Calibration Of Water Distribution Models By Genetic Algorithms

The paper discuses the calibration of water distribution models by means of genetic algorithms as the optimization method. Calibration of hydraulic models is a procedure of determining individual unknown parameters of a hydraulic model, which minimizes the differences between the measurements performed on a real water distribution system and the results of the hydraulic model. The applied calibration approach is consisting of the “macro” and “micro” calibration level. The “macro” calibration allows the hydraulic model to become a "rough" approximation of the real system, by ensuring that the system variables are in reasonable agreement with collected measurements. At this stage all major possibilities of model discrepancies should be uncovered and resolved. Afterwards the “micro” calibration procedure is applied to identify the unknown model parameter values by using the optimization method of genetic algorithms. The optimization problem is structured so as to search for values of optimization variables or unknowns, which minimize the objective function while in the same time fulfilling all constraints. The objective of the calibration optimization problem is expressed in a form, which allows minimization of the differences between measurements and model predictions. The aforementioned modeling and calibration approaches were applied to a real water distribution system of a part of the Slovenian capital Ljubljana. It can be concluded that the calibration approach was successfully applied and that genetic algorithms have proven its robustness in identifying near optimal solutions of the calibration problem.