CCWI2017: F119 'WDNs Calibration Using k-means Algorithm for Pipes Clustering and a Hybrid Model for Optimization'

The modelling of Water Distribution Networks (WDNs) is an important issue for an efficient operation of water systems. The hydraulic simulations are used to establish the optimal conditions for pumps and valves, helping stakeholders to manage the systems. A reliable model should guarantee the minimal uncertain becoming the simulations the most real as possible. The calibration processes is used to reduce the uncertainties associated to pipe roughness and nodal demand, pipe status or leakage flow. Due to the high degrees of freedom surrounding this process, the improvement of available information of the hydraulic state of the network and the reduction of parameters to be calibrated can conduct for more reliable calibrated model. In this sense, this work present a hybrid calibration method, consisting in three stages defined as grouping pipes with k- means algorithms, pressure estimation with artificial neural networks (ANN) and an optimization process using particle swarm optimization (PSO) algorithms. A comparison between the classical approaches for roughness calibration is presented, reinforcing the improvement of the calibration process through the uncertain reduction.