混合约束多峰优化问题的一个协同共轭进退粒子群算法
2023-04-29陈相兵陈晨闵心畅
陈相兵 陈晨 闵心畅
为解决混合(等式和不等式)约束的多峰优化问题(MOPs),本文在粒子群算法框架下提出了粒子优度比较准则和局部协同与共轭进退寻优两种迭代进化策略. 优度比较准则在适应度和约束违反度的双重限制下指导粒子高效地执行进化策略,局部协同策略可使粒子能通过局部抱团收敛到多个全局最优解,而共轭进退寻优策略则提升了寻优的速度和精度. 基于优度比较准则与两种进化策略的有效结合,本文设计了一个协同共轭进退粒子群(CCARPSO)算法,以充分融合粒子群算法的全局搜索能力和共轭进退法的局部快速寻优能力. 数值仿真表明, 该算法能有效解决复杂约束MOPs和非线性方程组的多根问题,在广义Logistic分布的参数估计中有全局优化能力和较高的计算精度.
多峰优化; 优度比较; 局部协同; 共轭方向; 进退法; 粒子群
O29A2023.011006
收稿日期: 2022-01-22
基金项目: 四川省科技计划(2022JDRC0068, 2021JDRC0080); 四川省教育厅项目(18ZB0363); 中国民用航空飞行学院校级项目(J2021-058)
作者简介: 陈相兵(1985-), 男, 安徽枞阳人, 博士, 副教授, 主要研究方向为应用数学. E-mail: chenxb85@sina.com
通讯作者: 陈晨.E-mail:chenchen_uni@foxmail.com
A cooperative conjugate advance-retreat particle swarm optimization algorithm for hybrid constrained multimodal optimization problems
CHEN Xiang-Bing1, CHEN Chen2, MIN Xin-Chang3
(1.Division of Mathematics, Sichuan University Jinjiang College, Meishan 620860, China; 2. College of Science, Civil Aviation Flight University of China, Guanghan 618307, China;3. School of Mathematics, Sichuan University, Chengdu 610044, China)
This paper aims at the multimodal optimization problems (MOPs) with equality and inequality constraints. A new algorithm is proposed following the particle swarm optimization idea. This algorithm consists of a superiority comparison criterion and two iterative evolutionary strategies. The superiority comparison criterion guides the particles on how to evolute according to the constructed constraint violation degree and the fitness (i.e., the objective function value). The local cooperation strategy ensures that all particles can converge to multiple global optimal solutions through local clustering. The conjugate advance-retreat optimization strategy improves the speed and precision of optimization. Our algorithm, named cooperative conjugate advance-retreat particle swarm optimization (CCARPSO) algorithm, integrates the global searching ability of PSO and the local fast optimization capability of conjugate advance-retreat method. In numerical simulations, the algorithm effectively solves MOPs with complex constraints and nonlinear equations with multiple solutions, and has high global optimization ability and calculation accuracy in estimating parameters of the generalized Logistic distribution.
Multimodal optimization; Superiority comparison; Local cooperation; Conjugate direction; Advance-retreat method; Particle swarm
1 引 言
随着大数据时代的到来,实时数据的数量和种类急剧增加,海量数据出现在多个领域,例如医疗诊断[1]、市场决策[2]、路径规划[3]和光伏阵列[4]等. 随之,多变量、多约束的多峰值优化问题[5](MOPs:Multimodal Optimization Problems)时常出现而且亟待解决.
传统的优化方法,例如牛顿法、共轭梯度法、单纯形法以及分支定界法等,通常要求目标和约束函数可导,且容易陷入局部极值. 智能进化算法选择则利用群体智慧,能够并行处理超大规模优化问题. 智能进化算法包含差分进化(Differential Evolution, DE)[6]算法、粒子群优化(Particle Swarm Optimization, PSO)算法和遗传算法(Genetic Algorithm, GA)[7]等. 在MOPs的优化算法中,智能进化算法是当前的热点算法之一,如基于DE的小生境方法[8-9]. 为了降低参数的影响,一些新的进化算子融入了小生境策略[10-14].基于DE的小生境方法已经成功地应用于无约束或带简单约束的MOPs. 然而,对于带复杂约束(如非线性约束)的MOPs,相关的研究工作还极为少见,适用的智能进化算法有待研究.
PSO算法模拟了自然群体生命现象的自组织、自学习和自适应性,依据适应度和优胜劣汰法则迭代地搜索解空间的最优个体[15]. 它不仅不需要目标函数的梯度信息,而且具有操作简单、可并行计算和模型参数少等优点,已在众多领域发挥了重要作用,如基数约束的投资组合优化[16]和多元线性回归参数估计[17]等.
针对带混合(等式和不等式)约束的MOPs,本文设计了新的PSO算法,其中的粒子基于优度比较准则在局部协同和共轭进退寻优迭代两种进化策略中选择合适的进化策略,我们称之为协同共轭进退PSO(Cooperative Conjugate Advance-Retreat PSO,CCARPSO)算法. 其中,优度比较准则在约束违反度和适应度(目标函数值)的双重限制下指导粒子有效地迭代进化,协同策略则解决了一般PSO算法容易早熟和难以同时寻找多最优解的问题,共轭进退寻优策略则提升了寻优的速度和精度. 该算法充分融合了PSO的全局搜索能力和共轭进退法的局部快速寻优能力. 数值实验表明,在求解带混合约束的MOPs和多根的非线性方程组时本文提出的CCARPSO算法具有优良性能.
