基于增量学习的电梯调度算法研究

Research on Elevator Dispatch Algorithm Based on Incremental Learning

摘要: 针对现有电梯调度算法对用户动态需求适应能力不足的问题，提出了一种基于增量学习的电梯调度算法。首先，搭建了电梯仿真模型，模拟了某小区16层楼栋用户出入情况，采集了规模为11,424的样本集；其次，分析了楼栋用户在工作日、休息日使用电梯的时间分布特性，提取了用户在不同时间下的动态需求变化规律；然后，搭建了增量学习预测模型，根据不同时间输出电梯当前应停靠的楼层数值；最后，经过测试集分析，所提出的预测模型能够准确预测用户的实时出行需求，达到了96.875%的准确率，同时与现有算法相比，所提出的算法能够更加准确地掌握用户的动态需求变化规律，提前停靠在相应的楼层，有效减少了用户的候梯时间，提升了电梯效率。该算法在电梯系统智能化调度建设方面具有重要意义。

Abstract: Aiming at the problem that the existing elevator dispatch algorithm is not able to adapt to the dynamic needs of users, an elevator dispatch algorithm based on incremental learning is proposed. First, an elevator simulation model was built to simulate the entry and exit of users in a 16-story building in a community, and a sample set of 11,424 was collected; secondly, the time distribution characteristics of the use of elevators by building users on weekdays and weekends were analyzed, and the dynamic demand change rules of users at different times were extracted; then, an incremental learning prediction model was built to output the floor value where the elevator should currently stop according to different times; finally, after the test set analysis, the proposed prediction model can accurately predict the real-time travel needs of users, reaching an accuracy rate of 96.875%. At the same time, compared with the existing algorithm, the proposed algorithm can more accurately grasp the dynamic demand change rules of users, stop at the corresponding floor in advance, effectively reduce the waiting time of users, and improve the efficiency of elevators. This algorithm is of great significance in the construction of intelligent dispatching of elevator systems.