A performance comparison between genetic algorithms and differential evolution variants for water network optimisation

Abstract

This paper aims to compare the performance of three optimisation techniques in terms of optimising the design of water distribution systems (WDSs). These are the genetic algorithm (GA), the standard differential evolution algorithm (SDE) and the self-adaptive differential evolution (SADE) algorithm. In the SADE algorithm, instead of pre-specification, the control parameters of mutation weighting factor (F) and the crossover rate (CR) are encoded into the chromosome of the DE algorithm and hence are adapted by means of evolution. The SADE algorithm differs from the SDE algorithm in that the mutation and crossover parameter values of the SADE algorithm apply at the individual level rather than the generational level. Additionally, a convergence criterion is proposed for the SADE algorithm as the termination condition, thereby avoiding pre-specifying a fixed number of generations or computational budget to terminate the evolutionary algorithm. One benchmark WDS case study (Hanoi Problem network) is used to examine the search properties of the three optimisation approaches considered in this paper. The results obtained show that the proposed SADE outperformed the other two algorithms in terms of solution quality as well as efficiency. In addition, another advantage of the proposed SADE algorithm is that it relieves the effort required to fine-tune algorithm parameter values.