6 结 论
本文构建了带混合约束的MPOs的CCARPSO算法. 该算法是一种基于优度比较准则来选择迭代策略的粒子群算法. 共轭进退寻优策略保证粒子能朝着可行域方向快速进化,局部协同策略使所有粒子能通过局部抱团收敛到多个全局最优解. 数值仿真实验表明,CCARPSO算法有效地解决了非线性约束MOP和多根非线性方程组的求解问题. 另外,基于碳素纤维硬度的实际数据,我们利用CCARPSO算法求解参数估计优化问题,给出了稳健、精确的碳素纤维硬度模型参数估计.
参考文献:
[1] Tavard F, Simon A, Leclercq C, et al. Multimodal registration and data fusion for cardiac resynchronization therapy optimization [J]. IEEE T Med Imaging, 2014, 33: 1363.
[2] Zaman F, Elsayed S M, Ray T, et al. Evolutionary algorithms for finding Nash equilibria in electricity markets [J]. IEEE T Evolut Comput, 2018, 22: 536.
[3] Zhao Y, Ioannou P A, Dessouky M M. Dynamic multimodal freight routing using a co-simulation optimization approach [J]. IEEE T Intell Transp, 2019, 20: 2657.
[4] 徐仪圆, 许祺峰. 适用于商业航天的全局MPPT优化算法[J]. 西北工业大学学报, 2020, 38: 133.
[5] Li X, Epitropakis M G, Deb K, et al. Seeking multiple solutions: an updated survey on niching methods and their applications [J]. IEEE T Evolut Comput, 2017, 21: 518.
[6] 李汶骏, 龙伟, 曾力. 基于差分进化和核主元分析的燃气轮机故障检测[J]. 四川大学学报: 自然科学版, 2021, 58: 022004.
[7] 邓希, 胡晓兵, 江代渝, 等. 基于混合遗传算法的柔性作业车间机器和AGV规划[J]. 四川大学学报: 自然科学版, 2021, 58: 022003.
[8] Thomsen R. Multimodal optimization using crowding-based differential evolution [C]//Proceedings of the 2004 Congress on Evolutionary Computation. Piscataway: IEEE, 2004.
[9] Gao W, Yen G G, Liu S. A cluster-based differential evolution with self-adaptive strategy for multimodal optimization [J].IEEE T Cybernetics, 2014, 44: 1314.
[10] Zhao H, Zhan Z H, Lin Y, et al. Local binary pattern-based adaptive differential evolution for multimodal optimization problems [J]. IEEE T Cybernetics, 2020, 50: 3343.
[11] Wang Z J, Zhan Z H, Lin Y, et al. Automatic niching differential evolution with contour prediction approach for multimodal optimization problems [J]. IEEE T Evolut Comput, 2020, 24: 114.
[12] Chen Z G, Zhan Z H, Wang H, et al. Distributed individuals for multiple peaks: A novel differential evolution for multimodal optimization problems [J]. IEEE T Evolut Comput, 2020, 24: 708.
[13] 陈宗淦, 詹志辉. 面向多峰优化问题的双层协同差分进化算法[J]. 计算机学报, 2021, 44: 1806.
[14] 廖作文. 基于差分进化算法的非线性方程组多根联解研究[D]. 武汉: 中国地质大学, 2019.
[15] Kennedy J, Eberhart R. Particle swarm optimization [C]//A Proceedings of the Fourth IEEE International Conference on Neural Networks. Piscataway: IEEE, 1995.
[16] 朱沙, 陈臣. 一种求解基数约束投资组合优化的混合粒子群算法[J]. 统计与决策, 2016, 4: 64.
[17] 李翼, 张本慧, 郭宇燕. 改进粒子群算法优化下的Lasso-Lssvm预测模型[J]. 统计与决策, 2021, 37: 45.
[18] Li X,Engelbrecht A, Epitropakis M G. Benchmark functions for CEC 2013 special session and competition on niching methods for multimodal function optimization [R]. Melbourne: RMIT University, 2013.
[19] Balakrishnan N. Handbook of the logistic distribution [M].New York: Marcel Dekker, 1991.
[20] 韩雪, 程维虎. 三参数I型广义Logistic分布参数的一类改进估计[J]. 数理统计与管理, 2016, 35: 445.
[21] Kao J H K. Computer methods for estimating Weibull parameters in reliability studies [J]. IRE T Re Qual C, 1958, 13: 15.
[22] Swain J, Venkatraman S, Wilson J. Least-squares estimation of distribution functions in Johnsons translation system [J]. J Stat Comput Sim, 1988, 29: 271.
[23] 陈海清, 曾婕, 胡国治. 三参数I型广义Logistic分布参数的改进最小二乘估计[J]. 数理统计与管理, 2018, 37: 835.