Estimating Water Losses in Water Distribution Networks Using a Hybrid of GA and Neuro-Fuzzy Models

2 Abstract: Modeling Leakage rate in water supply networks is important due to some problems namely: quality of service, the cost of system expansion and wasting energy resources. Monitoring the pressure in certain parts of network and then finding a relation between pressure changes and leakage rate is a quick way for detecting leakage. This relation is nonlinear and complex for modeling and there exist some problems related to mathematically modeling of it by hydraulic fundamentals. Thus, regarding to the abilities of soft-computing methods for modeling nonlinear processes, it seems they are very useful to apply in this field. Consequently, in this study, a procedure for developing fuzzy models is introduced that employs Genetic Algorithm (GA) and Adaptive Neuro-Fuzzy Inference System (ANFIS) for optimizing them in terms of accuracy and complexity. Two water supply networks located in Kerman Province (Iran) are considered as case studies in order to illustrate the efficiency of our modeling procedure. Simulation results apparently show, applying proposed method results in achieving compact and accurate fuzzy models for estimating leakage rate.