摘要
高精度的短时进站客流量预测对城市轨道交通日常客流组织具有重要意义,利用客流预测结果在事前实施限流、疏导等措施,较事后控制更及时、先进。通过采集15 min间隔的地铁进站客流数据,利用上周同期进站量、本日上一时段进站量以及高峰和非高峰时段参数作为输入变量,尝试分别采用加权历史平均自回归模型、ARIMA模型及小波神经网络模型进行短时预测,以获得精度最高的模型。在此基础上,进行三种方法组合预测,探究组合预测效果。通过案例分析,发现当考虑时段因素时,小波神经网络预测精度最高,为91.05%;ARIMA模型误差结构最好。当采用所提出的组合预测模型后,预测精度指标较独立预测模型均有提升,但误差结构没有得到改善。研究表明,所提组合预测模型可以有效地应用于城市轨道交通进站客流的短时预测中。
Accurate prediction of short-term passenger entry flow plays an important role in the daily operation of urban rail tran- sit. The prediction can provide evidence for advanced approach for passenger flow control and induction. Using the same period volume in the previous week, last period volume and peak and off-peak hour＇s data of a day as input variables, this paper uses weighted historical average auto regression model, ARIMA model and wavelet neural network model to forecast the passenger en- try flow of urban railway transit stations, by collecting real time passenger flow data with 15min intervals. Based on these approa- ches, a combined method is employed to test the prediction power. A case study is conducted and the forecast results are com- pared. It is found that, when the peak hour is considered, the prediction accuracy of wavelet neural network model is 91.05%, which ranks the highest. The error structure of ARIMA model is the best. When the combination method is adopted, the predic- tion performance indicator improves significantly, while the error distribution does not improve at all. It is concluded that the proposed combination model is effective in short-term prediction of passenger entry flow in urban rail transit.
出处
《都市快轨交通》
北大核心
2017年第1期54-58,64,共6页
Urban Rapid Rail Transit
基金
北京市自然科学基金(9132015)
基本科研业务费(2016JBM030)
北京市高等学校青年英才计划(YETP0555)
关键词
城市轨道交通
进站客流量
短时预测模型
组合预测
urban rail transit
passenger entry flow
short-term prediction model
combination prediction
