基于神经网络的小时间粒度交通流预测模型

为解决传统车队离散模型基于概率分布假设和现有交通流预测时间粒度过大不能应用于自适应信号配时优化等问题.在车队离散模型的建模思路上,先分析了下游交叉口车辆到达与上游交叉口车辆离去之间的关系,基于此构建了基于神经网络的小时间粒度交通流预测模型.该模型以上游交叉口离去流量分布为输入,下游交叉口到达流量分布为输出,时间粒度为5 s.最后,通过实际调查数据标定模型参数并应用模型预测下游交叉口到达流量.结果表明,与Robertson模型相比,本文模型预测结果能够更好地反映交通流的变化特征,平均预测误差减少了8.3%.成果可用于信号配时优化.

Traffic Flow Prediction Model Based on Neural Network in Small Time Granularity

The traditional platoon dispersion model is based on the hypothesis of probability distribution, and the time granularity of the existing traffic flow prediction is too big to be applied to the adaptive signal timing optimization. In order to solve these problems, from the view of the platoon dispersion model, the relationship between vehicle arrival at the downstream intersection and vehicle departure from the upstream intersection is analyzed, then, a traffic flow prediction model based on neural network in small time granularity is proposed. The departure flow at the upstream is taking as the input and the arrival flow at the downstream intersection is taking as the output in this model, which has the time granularity of 5 s. Finally, the proposed model parameters are calibrated by the actual survey data, and this model is applied to predict the arrival flow of the downstream intersection. The results show that the proposed model can better reflects the fluctuant characteristics of traffic flow compared with Robertson model, and the prediction error is reduced by 8.3%. As a result, this result provides theoretical support for signal timing optimization.