Calibration in Water Distribution Networks with Pressure-Driven Analysis

Abstract

Water distribution networks (WDN) connect consumers to the water sources, and its goal is to fulfil water demand. However, it is a well-known fact that WDN have losses and an important part of them occur at pipe level. Despite all the research efforts focused on this subject, the identification of leaky pipes is still a major challenge. EPANET is frequently used to simulate WDN’ models, using a link-node formulation, similar to a graph, where the water demands are assigned to the nodes. A linearized system of equations (mass and energy conservation laws) is iteratively solved by a Newton-Raphson algorithm. The EPANET is demand-driven, since it assumes the water pressure is always enough to satisfy the demands. However, on real WDN, states of insufficient pressure also occur. Besides that, the demand-driven approach is not suitable for pipe leakage simulation, which depends on the pressure. WaterNetGen — an EPANET extension— allows both demand and pressure driven simulations, including pipes’ leakage modelling. However, the leakage parameters (bursts and background leakage coefficients and exponents) must be set manually by an expert — manual calibration — for the whole network or for each pipe. This work proposes a calibration methodology to estimate the pipe background leakage parameters. The approach is tested on a set of synthetic models, generated by WaterNetGen, and then applied to a real WDN to assess its performance on real world conditions.