道路交通流多步预测流量控制建模研究

Study on Multi-Step Traffic Flow Prediction and Control Model

对道路交通流多步预测流量控制进行研究,解决了传统模型只反映交通流的部分信息,且预测精度较低的问题。结合交通流量数据的周期性特征,提出了一种根据PCA周期分量提取的交通流多步预测模型。上述模型对历史交通流采用循环移位方法构造出样本矩阵,对其进行主分量分析和重构,提取出占交通流绝大部分能量的周期分量,对交通流的非周期分量进行奇异值分解和滤波,保障了交通流量多步预测的精度。上述模型不需要考虑路网复杂的拓扑结构和时空关系,仅根据交通流历史流量,即可实现多步预测。仿真结果表明,改进速路交通流进行有效的多步预测,且具有较高的预测精度。

In light of the periodic characteristics of traffic flow, a multi-step traffic flow prediction model based on PCA periodic components extraction was proposed. Utilizing cyclic shift method, the proposed model constructs the sample matrix, meanwhile carries out principal component analysis and reconstruction to extract the periodic components which possesses most of the energy of traffic flow. Moreover, it employs singular value decomposition to filtrate the aperiodic component of traffic flow to ensure the accuracy of multi-step. Without considering about the complex topological structure and space-time relation of highway network, the proposed model can realize muhi-step prediction only according to the historical traffic flow. The simulation results indicate that the proposed model is valid in multi-step traffic flow prediction with better prediction accuracy.