融入概率学习的混合差分进化算法求解绿色分布式可重入作业车间调度

Hybrid differential evolution integrated with probability learning for green distributed reentrant job shop scheduling

中文关键词  差分进化  绿色调度  分布式调度  可重入作业车间调度问题

英文关键词  differential evolution  green scheduling  distributed scheduling  reentrant job shop scheduling problem

中文摘要

      本文针对绿色分布式可重入作业车间调度问题(GDRJSSP), 提出一种融入概率学习的混合差分进化算法 (HDE PL), 以实现最大完工时间和总能耗最小. 根据GDRJSSP的问题特点, 设计编码和解码规则, 并采用差分进化 算法执行全局搜索来发现优质解区域. 为能更明确地引导全局搜索方向, 设计基于贝叶斯网络结构的多维概率模型 合理学习和积累优质解(即当前种群中的较优解)的模式信息. 结合问题解的结构特征, 提出基于关键路径的4种邻 域结构来构造局部搜索, 并设计基于非关键路径的节能策略来提升算法获取低能耗非劣解的能力. 仿真实验和算 法对比验证了HDE PL可有效求解GDRJSSP.

英文摘要

      Aiming at the green distributed reentrant job shop scheduling problem (GDRJSSP), a hybrid differential evolution incorporated with probabilistic learning (HDE PL) is proposed to minimize the maximum completion time and the total energy consumption. According to the problem characteristics of the GDRJSSP, the rule of job allocation among factories and the encoding and decoding rules are designed, and the differential evolution algorithm is used to perform global search to find high-quality solution regions. In order to guide the global search direction more clearly, a multidimensional probability model based on Bayesian network structure is designed to reasonably learn and accumulate the pattern information of high-quality solutions (i.e., the better solutions in the current population). Combined with the structural characteristics of the problem solution, four neighborhoods based on the critical path are proposed to construct the local search, and an energy saving strategy based on the non-critical path is devised to enhance the ability of the algorithm to obtain low-power non-dominated solutions. Simulation experiments and algorithm comparisons verify that HDE PL can effectively solve the GDRJSSP.