Distance-Based Burst Detection Using Multiple Pressure Sensors in District Metering Areas

Abstract

It is a major challenge for water companies worldwide to quickly react to bursts so that water loss is reduced and service improved. This paper proposes a novel data-driven method for district metering areas (DMAs) with multiple pressure sensors that detects bursts in a timely manner. With the use of data correlation between multiple pressure sensors, the problem of burst detection is transformed into one of outlier detection. A distance-based algorithm is employed to distinguish outliers (i.e., burst-induced data) from normal data by evaluating the similarity (or dissimilarity) between data. Furthermore, approximate location information of bursts can be provided by comparing abnormality degrees, which represent different magnitudes of pressure drop at each sensor. The method was applied in a gravity-based DMA with four pressure sensors to evaluate its performance. The results show that the method can detect relatively large bursts and raises few false alarms. In addition, the method is able to identify the pressure sensor nearest to the burst event, thus giving some approximate location information.