基于SVR的电梯交通流时间序列预测  

Time Series Prediction of Elevator Traffic Flow Based on SVR

作　　者：唐海燕[1] 于德亮[1] 丁宝[1] 齐维贵[1] 

机构地区：[1]哈尔滨工业大学电气工程及自动化学院,黑龙江哈尔滨150001

出　　处：《控制工程》2011年第5期723-726,792,共5页Control Engineering of China

基　　金：国家十五科技支撑计划重大项目(2006BAJ03A05)

摘　　要：为了使电梯群控系统更好地跟踪电梯交通流的变化以提高群控系统的性能,提出了基于支持向量回归(Support Vector Regression,SVR)的电梯交通流预测方法。针对电梯交通流时间序列小样本的特性,考虑了电梯交通流的横向和纵向变化趋势,采用SVR算法建立了电梯交通流时间序列的预测模型。给出了预测的评价指标,研究了SVR模型中的参数对预测效果的影响,利用试验寻优的方法确定了SVR预测模型的最优参数。最后,与电梯交通流RBF神经网络预测模型进行了比较研究,分析了数据样本中波动较大部分的预测效果,结果表明SVR算法比RBF神经网络方法具有更好的预测性能、泛化能力和鲁棒性,实现了电梯交通流较好的拟合和预测。In order to make elevator group control system better follow the change of elevator traffic flow and improve the system per-formance, the prediction method of elevator traffic flow based on support vector regression （SVR） is proposed. The time series prediction model of elevator traffic flow based on SVR is founded due to the characteristics of small sample, and the transverse change trend as well as lengthways change trend of elevator traffic flow are considered. The influences on the prediction effects by parameters of the model are studied, and the evaluation index are given. Then the optimum parameters are confirmed by experiment. In order to verify the superiority of prediction model, comparing research with RBF neural network prediction model of elevator traffic flow is carried through. The prediction effect on the part of sample which has higher wave is analyzed. Simulation results show SVR algorithm has much better prediction performance, generation ability and robust ability. Meanwhile, the fitting and prediction of elevator traffic flow with better effect are realized.

关 键 词：支持向量回归 电梯交通流 预测 RBF神经网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]