A Modified Comprehensive Learning Particle Swarm Optimizer

摘要

To overcome premature convergence and falling into local optima of particle swarm optimization (PSO), a comprehensive learning particle swarm optimizer (CLPSO) has been proposed. However, it is not good at solving unimodal problems. In this paper, we propose a modified CLPSO (MCLPSO) with three improvements. Firstly, we use opposition-based learning (OBL) to improve the initial population. Secondly, we add the best solution of the population to the list of selected particles in order to improve the convergence ability while maintaining the population diversity. Finally, we use the mean velocity of the population with a linearly decreasing probability to update the particle velocity to further improve the performance of CLPSO. The MCLPSO algorithm is tested on CEC2005 in 30 dimensions. Furthermore, the MCLPSO is conducted to solve hydrothermal scheduling problems. The experimental results demonstrate that the solution accuracy of MCLPSO is overall better than those of comparison algorithms.