An Improved Particle Swarm Optimization Algorithm Based on Multi-Strategy Coupling

To enhance the global convergence and solution accuracy of the Particle Swarm Optimization (PSO) algorithm in complex optimization problems, a MultiStrategy Coupled Improved Particle Swarm Optimization algorithm (MSC-PSO) is proposed. This algorithm systematically reconstructs the search mechanism of the traditional PSO by synergistically integrating Cauchy mutation-based velocity perturbation for particle updates and the natural selection mechanism from genetic algorithms. Firstly, after adaptive adjustment using a piecewise nonlinear inertia weight function, the particle velocity update is perturbed via Cauchy mutation so that the algorithm is better equipped to escape from local optima and locate the global optimum or solutions closer to it. Secondly, the natural selection mechanism of genetic algorithms is introduced. A roulette wheel selection strategy is employed to dynamically remove ineffective particles, thereby augmenting population diversity. Meanwhile, through the coordinated implementation of cosine-adaptive inertia weight adjustment and nonlinear tuning of learning factors, the algorithm achieves a natural equilibrium between global exploration and local exploitation during the search process, while mitigating oscillations and instability. Finally, quantitative validation of the improved algorithm's advantages in terms of global convergence and accuracy is conducted through tests on Sphere Function and Generalized Rastrigin's Function. The results demonstrate that the MSC-PSO algorithm exhibits high solution accuracy and global convergence capabilities for both unimodal and multimodal functions. Notably, it significantly outperforms other PSO variants in solution accuracy and global convergence for unimodal functions.