Iteratively tuning the regularisation parameter in an inverse method for localising leaks in water distribution networks

Abstract

[EN] Novel methods for detecting and localising leaks in water distribution networks are being increasingly investigated as water utilities face unprecedented financial and environmental challenges in reducing water losses. A promising method includes the solution of the regularised inverse problem that minimises the difference between simulated and measured data in addition to the regularisation term. However, the results of leak localisation are sensitive to the choice of the regularisation parameter. In this paper, we propose and investigate a method for iterative tuning of the regularisation parameter to improve the leak localisation performance in case of multiple time steps measurements. The numerical investigation utilises a benchmarking network model, and different metrics are applied to evaluate the results of the leak localisation performance, such as the performance and the distance metric.