The calibration process of water distribution system models allows for accurate and reliable hydraulic analysis results. Thus, calibration is of utmost importance if adequate operation and maintenance model-based procedures are sought. However, in emerging economies, there is a series of factors that make it more difficult to construct accurate models, including very poor information management, unusually high leakages and the presence of a large number of illegal connections. While some of the model variables are assumed to be known under normal circumstances, these factors make it necessary to consider them for calibration as well. This paper presents a calibration methodology flexible enough to address such problems allowing the calibration of pipe diameter, roughness and minor losses, and nodal demands and leakages. A genetic algorithm was implemented as well as a constraint programming algorithm that makes direct use of hydraulic criteria to advance in the solution space. The methodology was tested on a real system in Colombia with a satisfactory outcome. The use of these techniques results in major reduction of calculation time and similar or superior results in comparison to manual methods.
Keywords: constraint programming, genetic algorithms, water distribution system calibration