SmartWaterConnect 5.0

Many water leaks make for a huge loss

Flanders is a region where the water stress level is high. At the same time water companies estimate they lose 60 million cubic meter of water a year due to leaks in the network, representing almost a fifth of the annual water consumption. More than half of the water loss is caused by small or historic leaks in the network. Another third could be attributed to leaks in the connection between the consumers’ water meters and the network pipes.

Small or historic leaks and leaks at the connection points are difficult to detect, and therefore remain largely unsolved nowadays. To improve this situation, the Flemish government set forth an ambitious target to reduce this water loss to half of what is internationally accepted as unavoidable loss in a water distribution network.

The missing link: connecting the data dots

The SmartWaterConnect 5.0 project aims to optimize the localization of leaks, both small and historic leaks in the network, as well as leaks in the connection between the network and the consumers’ smart meters. The goal is to reduce the loss, often referred to as non-revenue water, for De Watergroep down to 8%. De Watergroep is the largest water company in Flanders, supplying drinking water for three million customers.

To do so, the project will build the missing link in water leak localization. We will bring together data from various sources and develop robust algorithms using hybrid AI for a fine-grained leak localization. The data will stem from a fine-grained deployment of fixed sensors (i.e., the smart meters) at the consumer’s location, together with dynamically positioned mobile sensors to pinpoint leaks in the public network.

Research goals

SWC’s project partners will tackle the challenge from two angles:

• A first effort will start from a fine-grained model of the public water network, a similar model used in the imec.icon predecessor SmartWaterGrid. To detect and localize leaks in that network, the project partners will design new hybrid AI models and adapt existing learning methodologies for hybrid AI.
• A second effort is meant to localize leaks between the consumers’ smart meters and the public water network. To do so, the researchers will design innovative hybrid AI models that take into account smart meter sensor data, including acoustic noise level, water pressure, and consumption data.

After researching both angles, the project will combine the results into a final proof of concept. It is expected that the project’s results will lead to a more efficient and economical strategy for placing sensors on the network, and from there to the envisaged improvement in the detection of leaks.

“SmartWaterConnect will enable the missing link in leak detection by bringing together data from different sources, including smart meters and other sensors, to target a considerably lower water loss in the public water networks.”