基于乘积耦合Volterra模型的短时交通流预测

摘要

An adaptive third-order Volterra filter for short-time traffic flow prediction was proposed, which was based on intrinsic nonlinearity and deterministic mechanism of chaotic time series and nonlinear expression of Volterra series. Concerning the problem that the coefficients of Volterra filter exponentially increase with its order, an approximate multiplication-coupled structure for Volterra filter was studied. At first, through properly choosing time delay and embedding dimension using mutual information method and false nearest neighbor method, respectively, the largest Lyapunov exponent was estimated by applying small data sets method so as to validate that chaos existed in traffic flow series. Then, the approximate multiplication-coupled structure for third-order Volterra filter, of which coefficients were adaptively adjusted by using an improved nonlinear Normalized Least Mean Square ( NLMS) algorithm, was employed to reduce computational complexity. Finally, by applying the proposed method to the real measured traffic flow data, the experimental results show that chaos exists in traffic flow series and the proposed scheme can effectively predict traffic flow series and reduce model complexity.%基于混沌序列固有的非线性和确定性机制以及Volterra级数的非线性表征能力,提出一种短时交通流预测的三阶Voherra模型.针对Volterra模型随阶数增加复杂度以幂次方增加的问题,研究了该模型的乘积耦合近似实现结构.首先,采用互信息法和虚假邻点法选取时间延迟和嵌入维数,并采用小数据量法计算Lyapunov指数判定交通流是否具有混沌特性；然后,建立三阶Volterra滤波器的乘积耦合近似实现结构,并采用一种改进的非线性归一化最小均方(NLMS)算法实时调整模型系数；最后,对高速公路实测交通流的预测结果表明,交通流中存在混沌特征,应用构建的预测模型可有效地对交通流进行预测,且降低了模型的复杂性.