Leakage Localization with Differential Evolution: A Closer Look on Distance Metrics

In this paper the impact of different distance metrics as well as different sortings of the parameter space on the shape of the objective function for model-based leakage localization is investigated. Leakage localization is formulated as an optimization problem solved with a differential evolution algorithm. Distance metrics and sortings are evaluated through the convergence speed of the algorithm and the quality of the results in terms of a topological distance from the leak found by the algorithm to the real leak. The algorithm is tested on a hydraulic model of a real-world network and has shown that a Cuthill-McKee ordering of the search space together with correlation distance metric performs the best.