论文部分内容阅读
提出了一种基于K-means全局引导策略的多目标微粒群算法(KMOPSO),通过K-means算法从归档集中选出K个均匀分布的非支配粒子作为全局最优引导,以保证种群中的粒子向整个Pareto前端移动,提高解的多样性.用基于最近邻居的剪枝算法控制归档集规模,同时保证其中非支配解的多样性.引入变异策略来加强算法的局部搜索能力,避免早熟收敛.用5个经典函数进行了仿真测试,实验结果表明,该算法能有效地解决多目标优化问题,不但能收敛于Pareto最优前端,而且在解的多样性方面优于改进的非劣分类遗传算法和基于拥挤距离的多目标微粒群算法.
A multi-objective particle swarm optimization algorithm based on K-means global guidance strategy (KMOPSO) is proposed. By K-means algorithm, K uniformly distributed non-dominated particles are selected from the archive set as global optimal guidance to ensure that Particle moves to the entire Pareto front and enhances the diversity of solution.Using the nearest neighbor pruning algorithm to control the size of the archive set while guaranteeing the diversity of the nondominated solutions.We introduce the mutation strategy to enhance the local search ability of the algorithm and avoid premature convergence The simulation results with five classical functions show that the proposed algorithm can effectively solve the multi - objective optimization problem, and can not only converge to the Pareto optimal front end, but also outperform the improved non - inferiority classification in the diversity of solutions Algorithm and multi-objective particle swarm optimization algorithm based on congestion distance.