IoT Leak Detection Reported to Cut Non-Revenue Water Losses by Up to 30 Percent

Utilities deploying IoT monitoring across distribution networks report reductions in non-revenue water losses of up to 30 percent. Sensors distributed along pipelines transmit continuous pressure, flow and quality data, letting operators pinpoint leaks within minutes rather than days. A Yorkshire Water deployment is cited as identifying more than 1,000 leaks in a single pilot phase.

Non-revenue water is the term for water a utility treats and pumps but never bills, and in many networks it runs between 20 and 40 percent of everything put into the system. That figure is why the business case here is unusual: the saving is not administrative efficiency but a physical commodity already paid for. Every litre not lost has already had energy and treatment chemicals spent on it.

The detection method is worth understanding because it explains why continuous data matters. A leak rarely announces itself; it shows up as a small persistent flow during hours when consumption should be near zero, or as a pressure gradient that does not match the expected profile. Neither pattern is visible in a meter read once a quarter. Machine learning approaches — random forests, k-nearest neighbours and decision trees are mentioned in the same material — are applied to separate those signatures from ordinary demand variation, which is a classification problem rather than a threshold one.