An improved electromagnetism-like mechanism algorithm for the optimization of maximum power point tracking / Tan Jian Ding

Tan, Jian Ding (2017) An improved electromagnetism-like mechanism algorithm for the optimization of maximum power point tracking / Tan Jian Ding. PhD thesis, University of Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (1755Kb)
Preview	PDF (Thesis PhD) Download (2953Kb) | Preview

Abstract

The Electromagnetism-Like Mechanism algorithm (EM) is a meta-heuristic algorithm designed to search for global optimum solutions using bounded variables. The search mechanism of EM mimics the attraction and repulsion behaviours in the electromagnetism theory. Despite its notable performance in solving various types of optimization problems so far, literature study shows that in general, EM is good at solutions exploration but shows insufficiency in its solutions exploitation ability. Based on this motivation, this study aimed to improve the EM by enhancing this algorithm with stronger exploitation mechanisms. This research can generally be divided into several phases. The first phase of the research was on the investigation of the relationship between the search step size and the convergence performance. The conventional EM was tested to search under two different extremes of step sizes separately, marked as EM with Large Search Steps (EMLSS) and EM with Small Search Step (EMSSS) respectively. Experiments on ten test functions showed that the EMSSS performed much detailed searches in all dimensions and yielded outcome with higher accuracies. The trade-off, however, was that the convergence processes were comparatively slower than the EMLSS. The second phase of the research focused on enhancing the EM. Two major breakthroughs were achieved. The first successful modification was recorded by introducing a Split, Probe and Compare (SPC) feature into the EM (SPC-EM). The SPCEM applied a dynamic strategy to regulate the search steps during the local search. The search scheme began with relatively bigger steps. The algorithm then systematically tuned the step sizes based on a specially designed nonlinear equation. This ensured accuracies of the final solutions returned, in the meanwhile not slowing down the whole convergence process by probing around too finely at the beginning of the search. The modified algorithm was tested out in the established test suite. The results indicated that SPC-EM outperformed the conventional EM and other algorithms in the benchmarking.

Actions (For repository staff only : Login required)