一种高斯反向学习粒子群优化算法

Particle Swarm Optimization Algorithm with Gaussian Opposition-based Learning

针对粒子群算法在处理多峰复杂问题时,收敛速度慢且容易陷入局部最优的缺点,提出一种高斯反向学习粒子群优化算法(GOL-PSO).针对历史最优粒子间无法相互交流,增加一种高斯反向学习机制来提高粒子的学习能力,进而提高算法的搜索能力,另外算法在更新公式中引入"历史最优平均值"因子来提高算法的收敛速度.经过在8个测试函数的仿真实验中,与一些改进的粒子群算法进行比较,GOL-PSO有5个测试函数的测试效果最好,且T检验结果表明算法结果有明显提高,同时算法收敛对比分析结果表明,本文算法具有良好的全局搜索能力和较快的收敛速度.

To the weaknesses of easily get trapped in the local optima and poor convergence speed when solving complex multimodal problems, a particle swarm optimization algorithm with Gaussian opposition-based learning is proposed. A strategy of gaussian and opposition-based learning is performed when the local best particle is not change with other. The strategy will improve algorithm＇s search ability, through enhancing learning ability. Then to enhance the convergence speed, when producing the velocity update pattern with the factor of local best average. The simulation results of the problem in 8 test functions show that, compared with other improved PSO variants, GOL-PSO is better than other algorithms in 5 test functions. What＇s more, both the T-test and the convergence analysis of the algorithm show that GOL-PSO has good global search ability and faster convergence speed.