On the assessment of meta-heuristic algorithms for automatic voltage regulator system controller design: a standardization process

Meta-heuristic algorithms (MHAs) have gained popularity in recent years due to their successful results in solving a wide variety of scientific problems. They offer ease of use, fast implementation, and effective convergence toward the optimal solution. Although MHAs have been extensively tested in solving well-known mathematical benchmark problems with one or more dimensions as well as civil and mechanical engineering problems in their initial demonstrations, controller design problems are not typically considered. Furthermore, the literature lacks a standardized optimization process for controller design problems using MHAs. Due to variations in iteration numbers, population sizes, number of trials, objective functions, and insufficient analysis presented in research papers, it becomes challenging to compare and evaluate the controller design performance of MHAs in a successful and fair manner. This work aims to establish a standardized approach for evaluating the performance of MHAs in controller design by proposing a consistent function evaluation metric. To achieve this goal, we present the most comprehensive and comparative study of MHAs’ performance in controller design conducted to date. In this paper, we utilize two commonly used objective functions in controller design: Zwe Lee Gaing and Integral Time Absolute Error. Additionally, we employ a total of twenty algorithms, consisting of ten classical algorithms and ten recently popular algorithms. We evaluate the performance of these algorithms on the “automatic voltage regulation” electric power system problem, which serves as a widely used benchmark for meta-heuristically optimized controllers. We consider three different controllers with three (PID), five (FOPID), and seven (FOPIDD) parameters. The performance results of the selected algorithms are thoroughly discussed, considering various analysis techniques such as box plot analysis, convergence curves, and transient response performances, as well as statistical tests like Wilcoxon and Friedman tests. As a result, symbiotic organisms search, teaching–learning based optimization, chaos game optimization, supply–demand based optimization, and jellyfish search algorithms generally emerge as the best-performing algorithms across all optimization processes for the three types of controllers. Researchers interested in conducting further analysis and comparing the improved algorithms can access all the models,parameters, and codes used in this study f