A linearization approach for improving the computational efficiency of water distribution system ranking-based optimization algorithms

This paper presents a linearization approach to be used within the framework of water distribution system optimization problems where the searching algorithm is guided by the ranking of the possible solutions characterized by a system performance indicator such as the resilience index. Typically, within these optimization problems the value of the system performance indicator associated to each solution is quantified by performing a hydraulic simulation. Indeed, even if hydraulic simulators of water distribution systems enable us to accurately calculate pipe discharges and nodal heads in a short time, when optimization problems are applied to real networks, the computational times tend to increase. Thus, given that within the framework of the previously mentioned optimization problems, the aim of hydraulic simulations is to quantify the selected indicator in order to rank the different hypothesized solutions, we evaluate the possibility of replacing the correct hydraulic simulator with a simpler simulator which relies on linearized energy balance equations to improve the computational efficiency. The results obtained in addressing two different problems applied to real water distribution systems, namely, identification of the optimal placement of flow meters in a district meter area and multi-objective design of a water distribution system, show that a linearized hydraulic simulator although not capable of correctly describing the hydraulic functioning of the network, nonetheless enables each solution it finds to be assigned a value of the performance indicator whose ranking closely approximates the ranking of the corresponding solution that would be obtained if the full hydraulic simulator was used. This ensures significant reductions in computational times.