Potential energy-guided optimization algorithm: a novel physics-based metaheuristic for global optimization

This paper proposes a novel Potential Energy-Guided Optimization Algorithm (PEOA), a physics-inspired metaheuristic rooted in the principle of minimum potential energy that enables natural systems to evolve toward low-energy, high-efficiency states. By emulating this principle, PEOA is designed to be an optimization algorithm that reflects this natural efficiency, delivering high robustness and excellent performance across a wide range of applications. The algorithm is rigorously evaluated against a range of classical and state-of-the-art (SOTA) methods on standard benchmark functions and real-world problems. PEOA delivers consistently top-tier performance across varied population sizes (N = 20, 50, 100) on standard benchmarks, as evidenced by its leading rankings, especially in high-dimensional problems where it achieves first place in approximately 80% of cases. The algorithm also maintains high-precision convergence across varying dimensions (D = 30, 50, 100), even with a small population size (N = 20), confirming its robustness to dimensional scaling. In real-world applications, PEOA yields the lowest objective values for engineering problems and achieves the lowest Mean Squared Error (10.7184) in neural network hyperparameter tuning in the shortest time (197 s). These outcomes highlight both its high predictive accuracy and computational advantages, collectively establishing PEOA as a robust and efficient solver for diverse and complex optimization challenges.