基于数据替补修正的高速铁路日常客运量VMD-GA-BP预测方法  

A VMD-GA-BP Method for Predicting Non-Holiday Passenger Flow of High Speed Railway Based on Data Replacement Correction

作　　者：史峰[1] 杨星琪 胡心磊 徐光明[1] 武润发 SHI Feng;YANG Xingqi;HU Xinlei;XU Guangming;WU Runfa(School of Traffic and Transportation Engineering,Central South University,Changsha Hunan410075,China)

机构地区：[1]中南大学交通运输工程学院,湖南长沙410075

出　　处：《中国铁道科学》2019年第3期129-136,共8页China Railway Science

基　　金：国家自然科学基金资助项目(71701216;U1334207);湖南省研究生科研创新项目(CX2018B098);中南大学中央高校基本科研业务费专项资金资助项目(2018ZZTS027)

摘　　要：针对高速铁路日常客运量预测问题,提出消除节假日因素影响的数据替补修正法和融合变分模态分解(VMD)、遗传算法(GA)和BP神经网络的日常客运量VMD-GA-BP预测方法。数据替补修正法是根据日常客运量超常波动判定阈值识别节假日延续期,采用VMD-GA-BP预测方法得到预测值,用该预测值替换节假日延续期内的客运量。VMD-GA-BP预测方法首先采用VMD对被替换数据之前的数据序列进行分解,得到不同频率的模态分量;其次通过GA优化初始权值和阈值的BP神经网络对各模态分量分别预测;然后重构各模态分量的预测值,用预测值替换节假日延续期内的客运量,得到修正数据序列,据此预测得到高速铁路日常客运量。实例应用表明,VMD-GA-BP的预测误差远低于BP,EMD-GA-BP,SVR,EMD-BP等方法,且基于修正数据序列的预测误差明显低于基于原始数据序列。可见,VMD-GA-BP预测方法精度较高。In order to solve the problem in predicting the non-holiday passenger flow of high speed railway,the data replacement correction method to eliminate the influence of holiday factors as well as the VMD-GA-BP prediction method for non-holiday passenger flow to integrate the variational mode decomposition(VMD),genetic algorithm(GA)and BP neural network are put forward.The data replacement correction method is to identify the holiday extension period according to the threshold value of the supernormal fluctuation of non-holiday passenger flow.The predicted value is obtained by using the VMD-GA-BP prediction method,and the predicted value is used to replace the passenger flow in the holiday extension period.The VMD-GA-BP prediction method firstly decomposes the data sequence before the replaced data by VMD,and obtains the modal components of different frequencies.Secondly,the BP neural network is used to optimize the initial weights and thresholds of GA to predict each modal component respectively.Then,the predicted values of all modal components are reconstructed,and the passenger flow during holiday extension period is replaced by the predicted value.The modified data sequence is obtained,accordingly,the non-holiday passenger flow of high speed railway is predicted.Examples show that the prediction error of VMD-GA-BP is much lower than that of BP,EMD-GA-BP,SVR,EMD-BP and so on.Moreover,the prediction error based on the modified data sequence is significantly lower than that based on the original data sequence.It can be seen that the accuracy of VMD-GA-BP prediction method is higher.

关 键 词：高速铁路 日常客运量预测 预测精度 变分模态分解 BP神经网络 遗传算法 

分 类 号：U293.13[交通运输工程—交通运输规划与管理]