城市道路网节点短时段交通量预测模型研究

准确的短时段交通量预测在良好的道路交通管理中将越来越成为至关紧要的一个步骤。本文应用BP学习算法及进行误差校正的SPDS算法,建立了基于BP网络的城市道路网节点短时段交通量预测模型。并依据哈尔滨市省政府交叉口2001年6月15日15min间隔的交通量调查数据,对中宣街进行了分时段的交通量预测。本文还对交通量神经网络预测模型提出了五种输入层方案,针对不同输入层方案,采用试算法确定最佳隐层单元数,根据各方案的训练时间和误差进行评价,确定了理想的交通量神经网络预测模型,并分析了输入层单元和隐层单元分别与训练时间和误差的关系。最后,采用确定的交通量预测模型进行预测,预测结果证明了本模型在较短时间内具有较高的预测精度。

RESEARCH ON SHORT-TERM TRAFFIC FLOW FORECASTING MODEL OF NODES IN URBAN ROAD NETWOEK

Accurate short-term traffic flow forecasting is becoming a crucial step in better road traffic management. Based on BP algorithm and SPDS algorithm,a short-term traffic flow forecasting model of nodes in urban road network is put forward. And a forecasting of Zhongxuan Street is done based on investigation data of 15 min interval in province government intersection in Harbin on June 15,2001. Five architecture projects as input units of this forecasting model are brought forward. By the test-calculate means the optimal hidden units are confirmed. And an optimal traffic flow forecasting model is evaluated after comparing training time and errors. At the same time, input units relating with training time and errors, and hidden units relating with training time and errors are analysed. Finally,it is validated that the optimal traffic flow forecasting model has certain precision in a short time.