Auxiliary population-assisted differential evolution for multi-area economic dispatch considering valve point effects

Abstract

Multi-area economic dispatch (MAED) is an indispensable task in the power system's operation. Nevertheless, the valve point effects of generating units make the problem highly nonlinear and non-convex. In this paper, a problem solver called AP-LSHADE-RSP is proposed. AP-LSHADE-RSP builds upon LSHADE-RSP (LSHADE with rank-based selective pressure strategy) by integrating three strategies to overcome its early convergence issue. (1) An auxiliary population (AP) is introduced to store fitter trial individuals to retain high-quality information rather than discard them outright. (2) An AP-assisted mutation is designed to generate mutant individuals from the AP or the main population to increase the population richness. (3) An improved crossover is presented to combine information from both parents and mutants to mitigate information homogenization. AP-LSHADE-RSP is validated on CEC2018 benchmarks and three MAED cases and many more. It outperforms other algorithms significantly based on Wilcoxon rank-sum test and finally ranks first based on Friedman test on CEC2018 benchmarks. Besides, compared with the suboptimal algorithm, AP-LSHADE-RSP reduces the average fuel cost by 8.25–130.51 $/h, its standard deviation is only 15.29–22.48 % of the competitor, and its convergence speed is improved by 32–59 %. Furthermore, it reduces the CPU time by 6.4–58.3 % compared with LSHADE-RSP.