Calibrating Water Distribution Model Via Genetic Algorithms

摘要: Computer models have been built for the simulation of water distribution systems since mid-1960s. However, a model needs to be calibrated before it can used analysis and operational study real system. Model calibration is vitally important, but time consuming task. Over last two decades, several approaches using optimization techniques proposed calibrations. Although most methods make agree with field observations, few are able achieve good level in terms determining correct parameters (pipe roughness coefficients, junction demands valve settings). The previously developed appear lacking versatility users accurately specify task given data This paper proposes comprehensive flexible framework calibrating hydraulic network model. Calibration tasks specified wat er system according availability application requirements. It allows user (1) flexibly choose any combination such as pipe roughness, demand link (pipes, valves pumps) status, (2) easily aggregate reduce problem dimension expeditious calculation, (3) consistently boundary conditions loadings that corresponding collection. A then defined an implicit nonlinear problem, which solved by employing powerful genetic algorithm (GA), generic search paradigm based on principles natural evolution biological reproduction. solutions obtained minimizing discrepancy between predicted observed values pressures flows. With this methodology, modeler fully assisted during process, thus possible high confidence. As result, conducting management. Example presented demonstrate efficacy robustness genetic-based methodology