Improved particle swarm optimization based on hyperbolic cross points algorithm

Particle swarm optimization (PSO) has many advantages, such as swarm intelligence, intrinsic concurrency, simple iteration format, and fast convergence speed, so it has attracted much research interest. An improved standard particle swarm optimization based on hyperbolic cross points algorithm (HCPA-SPSO) has good global and local search capability, which can significantly improve the search accuracy and success rate of SPSO. First, a particle initialization strategy is proposed to generate the initial particle swarm by global HCPA. Second, HCPA is introduced as a local evolution operator. The fuzzy C-means clustering method is used to classify the particle swarm at each fixed iteration step, and a local HCPA search for the representative particles in each class is performed. Finally, the proposed algorithm is compared with SPSO at typical test functions. For the objective function of Multi-local minima-shaped, Bowl-shaped and Valley-shaped, HCPASPSO can significantly improve the success rate and accuracy of the optimal solution. For the Plate-shaped objective function, HCPA-SPSO can further improve the success rate and the accuracy of the optimal solution. The accuracy of the proposed algorithm is similar to that of SPSO for the Steep ridges-shaped objective function whose optimal solution is infinite non-repeating decimal. By combining the global search capability of SPSO and the local search capability of HCPA, the proposed algorithm can effectively balance the exploration and development capability of the whole algorithm. The research results can provide an optimization algorithm with higher accuracy for optimal design and fault diagnosis.