Improved performance of PSO with self-adaptive parameters for computing the optimal design of Water Supply Systems

This paper deals with a new variant of Particle Swarm Optimization (PSO) in which no a priori parameter tuning is necessary. PSO, as an efficient and powerful problem-solving technique, has been widely used, but, as in other Evolutionary Algorithms (EA), appropriate adjustment of its parameters is cumbersome and usually requires a lot of time, effort and luck. Thus, a self-adaptive framework is proposed to improve the robustness of the PSO. In this paper, within a framework that also includes other variants previously introduced by the authors, the algorithm’s parameters are co-evolved with the particles. Its performance results show that the use of this self-adaptive feature averages out the performance of standard PSO and other EA applied to the same problems, namely the design of Water Supply Systems (WSS), while avoiding the process of localizing and fine-tuning suitable parameters values, when using two benchmarking problems presented in the literature, namely the Hanoi Water Supply System and the New York Tunnel Water Supply System. The results provided in the case of a real-world problem demonstrate the scalability of the proposed variant to the realistic water distribution design problems, which are much larger.