采用异构搜索的多子群协同进化粒子群算法

针对传统的单种群粒子群优化算法易陷入局部最优、搜索精度低的问题,提出一种异构多子群粒子群算法。算法由自适应子群、精英子群和若干普通子群构成,精英子群由普通子群和自适应子群中的优秀个体组成,每个子种群采用不同策略进行进化,根据种群的早熟收敛程度和粒子的适应度值自适应地调整惯性权重,自适应子群根据普通子群的适应度值和速度自适应调整飞行方向,采用免疫克隆选择算子对精英子群进行精细搜索,普通子群、自适应子群与精英子群之间通过迁移操作实现信息的充分交流。针对典型的Benchmark 函数优化问题测试,仿真结果表明所提算法能较好地保持粒子多样性,收敛精度高且全局搜索能力强,具有良好优化性能。

Multi-swarm cooperative particle swarm algorithm with heterogeneous search strategy

Conventional particle swarm optimization is easily trapped in local optima and has the problem of low search accuracy. A multi-swarm particle swarm optimization with heterogeneous search is proposed. The proposed algorithm consists of one adaptive sub-swarm, one elite sub swarm and several ordinary sub-swarm, particles in elite sub-swarm are outstanding individuals migrate from adaptive sub-swarm and ordinary sub-swarm. Each sub-swarm evolutes with heterogeneous strategies. The inertia weight is changed adaptively according to the degree of population premature convergence. The flight direction of the particles in adaptive sub-swarm adjusts their flight direction according to fitness value and speed of ordinary sub-swarm. The immune clonal selection operator is employed for optimizing the elite sub swarm while the migration scheme is employed for the information exchange between elite sub swarm and others sub swarm. Experiments on four benchmark function show that the proposed method can maintain the diversity of particles with strong global search capability, it converges with high precision and with better optimization performance.