Comparison of Three Metaheuristics for Leak Diagnosis in Water Distribution Networks

This article presents a comparison between three different metaheuristic methods: the genetic algorithm (GA), the simulated annealing (SA), and the particle swarm optimization (PSO) in an experimental setup for leak diagnosis in a water distribution network. First a detection stage estimates the value of the flow-rate residual between the input and the output of the water distribution network until it exceeds a preestablished threshold value. The estimated value of the leak is analyzed during a previous time window and when its value is stable around a constant value a detection alert is emitted. Then, an optimization algorithm is implemented. A cost function with two search variables (location and magnitude of the leak) is then defined. The optimization of the cost function is performed using the three metaheuristic methods. Finally, results using experimental data are presented: the computing time of each metaheuristic method, the estimated leaked outflow and the success/failure rate for each method.