Leak Estimation in Water Distribution Systems by Statistical Analysis of Flow Readings

Abstract

A new method is presented for detecting the magnitude of leaks in small residential service zones (under 1,000 homes) of a drinking water distribution system. It is assumed that continuous measurements of flow rates through the main supply line into a residential service zone are available during periods of low use. The sample mean and variance from the set of measured flow rates are computed as the set is truncated progressively from below. Trajectories of the sample statistics and their derivatives are plotted versus the level of data truncation. In the presence of leaks, these trajectories diverge from their expected theoretical path when plotted on a standardized graph derived from a mixed truncated normal distribution. The point of departure on the standardized graph indicates where the truncation threshold matches the maximum rate of network leakage. A performance limit for the proposed method is derived to account for network size and time averaging. Several examples, based on observed and simulated pipe flows, demonstrate application of the leak detection method.