| Texto completo | |
| Autor(es): |
Borges, Livia Teresa Minami
;
Da Silva, Fabiano Borges
;
Soler, Edilaine Martins
;
Nepomuceno, Leonardo
Número total de Autores: 4
|
| Tipo de documento: | Artigo Científico |
| Fonte: | APPLIED SOFT COMPUTING; v. 186, p. 21-pg., 2025-11-06. |
| Resumo | |
This work proposes the Directed Genetic Algorithm (DGA) and demonstrates its global convergence through tools from the theory of stochastic dynamic systems. The DGA is primarily designed to solve the Economic Dispatch Problem with Valve-point loading and Prohibited Zones (EDPVPZ) representation, and the proposed Network-Constrained EDPVPZ (NC-EDPVPZ), which are multimodal, non-convex, non-differentiable, and discontinuous optimization problems in the field of power systems. Both problems are concerned with the calculation of the optimal generation dispatch, while taking into account physical and operational constraints related to the generation and transmission systems. Since these problems often have solutions concentrated at singular points of their objective functions, we introduce the innovative concept of a "pivot individual", which aims to explore these singular points and improves the algorithm's convergence rate and enables the discovery of optimal solutions by directing the search towards problem singularities. We also provide a formal proof that the cost associated with the best individual in the DGA population converges almost surely to the global minimum cost. We derive a computationally efficient linear model for the transmission network for the NC-EDPVPZ, where both losses and power flow limits are enforced for all individuals in the population. The fast computational feature of the proposed network model suggests its suitability for use within other metaheuristic approaches. We solve the NC-EDPVPZ for systems with up to 300 buses. To assess the performance of the DGA, we calculate the average convergence rate metric and conduct comparative analyses with the classical real-coded Genetic Algorithm (GA), Collaborative Optimization Differential Evolution (CODE), Particle Swarm Optimization (PSO), Covariance Matrix Adaptation Evolution Strategy (CMA-ES), Chaotic Adaptive Differential Evolution with Optional External Archive (CJADE), and Hybrid Sampling Evolution Strategy (HSES). Furthermore, the introduction of new genetic operators significantly improves the quality of optimal solutions compared to those obtained by traditional methods. The performance of the DGA is also compared using non-parametric statistical tests, involving the Wilcox signed-rank test, the Friedman test and the Nemenyi test. These results reveal that the DGA is competitive with the state-of-the-art methods for solving the EDPVPZ. (AU) | |
| Processo FAPESP: | 18/16568-0 - Um princípio de médias para equações diferenciais estocásticas |
| Beneficiário: | Fabiano Borges da Silva |
| Modalidade de apoio: | Auxílio à Pesquisa - Regular |