In November 2020, the Seine-Normandy Water Agency launched the call for innovation projects for water management, “Digital Transition and Circular Economy” to support local authorities, companies and associations involved in the protection of the resource and aquatic environments. Two selected projects, based on artificial intelligence, are currently being tested by the Val de Marne department.

Urbanization and changes in agricultural practices, such as the removal of hedges, increase the risk of flooding due to violent storms or heavy rainfall. It has become essential to take preventive and curative measures.

The Val de Marne, a highly urbanized department, located at the confluence of the Marne and Seine rivers, with a hydrographic network of about one hundred kilometers, is particularly exposed to the risks of flooding. To minimize these risks, it has set up the “Departmental Strategic Plan for Sanitation by 2030” and is developing two AI-based projects.

Self-monitoring and real-time management

The purpose of self-monitoring is to better control discharges, whether in exceptional circumstances (accidents, special weather events) or during work.

The first project, costing 632,000 euros, financed up to 80% by the Seine-Normandy Water Agency, aims to improve self-monitoring and real-time management of the sewerage network.

Sheila Aboulouard, deputy director in charge of operations at the Val-de-Marne’s environmental and sanitation services department, states:

“A self-monitoring order of June 21, 2015, amended on July 30, 2020, requires us to measure everything that enters the sewerage system and everything that is discharged into the natural environment, into the Seine and the Marne, in order to check whether our system is performing well.”

Seven hundred sensors were therefore installed on the departmental network to measure flows (water level, speed, flow rate). The data is then validated and sent to the State control services and the Water Agency. AI could replace human validation of data, which is very time-consuming, according to Sheila Aboulouard.

She explains:

“We provided AI with 10 years of data, from 2008 to 2018, to study the information transmitted by the sensors. This will allow it to check whether the data that will be taken next is consistent.”

She adds:

“Based on the data it has, it will be able to predict, for example, the flow of water in such and such a pipe… The artificial intelligence will predict the data at 2 hours, 6 hours and 12 hours so that we can act in real time, in case of a storm for example. If there is a risk of a spill, we will divert the effluent to limit the impact on the natural environment.”

This AI system will be deployed in nine catchments later this year and in all 25 catchments by 2024.

Predictive maintenance

The second project involves predictive maintenance for the 113 pumping stations in the Val de Marne. Sensors will soon be installed on a pilot station where no maintenance will be carried out for six months. When a breakdown occurs, all the parameters responsible for this failure will be recorded and analyzed. It will then be possible to prevent incidents on the network and thus minimize the risk of flooding.

This project, which costs 420,000 euros, also benefits from an 80% subsidy from the AESN.

Translated from Le Val de Marne mise sur l’intelligence artificielle pour améliorer son réseau d’assainissement