基于小波和回声状态网络的交通流多步预测模型

针对交通流的含噪混沌特征,提出了一种基于小波回声状态网络的交通流多步预测模型。该模型利用小波多尺度分解方法,屏蔽了噪声成分对交通流动力学特性的干扰,同时提取了占有交通流绝大部分能量的混沌低频成分。在采用多路分量并行预测的方式下,充分发挥了回声状态网络对混沌低频分量的强大多步预测能力,从而保障了交通流多步预测的精度。对北京市西直门桥的实测交通流的预测结果表明:该模型的多步预测精度比传统的回声状态网络模型有了较大幅度的提升,在保证预测精度的前提下,最大可预测的步长也相应的增加。

Multi-step traffic flow prediction model based on wavelet and echo state network

In light of the noisy chaotic characteristics of traffic flow,a new multi-step traffic flow prediction model based on wavelet and echo state network was proposed.Utilizing multi-scale decomposition method of wavelet,the proposed model restricts the interference of noisy components to the dynamic behavior of the traffic flow;meanwhile it extracts the chaotic low-frequency component,which possesses most of the energy of traffic flow.In predicting the multi components concurrently,the strong prediction capacity of echo state network for chaotic low-frequency component was utilized effectively to ensure the accuracy of multi-step traffic flow prediction.The results of prediction of the real traffic flow in Xizhimen Bridge of Beijing show that the prediction accuracy is significantly improved by the proposed multi-step model in comparison with the traditional echo state network model.Under the condition of high prediction accuracy,the maximum predictable step is also increased by the proposed model.