Abstract
The Dung Beetle Optimization (DBO) algorithm is a relatively recent metaheuristic known for its simplicity, versatility, and low parameter dependence, making it a valuable tool for solving complex optimization problems. Despite its potential, DBO suffers from limitations such as slow convergence and premature stagnation in local optima. To address these critical issues, this paper introduces a novel enhanced variant named Elite Bernoulli-based Mutated Dung Beetle Optimizer with Local Escaping Operator (EBMLO-DBO), specifically designed to improve the convergence speed, search capability, and robustness of the original DBO algorithm. The motivation for this enhancement stems from DBO's limited performance in high-dimensional and non-convex problems, where it often fails to maintain an effective balance between exploration and exploitation. The novelty of the proposed EBMLO-DBO lies in the integration of four key strategies tailored to overcome these weaknesses: (i) Bernoulli map-based initialization to enhance population diversity and ensure a better global search foundation; (ii) Morlet Wavelet mutation to introduce adaptive local refinements and help the algorithm escape local optima; (iii) elite guidance to accelerate convergence by directing the population toward high-quality regions; and (iv) a local escaping operator (LEO) to dynamically refine the search process and strengthen exploitation without sacrificing exploration. The performance of EBMLO-DBO is rigorously validated using the CEC2017 and CEC2022 benchmark suites, where it achieves Friedman ranks of 1.83 and 2.7 respectively, consistently surpassing eleven state-of-the-art algorithms including PSO, HHO, WOA, and advanced methods like CMAES and IMODE. In benchmark function optimization, EBMLO-DBO demonstrates superior performance by achieving first rank in 50% of CEC2022 functions and obtaining the lowest average fitness values in 18 out of 29 CEC2017 functions. For photovoltaic parameter estimation applications, EBMLO-DBO exhibits exceptional accuracy with RMSE values of 9.8602E-4 for single diode models, 9.81307E-4 for double diode models, and 2.32066E-3 for PV module models, achieving top performance ranks of 1.45, 1.42, and 1.74, respectively. Statistical analysis using Wilcoxon signed-rank test at significance level [Formula: see text] confirms the significant superiority of EBMLO-DBO over all compared algorithms, thereby validating the effectiveness and reliability of the proposed enhancements. Overall, the results state that EBMLO-DBO offers a significantly improved search performance and solution quality compared to the original DBO and related methods, thereby justifying the necessity and effectiveness of the proposed enhancements.