一种自适应强制进化随机游走算法应用于换热网络综合
An adaptive random walk algorithm with compulsive evolution algorithm for heat exchange network synthesis
段欢欢 1易智康 2张笑恬 2肖媛 2崔国民2
作者信息
- 1. 上海理工大学能源与动力工程学院上海市动力工程多相流动与传热重点实验室,上海 200093;河南牧业经济学院能源与智能工程学院,河南郑州 450000
- 2. 上海理工大学能源与动力工程学院上海市动力工程多相流动与传热重点实验室,上海 200093
- 折叠
摘要
RWCE(强制进化随机游走)算法应用于系统热集成时,最大步长既影响当前可行搜索域的范围,又影响整型变量的进化,固定参数设置降低了更优解产生的几率.因此提出一种融合自适应步长和自适应反向学习策略的RWCE算法.建立随机动态步长,在导向参数牵引下自动激励有利步长值持续进化;在此基础上,建立自适应反向学习策略改变个体进化路径,使算法在优化的不同阶段能够自动搜索最佳步长,并挖掘尽可能多的结构,充分发挥算法全局搜索和局部开发能力.最后研究并计算H6C10、H10C10、H13C7 3个典型中大规模算例,结果表明该方法能够进一步提升算法的寻优能力.
Abstract
When the RWCE(random walk algorithm with compulsive evolution)algorithm was applied to energy system optimization,the maximum step length affected both the range of the current feasible search domain and the evolution of integer variables.The fixed parameter setting further reduced the probability of optimal solutions.Therefore the RWCE algorithm that integrated adaptive step size and opposition-based learning strategy was proposed.A stochastic dynamic step size was established to automatically motivate the beneficial step size valued to evolve continuously under the traction of the guiding parameter.On the basis,the individual evolution path was changed by the adaptive opposition-based learning,so that the algorithm could automatically search for the better step size at different stages of optimization and explore as many structures as possible,so as to give full play to the global search and local exploitation capability of the algorithm.Finally,three typical medium-to-large scale cases H6C10,H10C10 and H13C7 were studied and evaluated.The results show that the proposed method can further improve the algorithm's search capability.
关键词
自适应/导向参数/反向学习/换热网络/RWCEKey words
adaptive method/guiding parameter/opposition-based learning/heat exchanger networks/RWCE引用本文复制引用
基金项目
国家自然科学基金资助项目(21978171)
国家自然科学基金资助项目(51976126)
中国博士后科学基金资助项目(2020T13043)
出版年
2024
国家科技期刊平台
NSTL
万方数据