论文部分内容阅读
针对广泛存在于化工生产过程中的并行多机间歇调度问题,提出了一种自适应分布估计算法,用于最小化最早完工时间(makespan)。首先,提出了一种具有自适应学习能力的改进策略,该策略根据当前解的改善状况自适应调节学习速率,有效克服了EDA对学习速率较敏感和依赖的不足,进而使得算法的搜索宽度和深度得到合理平衡;其次,设计了一种基于双精英个体的协同进化策略,该策略通过双概率模型协同进化,使算法能充分利用优秀个体的信息来指导搜索方向。仿真实验和算法比较验证了AEDA的有效性和鲁棒性。
Aiming at the problem of parallel multi-machine intermittent scheduling widely existed in the process of chemical production, an adaptive distribution estimation algorithm is proposed to minimize the makespan. First of all, an improved strategy with adaptive learning ability is proposed, which adaptively adjusts the learning rate according to the improvement of the current solution and effectively overcomes the shortcomings of the EDA being more sensitive and dependent on the learning rate, so that the search width and The depth is reasonably balanced; Secondly, a co-evolutionary strategy based on dual-elite individuals is designed, which co-evolves through the dual-probabilistic model so that the algorithm can make full use of excellent individual’s information to guide the search direction. The simulation experiment and algorithm comparison verify the validity and robustness of AEDA.