基于混沌理论的短期交通流量多步预测

Multi-step Short-term Traffic Flow Prediction Based on Chaotic Theory

短期交通流量预测是智能交通系统的核心研究内容之一。针对城市交通流具有的混沌特性,提出1种具有较高精度的短期交通流量多步预测方法,以支持交通控制和交通流诱导。利用最大Lyapunov指数方法判别交通流量时间序列的混沌特性,对交通流量时间序列进行相空间重构,并在此基础上结合加权一阶局域方法设计了基于混沌理论的交通流量多步预测算法。将此方法运用于实际道路交通流量的多步预测,比较多步预测值与实际流量值,其平均绝对百分比误差为3.33%,平均绝对误差为9.05/[pcu·(5min)-1],均方根误差为10.36/[pcu·(5min)-1]。应用结果表明,该预测方法具有较高的精度。

Short-term traffic flow prediction is one of the key issues in intelligent transportation system studies. Through considering the chaotic characteristics of urban traffic flow,a method with high accuracy of multi-step shortterm traffic flow prediction is proposed to support the traffic control and route guidance.The chaotic characteristics of traffic flow time series can be distinguished by using the largest Lyapunov exponent method.Then,the phase space reconstruction of traffic flow time series is conducted.On this basis,multi-step traffic flow prediction algorithm is designed based on chaotic theory by further using adding-weight one-rank local-region method.The proposed traffic flow prediction algorithm is applied to a case study,and the multi-step predicted values are compared with the real traffic flow value.The results show that the proposed algorithm has a high accuracy with the mean absolute percentage error 3.33%,the mean absolute error 9.05,and the root mean square error 10.36.