关于多目标优化算法搜索性能优化研究

Research on Searching Performance Optimization of Multi-Objective Optimization Algorithm

在约束多目标优化问题中,约束条件的限制使得优化算法在收敛到最优解或保持解集多样性方面存在很大困难,为了提高算法的多样性和收敛性,提出一种将人工免疫系统与多Agent系统相结合的约束多目标优化算法。算法结合二者的优点,通过邻域克隆选择行为、邻域竞争行为、邻域协作行为以及自学习行为来完成高效的局部和全局搜索。算法用距离值和惩罚项对Agent个体的目标函数值进行修正。在进化过程中,充分利用约束偏离值较小的不可行解,以保持种群多样性,避免早熟收敛。在标准测试函数(CTP测试集)上,将提出的算法与其它3种优秀算法进行对比实验,实验结果表明,提出的算法所求解集的多样性和收敛性比其它3种算法均有一定的提高,搜索性能得到了优化。

In constrained multi-objective optimization problems, constraint conditions make them difficult for optimization algorithms to converge to the optimal solution or keep individual diversity. To improve the diversity and convergence of algorithms, a constrained multi-objective optimization algorithm combining artificial immune system with multi-agent system is proposed. The proposed algorithm combines the advantages of the artificial immune system and the multi-agent system, and completes the local and global search efficiently through neighborhood clone selec- tion operator, neighborhood competition operator, neighborhood collaboration operator, and self-learning operator. The algorithm uses distance value and penalty to modify the objective values of agent individuals. During the evolutionary process, the algorithm utilizes the infeasible solutions with smaller constrained violation values to keep individual diversity and avoid prematurity. On the standard test functions （ CTP series）, the proposed algorithm is compared with another three excellent algorithms. Experimental results show that the optimal solutions of the proposed algorithm are better than those of another three algorithms in terms of diversity and convergence, and the searching performance is optimized.