Comparison of Heuristic and Metaheuristic Evolutionary Algorithms on Optimal Design of Water Distribution Networks

Complexity in the design of water distribution networks (WDNs) is primarily due to the discrete nature of available pipe sizes and the nonlinear relationship between pipe discharge and head loss through it. Several evolutionary algorithms have been suggested by researchers for the optimal design of WDNs in the last two-decades considering different complexities. These have been tested and evaluated on many benchmark networks. Most of these algorithms are inspired by natural phenomenon of selection and called metaheuristic evolutionary algorithms. Metaheuristic algorithms are parameter-based, and thus fixing their values for a particular problem is always a challenge. To overcome this issue, heuristic evolutionary algorithms like Jaya, Rao-I, and Rao-II have been suggested by Rao (2016, 2019) which are parameter-free algorithms. How best these three algorithms are applicable to design WDNs of various complexities which have been explored through their applications on various bench mark problems. Further, the efficiency and efficacy of these algorithms are compared with a metaheuristic algorithm ‘particle swarm optimization’ (PSO) which is based on the phenomenon of bird flocking and require few parameters to be defined. PSO is successfully used by many researchers and found to be better in comparison with other metaheuristic algorithms. In this study, Rao-I algorithm is found to be faster in providing optimal solution as compared to Jaya and Rao-II. In comparison with PSO, Rao-I is observed to provide similar solution in more or less same number of evolutions. Even though these heuristic evolutionary algorithms do not mimic any natural phenomenon, they are observed to provide better optimal solutions for WDNs.