基于收费数据的高速公路站间旅行时间预测

针对高速公路断面检测数据密度不足现状，采用收费数据预测收费站间车辆旅行时间。首先，研究收费数据实时修正处理方法，改进平均旅行时间计算模型；其次，引入分段线性插值方法构建卡尔曼滤波模型，以减小卡尔曼滤波线性化产生的模型误差问题；接着，依据旅行时间预测业务逻辑开发应用系统，实时主动预测高速公路站间旅行时间。示范路段应用表明，插值后预测算法在正常、事故、小长假三种交通流状态下所有周期平均相对误差控制在10%内，事故周期平均相对误差控制在13%内。插值后算法预测精度有效提高，可为高速公路公众出行提供时间参考。

Highway Travel Time Prediction Between Stations Based on Toll Ticket Data

Considering the less section inspection data, toll ticket data are used to predict travel time between highway toll stations. First, a processing method that can revise the toll ticket data in time was researched, and also improved the average travel time calculation model. Second, in order to decrease the model deviation caused by Kalman filter model linearization, piecewise linear interpolation method was introduced to build the Kalman filter model. Last, the application system was developed according to the travel time prediction business logic, the system can accurate predict travel time between highway toll stations in time. Actual road application shows that the interpolation algorithm can improve travel time prediction accuracy compared to the conventional Kalman filter method under the normal, accident and holiday traffic flow. The relative error of all prediction periods is less than 10%, and the relative error of accident prediction periods is less than 13%. The prediction accuracy of interpolation algorithm is improved effectively, and can provide an effective time reference for public in highway.