基于二次分解和MVO-TCN的城轨站点客流预测研究  

Research on Passenger Flow Prediction at Urban Rail Stations Based on Quadratic Decomposition and MVO-TCN

作　　者：周建国[1] 蔡晨昊 ZHOU Jian-guo;CAI Chen-hao(Department of Economics and Management,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学经济管理系,保定071003

出　　处：《武汉理工大学学报》2024年第9期100-108,共9页Journal of Wuhan University of Technology

摘　　要：为提升城轨站点客流预测精度,提出一种基于变分模态分解(VMD)、经验小波变换(EWT)和多元宇宙优化算法(MVO)优化的时间卷积神经网络(TCN)的复合预测模型。使用VMD对城轨站点客流数据进行初步分解,生成若干本征模态函数(IMF)。随后,计算每个IMF的样本熵,并利用K-means算法根据样本熵值将IMFs聚类为复合分量,对高频复合分量进一步应用EWT进行细化分解,以减少波动性强的城轨站点客流数据的分解残留。通过MVO算法对TCN模型的参数进行优化,以强化模型在各个分量上的预测能力。最后,将各分量的预测结果进行重构,得到最终的客流预测值。实验结果表明,文中所提出的VMD-EWT-MVO-TCN模型通过二次分解改善了分解残留问题,结合对TCN参数的寻优提升了城轨站点客流预测精度,RMSE和MAE值分别为14.936 5和5.789 3,相较TCN模型提升了45.46%和50.28%。该模型能够为城轨站点客流管理提供新的技术支持和决策参考。In order to improve the accuracy of passenger flow prediction at urban rail stations,a composite prediction model based on the Variational Mode Decomposition(VMD),Empirical Wavelet Transform(EWT),Multi-verse Universe Optimization(MVO)algorithm,and Temporal Convolutional Neural Network(TCN)is proposed.First,VMD is used to initially decompose the passenger flow data at urban rail stations to generate a number of intrinsic modal functions(IMFs).Subsequently,the sample entropy of each IMF is calculated and the K-means algorithm is used to cluster the IMFs into composite components based on the sample entropy values,and the EWT is further applied to the high-frequency composite components for a refined decomposition in order to minimize the decomposition residuals of the highly fluctuating passenger flow data of the urban rail stations.The parameters of the TCN model are optimized by the MVO algorithm to strengthen the prediction ability of the model on each component.Finally,the prediction results of each component are reconstructed to obtain the final passenger flow prediction value.The experimental results show that the proposed VMD-EWT-MVO-TCN model improves the decomposition residual problem through the quadratic decomposition and combines with the optimization of the TCN parameters to enhance the accuracy of passenger flow prediction at urban rail stations,with the RMSE and MAE values of 14.9365 and 5.7893,respectively,which are improved by 45.46%and 50.28%compared with the TCN model.The model can provide new technical support and decision-making reference for passenger flow management in urban rail stations.

关 键 词：城轨站点客流预测 预测分析 二次模态分解 多元宇宙优化算法 时间卷积神经网络 

分 类 号：U121[交通运输工程]