摘要
准确预测铁路货运量对铁路货运组织工作的开展极为重要,特别是短期(月、日)货运量数据直接关系到铁路各项运输计划的编制。人工神经网络模型因其强大的学习能力而被广泛运用于各领域的预测,其中的LSTM网络适合处理和预测铁路货运量这类间隔和延迟相对较长的时间序列。考虑不同时期货运数据的特点分别建立基于月货运量数据的LSTM多变量预测模型和基于日货运量数据的LSTM时间序列模型。基于广铁2010—2017年的货运量数据,运用所建模型预测各月和每日的货运发送量,并与ARIMA模型预测方法和BP神经网络方法的预测结果相比较。结果表明,LSTM网络预测效果更佳。
Accurate prediction of railway freight volume is important for railway freight transport organization.In particular,the short-term(monthly/daily)freight volume data is directly related to the preparation of railway transport plans.Artificial neural network(ANN)models have been widely used in predictions in various fields for their strong learning ability.Long Short-Term Memory(LSTM)network,one of them,is suitable for processing and forecasting the time series with long intervals and delays,such as railway freight volume.Considering the characteristics of freight data in different periods,this paper established an LSTM multivariate forecasting model based on monthly freight volume and an LSTM time series model based on daily freight volume.Based on the freight volume data of China Railway Guangzhou Bureau Group Co.,Ltd.from 2010 to 2017,two models were used to predict its monthly and daily freight shipments separately.The results were compared with the forecasting results of ARIMA model and BP neural network method,which show that LSTM network has better performance.
作者
程肇兰
张小强
梁越
CHENG Zhaolan;ZHANG Xiaoqiang;LIANG Yue(School of Transportation and Logistics, Southwest Jiaotong University, Chengdu 611756, China;National Engineering Laboratory of Application Technologyof Integrated Transportation Big Data, Chengdu 611756, China;National United Engineering Laboratory of Integrated and Intelligent Transportation, Chengdu 611756, China)
出处
《铁道学报》
EI
CAS
CSCD
北大核心
2020年第11期15-21,共7页
Journal of the China Railway Society
基金
国家铁路局科技开发项目(KF2019-101-B)
成都市科技局软科学项目(2017-RK00-00050-ZF)
广州铁路(集团)科技开发项目(2017K020)。
关键词
铁路运输
货运量预测
LSTM网络
时间序列
深度学习
railway transportation
freight volume forecasting
LSTM network
time series
deep learning
