells us that “Helios seeks to stand out from the competition and increase the added value of our operations thanks to artificial intelligence. It should also allow us to make our operations more efficient ”.

One of the keys to the integration of artificial intelligence is the massive collection of data. To this end, Hélios has developed an application, Colibri, allowing operators to record data during their working day. For each wastewater station, the operator will come and scan the QR code placed on each machine to record his data. These bring a better knowledge of the network and they will make it possible to train algorithms in AI.


One of the concrete applications of AI would be the prediction of drinking water consumption. Seasonal data such as time of day, time of year, location or even weather are known to influence drinking water consumption. Today, this prediction is made by the operator and its finesse depends greatly on his experience.

An algorithm to predict this consumption by integrating historical consumption data with seasonal data is currently in development. It would make it possible to program the pumping of drinking water in a finer way and thus reduce restriction measures. This is an important issue in municipalities that experience water stress every summer.

It is also possible, thanks to this tool, to detect leaks on the network. They currently represent 20% of the water introduced into them. Their detection therefore has both an economic and an environmental stake. After making a map of the consumption on the network, the system monitors the data and generates alerts in the event of anomalies, even minor. Hydroscan, a Belgian company, thus reduced leaks on an urban network by 5% in 2017, saving 5 million liters.


Wastewater treatment is another axis for the use of artificial intelligence. 70% of wastewater treatment plants are aerated ponds. Coagulating chemicals are used to agglomerate and then remove suspended matter. Then oxygen is pumped into the pond so that carbonaceous pollution is assimilated by bacteria.

“An artificial intelligence algorithm would make it possible to have a fine prediction on the quantity of chemicals and oxygen needed to purify the water. Currently, the oxygenation of the pond is the first item of expenditure for wastewater treatment and its control is carried out manually by an operator. ” – Quentin Deroo – Technology Director – The Hélios group

Rainfall influences the amount of suspended matter in the water. Thus, data on water quality or weather are necessary for the construction of this algorithm. In addition, ponds have a different bacterial fauna. It is therefore necessary to come and train the algorithm with data specific to each of them.

To conclude, Mr. Deroo sees AI as a real opportunity to optimize operations. This by reducing energy consumption, the use of chemicals for suspended solids and by optimizing operations to reduce the environmental footprint of activities. An application as concrete as it is necessary, and which should evolve further with the addition of new data in the years to come.