Yorkshire Water – Implementing real-time network event detection to detect potential supply interruptions

Project Description

Yorkshire Water has been working with Artesia to implement a water network event detection system that aims to detect and warn of potential supply interruptions. Along with all water companies, Yorkshire Water faces challenging regulatory targets in AMP7 to minimise the time customers go without water and is using Artesia’s machine learning AI system, eVader, to provide alarms when unusual flow or pressure events occur in the water network.

The eVader system was initially installed on about 180 DMAs in the Sheffield and Leeds areas as part of a 6-month head to head evaluation with a competitor system. After the successful 6-month evaluation, eVader was selected to enter the next phase of deployment to operate live on the whole DMA network.

The system, a cloud-based software as a service, receives data every 5 minutes and returns a result for each data point before the next data point for that site is received. This ensures that any anomalies, such as a burst or abnormal flow event is flagged to the control room as quickly as possible.

eVader uses a series of self-learning algorithms to check the incoming data, automatically gap fill where necessary, and apply robust statistical checks that determine what is normal for any site it is monitoring. Alarms can be generated when abnormal events are detected. The system dynamically adapts to network changes and takes into account seasonal variations in network performance.

The eVader system has been operating seamlessly for over 12 months, producing automated alarms such as the example above, in real-time. The rapid response of the alarms allows the control room to respond more quickly events such as bursts or changes, that could lead to a supply interruption. This minimises the length of time that customers could go without water. This will in turn improve service to customers and enable Yorkshire Water to minimise the risk of failing regulatory targets.

Click image for larger view

See more case studies