城市轨道交通短时客流不确定性预测模型

Uncertainty Prediction Model of Short-term Passenger Flow in Urban Rail Transit

以城市轨道交通实际运营客流数据为基础,针对现有短时客流预测存在的问题,从运营时段特征、客流类型及站点周边用地类型等影响因素出发,剖析了短时客流存在不确定性的原因;基于周期性差分自动平滑回归模型和支持向量机理论,构建了短时客流预测组合模型,捕捉短时客流的周期性特征和局部非线性性特征;为提高短时客流预测结果的可信度,引入广义自回归条件异方差模型来构建短时客流不确定性预测模型。通过实例,验证结果表明,周期性差分自动平滑回归-在线支持向量机组合模型对于周期性强且稳定的客流具有优越的预测性能,广义自回归条件异方差模型的短期客流不确定性预测结果更为准确可靠。

Based on the actual data of passenger flow collected from urban rail transit operation,and aiming at the existing problems in short-term passenger flow prediction,the causes of uncertainty in short-term passenger flow is analyzed according to the influencing factors such as operation period characteristics,passenger flow and land use types around stations.Based on the SARIMA(autoregressive integrated moving average)model and the SVM(support vector machine)theory,a prediction model of short-term passenger flow is constructed to capture the periodicity and local non-linearity characteristics of short-term passenger flow.To improve the reliability of the of short-term passenger flow,GARCH(autoregressive conditional heteroskedasticity)model is introduced to construct the uncertainty prediction model.Through actual case analysis,the testified results show that the SARIMA-OLSVM model has superior prediction performance for the strong periodical and stable passenger flow,and the GARCH model is more accurate and reliable for the short-term passenger flow uncertainty prediction.