AI for Water Network Intelligence & Leakage Detection

AI water-network intelligence uses statistical and machine-learning methods to interpret information from sensors, meters, asset records and network systems. It can help utility teams identify unusual flow or pressure patterns, forecast demand and prioritise locations for investigation. Engineers review the findings alongside hydraulic information and field evidence. 

Yes. The course compares threshold alarms, statistical baselines and machine-learning methods using representative district data. Participants use measures such as precision, recall and false alerts to judge how often a model identifies relevant events, which events it misses and how much investigation work its alerts generate. 

Familiarity with water-network operations, applied analytics or basic time-series concepts is helpful. The course focuses on interpreting model outputs and evaluating their relevance to utility decisions. Confirm the tools and any Python requirements with Excentra before registering. 

Anomaly detection flags an unusual pattern, such as a rise in night flow or an unexpected change in pressure. Demand changes, sensor faults and operating events can also produce unusual signals. A suspected leak is assessed against other evidence, such as pressure records, operating logs and, where needed, field observations. 

Yes. The course examines how digital twins combine network models and measurements to explore operating scenarios. Participants consider how model assumptions, sensor coverage and uncertainty affect the conclusions a utility can draw from a simulated change in demand or pressure. 

Participants learn to look for changes in sensor behaviour, seasonal demand and false alert rates that may affect model performance. They consider how to use quality checks and performance measures to decide when a model needs investigation or retraining. They also define when operators should return to established procedures. 

Yes. Excentra can discuss scenarios relevant to your network, such as recurring night-flow alerts, incomplete pressure records or demand variability across districts. Synthetic or approved data can be used to reflect the team’s operating conditions. Excentra Discuss your utility’s requirements with Excentra. 

Participants develop a data-readiness checklist, a model-performance scorecard and a workflow for prioritising leakage alerts. The checklist records which measurements are available and where gaps affect confidence. The scorecard and workflow set out how to assess model outputs, route alerts for review and decide which findings warrant field investigation.