求解约束优化问题的动量粒子群算法  

Momentum Particle Swarm Optimizer for Constrained Optimization

作　　者：马瑞新[1] 刘宇[1] 覃征[2] 王晓[1] 

机构地区：[1]大连理工大学软件学院,大连116621 [2]清华大学软件学院,北京100084

出　　处：《系统仿真学报》2010年第11期2485-2488,共4页Journal of System Simulation

基　　金：国家自然科学基金(60803074);辽宁省自然科学基金(60803074)

摘　　要：为解决约束优化问题,提出使用双可行域吸引子策略改进动量粒子群算法。该算法只需初始种群中有一个粒子位于可行域内,随着搜索过程的进行,整个种群自动进入可行域内搜索。一方面,在搜索过程早期,由于可行域内粒子少,所有粒子移向相同的吸引子,整个种群迅速进入可行域内。另一方面,随着进入可行域粒子的增多,由于每个粒子使用距本身最近的可行域吸引子,较好地维持了种种群的多样性,避免早熟现象的发生,使算法具有较好的寻优性能。与国际上当前解决约束优化问题的粒子群算法在4个标准约束优化函数上测试比较,实验结果表明本算法取得的最优值要优于其它粒子群算法。The strategy that two good positions in feasible region worked as attractors was incorporated into momentum particle swarm optimization algorithm in order to resolve constrained optimization problems. The resulting algorithm only requires that one of the initial particles is in the feasible region, and then all particles in the swam automatically move into the feasible region. On the one hand, in the early iterations few particles appear in the feasible region and hence all particles move toward the same attractors, so the particles soon enter into the feasible region. On the other hand, as the number of particles in the feasible region increases, each particle adopts the most near attractor so that each particle has different attractor. Therefore, the algorithm maintains the diversity of the population, alleviates the premature, and hence achieves good performance. The algorithm is compared with other particle swarm optimization algorithms on four benchmark functions. The experimental results show that the solution of the algorithm is better than that of others.

关 键 词：粒子群算法 约束优化问题 可行域 进化计算 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]