摘要
多峰优化问题既要求尽可能多的全局最优解,又要求其具有较高的精度,这就需要算法既能够协调种群的多样性和收敛性,又要具有较强的全局搜索能力。为此,提出邻域变异的黑猩猩多峰优化算法。它包括三个创新点:多变异邻域定向引导机制,随机变异增加种群多样性,最佳位置变异能在划分不同的子群内逐步定位到全局最优解,从而避免个体陷入局部最优;全局正向向量扰动机制,利用在进化过程中存储对当前个体更有潜力的进化方向,并在此方向上生成新的子代,使种群逐步向全局最优解收敛,提高其求解精度;参数的动态调整调整机制,能够有效降低算法对参数的敏感性。最后,利用典型测试函数进行验证。仿真结果表明,该算法在多种测试函数中表现优异,能有效处理复杂多峰优化问题。
In order to solve multipeak optimization problem,it requires as many global optimal solutions as possible and high precision,the algorithm is required to coordinate the diversity and convergence of the population and also have strong global search capability.Therefore,this paper proposed neighborhood variant chimp multipeak optimization algorithm.It included three innovative points.Multivariant neighborhood oriented guidance mechanism could increase population diversity by random variation and the best position variation could gradually locate the global optimal solution within different subgroups,thus avoiding individuals falling into local optima.Global forward vector perturbation mechanism utilized the storage of more potential evolutionary directions for the current individuals during the evolution process,and generated new offspring in directions,enabled the population to gradually converge towards the global optimal solution,improved its solution accuracy.Mechanism of parameters dynamic adjustment effectively reduced the sensitivity of the algorithm to parameters.Finally,typical test functions were used for validation.Simulation result shows algorithm performs well in various test functions,can effectively handle complex multipeak optimization problems.
作者
齐名军
王志宝
Qi Mingjun;Wang Zhibao(Dept.of Public Base Education,Hebi Polytechnic,Hebi Henan 458030,China;School of Computer&Information Technology,Northeast Petroleum University,Daqing Heilongjiang 163318,China)
出处
《计算机应用研究》
北大核心
2025年第6期1784-1793,共10页
Application Research of Computers
基金
国家自然科学基金资助项目(50138110)
河南省重点研发与推广专项科技攻关项目(242102210054)。
关键词
向量扰动
变异因子
定向引导
控制参数
预测模型
vector disturbance
variation factor
directional guidance
control parameter
prediction
