基于EMD-LSTM模型的APU排气温度预测
APU exhaust temperature prediction based on EMD-LSTM model
王晓燕 1白贤明 2宋辞 1毛子荐1
作者信息
- 1. 沈阳航空航天大学经济与管理学院,沈阳 110136;辽宁省飞机火爆防控及可靠性适航技术重点实验室,沈阳 110136
- 2. 辽宁省飞机火爆防控及可靠性适航技术重点实验室,沈阳 110136;沈阳航空航天大学安全工程学院,沈阳 110136
- 折叠
摘要
为了提高排气温度(EGT)的预测精度需要减少数据的复杂性.提出一种经验模态分解(EMD)和长短期记忆神经网络(LSTM)组合方法来预测EGT.将具有时间序列特征的EGT数据,利用EMD分解成含有相同特征的本征模态函数(IMF)和残差(RES);利用LSTM模型对分量进行预测;将所有分量预测出来的结果进行叠加得到EGT的预测值.并对EMD-LSTM模型与单一的LSTM模型的预测结果进行对比分析.结果表明:前者比后者的方均根误差和平均相对误差分别降低了 35％和42％.说明此模型在预测APU的EGT值上具有更好的预测精度.
Abstract
To improve the prediction accuracy of exhaust gas temperature(EGT),the complexity of the data should be reduced.A combined empirical modal decomposition(EMD)and long short-term memory neural network(LSTM)method was proposed to predict EGT.First,EGT data with time series characteristics were decomposed into intrinsic mode function(IMF)and residual(RES)containing the same characteristics using EMD;the components were predicted using LSTM model;and the results predicted from all components were superimposed to obtain the predicted values of EGT.The prediction results of EMD-LSTM model and single LSTM model were compared and analyzed.The results showed that the former had 35％and 42％lower root mean square error and average relative error than the latter.It indicated that this model has better prediction accuracy in predicting the EGT value of APU.
关键词
排气温度/预测精度/经验模态分解/长短期记忆神经网络/本征模态函数Key words
exhaust temperature/prediction accuracy/empirical modal decomposition/long and short-term memory neural network/intrinsic mode function引用本文复制引用
出版年
2024
NETL
NSTL
万方数据