基于强化学习的改进NSGA-Ⅱ求解柔性作业车间节能调度问题  

Improved NSGA-Ⅱbased on reinforcement learning to solve energy-saving scheduling problem of flexible job shop

作　　者：陆心屹 韩晓龙 LU Xinyi;HAN Xiaolong(Institute of Logistics Science and Engineering,Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海海事大学物流科学与工程研究院,上海201306

出　　处：《现代制造工程》2023年第8期22-35,共14页Modern Manufacturing Engineering

摘　　要：针对绿色制造背景下的柔性作业车间调度问题,建立以最小化完工时间、机器负荷及车间能耗为目标的多目标整数规划模型,并提出一种基于强化学习中的Q学习(Q-learning)的改进快速非支配排序遗传算法(Non-dominated Sorting Genetic Algorithm-Ⅱ,NSGA-Ⅱ)进行求解。首先,采用多启发式算法初始化种群以平衡机器负荷,引入精英库实现双策略混合交叉以提升种群质量。其次,根据种群度量指标构建强化学习状态空间并通过Q-learning训练调整混合交叉比例,保证种群的分布均匀性和多样性,避免算法早熟。最后,通过不同规模的Kacem和Brandimarte基准算例及关于汽车发动机冷却系统零/部件制造的生产实例,对算法性能进行分析评估,验证了模型和算法在求解柔性作业车间调度节能问题的有效性以及在平衡机器负荷和能耗指标上的优越性。For the flexible job shop scheduling problem in the context of green manufacturing,a multi-objective integer programming model was established to minimize the completion time,machine load and energy consumption of the workshop,and an improved Non-dominated Sorting Genetic Algorithm-Ⅱ(NSGA-Ⅱ)based on Q-learning was proposed to solve it.Firstly,multiple heuristic algorithms were used to initialize the population to balance the machine load,and an elite pool was introduced to conduct a dual-strategy hybrid crossover for improving the population quality.Secondly,the state space of reinforcement learning based on two metrics of the population was constructed and the hybrid crossover ratio was adjusted by Q-learning training to ensure the uniformity and diversity of the population distribution,as well as avoiding premature maturity of the algorithm.Finally,the performance of the algorithm was analyzed and evaluated by solving Kacem and Brandimarte benchmark cases and a production example of manufacturing parts for automotive engine cooling system.The effectiveness of the model and algorithm in solving the flexible job shop scheduling problem and the superiority in balancing machine load and energy consumption are verified.

关 键 词：多目标柔性作业车间调度 能耗优化 改进NSGA-Ⅱ 强化学习 多样性度量指标 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]