A better estimation of wave arrival time in water distribution networks using WAvelet kNEe (WANE)

Abstract

A transient pressure wave is a sudden pressure change that occurs in a short time, which can be induced by sudden changes in valve and pump operation, and pipe bursts in a Water Distribution Network (WDN). An accurate estimation of a transient wave arrival time is crucial because it facilitates pipe condition assessment, hydraulic model calibration, and accurate localization of pipe burst events. Due to the noisy and highly fluctuating nature of the pressure signals, estimating an accurate transient pressure wave arrival time is not a trivial task. Among many methodologies proposed for detecting abrupt pressure changes, Discrete Wavelet Transform (DWT) and Cumulative Sum (CUSUM) were the two most popular approaches. However, several limitations involved with these two approaches can easily lead to unsatisfactory results. Moreover, some of the existing methodologies were only tested on either a single pipeline, engineered events, or a small sample size of events, making these methodologies suitable and accurate only for a limited number of scenarios. Driven by these limitations, a novel approach is proposed to estimate the wave arrival time in water distribution networks (WDNs). The backbone of this approach is the integration of wavelet decomposition and a knee point detection algorithm, thus gaining the name WAvelet kNEe (WANE). Through a comparative study against the other methodologies using 90 recorded transient events detected in a real WDN, WANE is found to provide the best wave arrival time estimation, with a Root Mean Square Error (RMSE) of 0.4 s. Based on the result, our estimation error is at least 15 s lesser than the other methodologies. With an improved wave arrival time estimation, WANE has the potential to minimize the response time of repair crews, service disruption time, as well as the associated water losses due to a pipe break.