Optimizing the parameters of hybrid active power filters through a comprehensive and dynamic multi-swarm gravitational search algorithm

Abstract

This paper introduces a dynamic comprehensive multi-swarm gravitational search algorithm with non-uniform mutation (cdGSA-2m) for optimizing hybrid active power filter (HAPF) parameters. The cdGSA-2m algorithm divides the population into two heterogeneous sub-populations with different learning mechanisms. The first sub-population focuses on exploitation with comprehensive learning, while the second sub-population emphasizes exploration with a dynamic multi-swarm strategy (DMS) with non-uniform mutation. Additionally, a non-linear adaptive parameter balances the search capabilities of the DMS sub-population. These mechanisms effectively control the exploration, exploitation, and diversity of the population. To analyse the effectiveness of implemented mechanisms in the cdGSA-2m, diversity and exploration–exploitation are discussed at length with a few experiments. An independent structural study of cdGSA-2m is also carried out to understand the effects of comprehensive learning, dynamic multi-swarm strategy, and non-uniform mutation mechanisms in the algorithm. The proposed cdGSA-2m is tested on IEEE-CEC 2017 benchmark problems, and the results of these problems are compared with twenty-two state-of-the-art algorithms in terms of mean error values and standard deviation of the fitness values. Based on the results, the algorithms are ranked against each other, along with statistical validation of the results. The proposed cdGSA-2 m also solves two case studies of the HAPF model and compares the results with existing state-of-the-art algorithms for the HAPF model. Experimental results demonstrate that the cdGSA-2m algorithm has better accuracy, convergence rate, search capability, and stability. The overall ranking of the algorithms suggests that cdGSA-2 m has outstanding potential to deal with difficult optimization problems.