Reliability-based design of water distribution networks using self-adaptive differential evolution algorithm

This paper presents the application of Self-Adaptive Differential Evolution (SADE) algorithm for reliability-based design of Water Distribution Networks (WDNs). The algorithm is first applied and tested for deterministic design of various benchmark WDNs. Reliability-based design is then performed on the benchmark WDNs as well as real WDN in India by employing the SADE algorithm considering the uncertainty in nodal water demands. Monte Carlo simulation approach is used to incorporate the uncertainty in nodal demands; and pressure driven analysis is adopted for simulating the distribution of flows and pressures in pipe networks with the help of EPANET toolkit. A thorough comparison is made between the performance of SADE and DE methods. The results illustrate that SADE performs much better than DE since SADE converges faster with a higher success rate, making it a more preferable option for WDN design problem. The results of the study also illustrate that the optimal cost increases with increasing uncertainty level, but the percentage increase in cost is problem specific, which requires a robust tool for handling such complexities. The consistent results obtained by the application of SADE algorithm makes it a preferable option for reliability-based design of large WDNs including real world problems.