摘要
准确的销量预测是新能源汽车企业制定科学合理的生产计划的主要依据。针对新能源汽车销量非线性、非平稳的特征,运用融合变分模态分解(VMD)方法和长短期记忆网络(LSTM)模型进行新能源汽车销量预测。首先,利用VMD方法对原始数据进行分解处理,获得本征模态分量与趋势项;其次,以模态分量和趋势项作为新的时间序列,构建各时间序列的LSTM模型;最后,重构各模型的预测值获得新能源汽车销量的预测值。研究结果表明:与BP神经网络模型、LSTM模型、融合VMD和BP神经网络模型对比,融合VMD和LSTM模型的预测精度更高,该模型在新能源汽车销量预测中具有可行性。
Accurate sales forecasting is the main basis for new energy vehicle enterprises to formulate scientific and reasonable production plans.To address the nonlinear and non-stationary problem of new energy vehicle sales,a fusion of variational modal decomposition(VMD)method and long short-term memory(LSTM)network prediction model were used for new energy vehicle sales forecasting.The VMD method was first used to decompose the original data to obtain the intrinsic mode functions and trend terms;then the LSTM model for each time series was constructed using the intrinsic mode functions and trend terms as the new time series;finally the forecast values of each model were reconstructed to obtain the forecast values of new energy vehicle sales.The results showed that the forecast accuracy of the fused VMD and LSTM model is higher compared with the BP neural network model,LSTM model and the fused VMD and BP neural network model,and the model is feasible in the prediction of new energy vehicle sales.
作者
卢志平
玉晓晶
陆成裕
LU Zhiping;YU Xiaojing;LU Chengyu(Department of Economics and Management,Guangxi University of Science and Technology,Liuzhou 545006,China;不详)
出处
《武汉理工大学学报（信息与管理工程版）》
CAS
2023年第4期546-551,共6页
Journal of Wuhan University of Technology：Information & Management Engineering
基金
国家社会科学基金项目(18XGL006).
关键词
新能源汽车
销量预测
长短期记忆神经网络
变分模态分解
时间序列
new energy vehicles
sales forecast
long short-term memory neural network
variational mode decomposition
time series
