Experiment of Different Mutation Strategies in Differential Evolution Algorithm Employed for Calibration of a Lumped Water Balance Model

Abstract

Differential evolution algorithm (DEA) is one featured kind of population based evolutionary algorithms. Although it is stated in previous studies that the algorithm is more sensitive to the crossover rate, it is tried out in this study whether DEA depends on mutation operation in which scaled differences of randomly chosen individuals existed in the population are used. Within this context, the presented study aims to carry out an empirical assessment regarding the comparison of DEA with the different mutation strategies for the calibration phase of a lumped water balance model. Both stable solution availabilities and convergence capabilities of DEA variants operated through five mutation approaches were performed on four parameterThorthwaite water balance model prepared for Gordes watershed. The findings derived from the model calibrations have indicated the usage of the fifth mutation strategy termed as current to best mutation strategy, which is more predominant in guaranteeing the achievement of stable solutions.