膜计算多粒子群算法

Membrane Computing Multi Particle Swarm Optimization(MC-MPSO) Algorithm

针对粒子群(Particle swam optimization, PSO)算法进化后期收敛速度较慢,易陷入局部极值点,精度较差等不足,提出膜计算多粒子群(MC-MPSO)算法。在该算法中,将原始PSO、标准PSO、中值导向粒子群(MPSO)、扩展粒子群(EPSO)、多作用力粒子群(MFPSO)、两阶段作用力粒子群(TFPSO)等六种具有不同优点的粒子群算法分别放入六个基本膜内,提出MC-MPSO算法的膜间交流与粒子更新机制,在进化前期,各粒子群算法按自身机制进行搜索寻优,即各基本膜各自进化来充分发挥各基本膜内算法的优点;在进化后期,各基本膜内算法与比自身更好的表层膜内最优解粒子交流,各表层膜逐步吞并搜索能力较差的基本膜,而最适合问题优化求解的基本膜长大并按照表层膜输出,使MC-MPSO算法集成了基本膜内六种粒子群算法的各自优势,并具有适应不同类型优化求解问题的寻优能力。通过与基本膜内六种粒子群算法的测试对比,与遗传算法、鱼群算法及其他基于膜计算的粒子群算法的比较,证明了MC-MPSO算法具有更好的寻优能力和适用性。最后,将MC-MPSO算法应用于串联和桥式系统可靠性优化问题,验证了所提算法的有效性。

Membrane computing multi particle swarm optimization(MC-MPSO) algorithm is proposed to overcome of particle swarm opmtimization(PSO) algorithm the defections of easy getting trapped in a local optimum, slow convergent speed and low accuracy in the later evolution process. In MC-MPSO algorithm, original PSO, standard PSO, median-oriented PSO(MPSO), extended PSO(EPSO), multi force PSO(MFPSO), two-stage force PSO(TFPSO) with different advantages of particle swarm algorithm are put into six membranes respectively. The communication among membranes and update mechanisms of particles are proposed in MC-MPSO. First of all, elementary membrane grow up according to their own searching mechanism with the advantages of each PSO algorithm. Secondly, six algorithms in the membranes exchange optimally with better membrane, and the surface membrane gradually swallow the membranes of poor searching ability. Then the membranes which can solve the problems properly grow up and the best membrane export through surface membrane. The MC-MPSO algorithm integrates the advantages of the six particle swarm optimization algorithms, and has the ability to adapt to different types of optimization problems. By comparing with the test of six algorithms in the membranes, the comparison of genetic algorithm, fish swarm algorithm and other particle swarm optimization algorithms based on membrane computing, the results show that the MC-MPSO algorithm has better search capability of optimal solution and wide applicability. Finally, the MC-MPSO algorithm is applied in the reliability optimization of series and bridge systems. The effectiveness of the proposed algorithm is verified.