基于强化蚁群算法的高层住宅电梯群调度研究  

Research on Elevator Group Control and Scheduling of HighResident-Rise ial Buildings Based on Reinforcement Learning and Ant Colony Algorithm

作　　者：顾玲丽 董佳琦 许洪华[1] GU Ling-li;DONG Jia-qi;XU Hong-hua(School of Electronics and Information Engineering,Suzhou University of Science and Technology,Suzhou Jiangsu 215009,China)

机构地区：[1]苏州科技大学电子与信息工程学院,江苏苏州215009

出　　处：《计算机仿真》2022年第1期412-417,共6页Computer Simulation

摘　　要：在电梯群调度系统研究中,蚁群算法应用较为广泛,但仍存在迭代次数多、收敛速度慢等问题,同时高层建筑电梯鲜有优化调度。针对上述问题,提出一种将强化学习和蚁群算法相结合的高层电梯群控调度方法:建立以用户乘梯体验和系统运行能耗的多目标函数优化调度模型,用Q-learning迭代寻优后的Q值初始化蚁群算法的信息素,同时也将Q值引入概率路径选择中,用并行蚁群算法进行最终派梯策略的寻优。用Python进行实验分析,结果表明所提方法较单蚁群算法有较快的收敛性,在很多大程度上减少了业主的待梯、乘梯时间以及系统运行能耗。In the study of elevator group scheduling systems, the ant colony algorithm is widely used, but there are still many problems such as many iteration times, slow convergence speed, and other problems, in addition, there is little optimal scheduling of elevators in high-rise buildings. Aiming at these problems, a high-level elevator group control scheduling method combined with reinforcement learning and the ant colony algorithm is put forward. A multi-objective function optimal scheduling model based on user elevator experience and system operation energy consumption was established. The pheromone of the ant colony algorithm was initialized with the Q value after Q-learning iterative optimization. At the same time, the Q value was also introduced into probabilistic path selection, and the parallel ant colony algorithm was used to optimize the final elevator dispatching strategy. The experimental analysis with Python shows that the method described in this paper has a faster convergence than the single ant colony algorithm, and it can reduce the time of waiting for and riding the ladder and the energy consumption of the system to a large extent.

关 键 词：高层住宅电梯群控调度 多目标优化 蚁群算法 信息素 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]