Anomaly Event Localization in Water Distribution Networks via Integrated Hydraulic-Acoustics Analysis

During the daily operations of water distribution networks, hidden anomaly events, which include, but not limited to, leaks and illegal water usage can occur unexpectedly and remain undetected over long periods. Early identification of the likely area coverage, to the smallest possible yet accurate extent, of the hidden anomalies, especially pipeline leaks, continues to be operationally challenging to water utilities. To address the challenge, this study develops a novel integrated hydraulic-acoustics analysis framework to localize the area coverage of detected anomalies in near real-time for pinpointing their likely locations within the network. Coupled with our prior anomaly detection in near real-time, our anomaly localization approach comprises of two key systematic steps, namely: (1) scoping the likely pipeline distance coverage from individual pressure-acoustic sensor stations, which detected anomaly occurrences with high confidence level, via a proposed autocorrelation method that estimates the time-delay between near-periodic power peaks in waveform profiles, as representative of continuous leak signals, to subsequently approximate the pipeline distance coverage from respective stations; and (2) performing the optimization runs for hydraulic model calibration with candidate locations (i.e. junctions) along the localized pipeline coverage from the previous step to identify the likely anomaly hotspots with defined confidence scores. In collaboration with PUB, Singapore’s National Water Agency, the proposed approach has been tested in the water supply zones with more than 1100km of underground water pipelines having 82 inline pressure-hydrophone sensors for 1 Jan 2020 to 31 Aug 2020. The reported leaks have been localized within 500m pipeline distance or less, with at least 71% confidence score in emulated near real-time.