基于LSTM的航空公司能耗序列预测

为提高航空公司能耗的预测精度,针对能耗数据的复杂非线性时序特性,提出一种基于长短时记忆网络(LSTM)的时间窗滑动航空公司能耗预估模型。该方法对能耗时序数据进行预处理,消除能耗时序数据的季节性趋势;依据滑动时间窗将数据转换成监督型数据,构建基于LSTM的模型来实现航空公司能耗预测,并利用网格搜索算法进行参数优选。实验结果表明,该模型预测精度优于传统ARMA模型、SVR模型,验证了其可行性。

AIRLINE ENERGY CONSUMPTION SEQUENCE PREDICTION BASED ON LSTM

In order to improve the prediction accuracy of airline energy consumption,aiming at the complex nonlinear timing characteristics of energy consumption data,we proposed a time window sliding airline energy consumption estimation model based on LSTM.Energy consumption time series data were preprocessed to eliminate the seasonal trend of energy consumption time series data.Then,data were converted into supervised data according to sliding time window.We constructed an LSTM-based model to realize airline energy consumption prediction,and parameters were optimized by grid search algorithm.The experimental results show that the prediction accuracy of this model is better than that of traditional ARMA model and SVR model,which verifies the feasibility of the model.