基于动态平衡的多目标粒子群优化算法
2019-10-21张海波
张海波


摘 要: 传统粒子群优化算法在求解动态优化问题时,种群将逐渐收敛,从而在问题变化后无法进一步寻优,针对上述问题,提出了一种基于动态平衡的多目标粒子群优化算法。采用双种群策略以动态平衡算法的探索能力与开发强度,其中一个子种群在动态调整的网格中运行混沌搜索,确保种群多样性符合要求的同时,能够有效提升搜索的效率。利用快速收缩多目标粒子群算法,对另外的子种群进行计算,收敛到Pareto前沿。通过一组标准测试问题对所提方法进行了驗证,实验结果显示所提算法无论在收敛速度还是在优化精度上都优于其它典型多目标进化算法。
关键词: 动态平衡; 混沌; 自适应网格; 多目标优化; 粒子群算法
中图分类号: TP311
文献标志码: A
文章编号:1007-757X(2019)06-0150-04
Abstract: In the traditional particle swarm optimization algorithm for solving the dynamic optimization problem, the population will gradually converge, thus the algorithm may lose further optimization ability after the environment dynamically changes. In order to solve this problem, a dynamic adaptive grid is presented based on multi-objective particle swarm optimization. In the algorithm, a double-population strategy is designed to explore and develop intensity balance algorithm, one of the subpopulation runs for adaptive grid search, so as to ensure the population diversity and the adaptive mechanism to improve the search efficiency, another subpopulation uses multi-objective particle swarm algorithm for rapid contraction. The algorithm can rapidly converge to the Pareto front under environmental change. Through a set of standard test problems, the experimental results show that the proposed algorithm both in convergence speed and optimization accuracy are better than other typical multi-objective evolutionary algorithms.
Key words: Dynamic balance; Chaos; Adaptive grid; Multi-objective optimization; Particle swarm algorithm
0 引言
动态多目标优化问题集合了当前优化问题的两个难点即需要优化的目标不仅为多个、相互冲突[1-3],解决上述问题不仅需要算法能搜索到分布尽可能广泛的Pareto最优解,而且还要快速适应环境变化,使优化结果能跟踪Pareto最优前沿移动。
进化计算、粒子群优化算法等基于种群随机搜索的自然启发式算法被用来求解动态多目标优化问题[4]。这些算法与传统算法之间的差异性总结为:第一,基于种群随机搜索的目标优化算法能够同时获取多个Pareto最优解[5]。第二、这些算法对环境变化具有自适应能力,而不需要重新优化[6]。第三、不要求获得模型信息,并且适用于求解计算复杂的优化问题[7]。……
