Reducing customer minutes lost by anomaly detection?

An method which compares measured and predicted water demands to detect anomalies, was developed and tested on three data sets of water demand of three years in which and 25 pipe bursts were reported. The method proved to be able to detect bursts where the water loss exceeds 30% of the average water demand in the area. By simultaneously running the method in adjacent supply areas, and combining the monitoring results the number of false alarms could be reduced. Further analysis of the reported bursts, showed that most burst (22 of 25) were isolated within 2 hours after occurrence. The anomaly detection method could not have reduced the number of Customer Minutes Lost (CML) of those bursts. The water loss and pressure drop of the other bursts was limited and caused no CML. The detection method was able to detect the bursts, but did not reduce the CML