Multivariate calibration of water supply networks using multiobjective genetic algorithm

This paper has aimed to propose a robust computational model with a view to multivariate calibration of hydraulic models of water distribution systems, enabling the identification of possible irregularities such as leaks, obstructions in the pipes, dead or strangled valves and incompatibilities in the modeling for the data real. Using a multiobjective genetic algorithm calibration process adjusts for differences of state variables of the network giving information consistent with their respective real networks. The proposed model allows use up to seven variables: the roughness, the demand, the headloss, the topographic, leaks, valves and diameters, simultaneously or any combination thereof. The experimental application was carried out at the Laboratory for Energy Efficiency and Water Sanitation Universidade Federal da Paraiba (UFPB) through its Pilot System for Water Distribution. The results showed a good convergence with respect to processing time and the approximation of the measured and calculated values, as well as enables the identification of problems by analyzing the parameters resulting from calibration, thus providing support for a more accurate rehabilitation.