Performance comparison of metaheuristics for controller tuning to damp low-frequency oscillations in power systems

This article presents a comparative analysis of four categories of metaheuristic algorithms: the first inspired by human behavior, the second by swarm optimization, the third by the laws of physics, and the last by natural evolution. All of these algorithms aim to stabilize multi-machine power systems. To present alternatives to classical control approaches, this research focuses on optimizing control and damping of low-frequency electromechanical oscillations. Furthermore, it seeks to identify which algorithms are most suitable for solving the problem, considering the convergence criteria of the studied algorithms and the performance indices obtained from their response times in the proposed control approach. Based on the average performance indices across both test systems, the league championship algorithm (LCA) and the multiverse optimizer (MVO) emerged as the most effective approaches under the imposed conditions. In Test System 1, LCA demonstrated superior signal tracking capabilities, achieving the lowest mean values in IAE (0.2243×10^-3), ITAE (0.3697×10^-3), and ITSE (0.0713×10^-7), effectively outperforming the standard GA and PSO approaches. MVO showed competitive results, notably in ISE (0.0685×10^-7), indicating strong error minimization during transient states. In the more complex Test System 2, LCA confirmed its robustness by maintaining the lowest mean IAE (0.6095×10^-3) and ISE (0.2871×10^-7) among all methods, validating its suitability for larger-scale power system stability problems. Although there are variations among the methods in the models used, LCA and MVO achieved better results and greater regularity than the other metaheuristics. In terms of algorithm regularity, MVO is more recommended. For better performance conditions over time, LCA achieved the best results.