基于DAM和CNN-LSTM的辅助动力装置性能参数预测模型

Prediction model of auxiliary power unit performance parameter based on DAM and CNN-LSTM

摘要:

针对辅助动力装置(auxiliary power unit, APU)性能参数难以准确预测的问题，提出一种基于特征与时序的双阶段注意力机制(dual-stage attention mechanism,DAM)和卷积神经网络(CNN)-长短期记忆网络(LSTM)的混合模型。所提的方法在特征提取阶段加入了通道注意力机制(Channel attention mechanism ,CAM)；输出阶段加入了时序注意力机制(Temporal attention mechanism ,TAM)，加强了CNN对重要特征的提取能力和历史关键信息对预测输出的影响，并利用改进的粒子群算法对模型关键参数寻优，提高预测精度。实验结果表明，所提出的新方法在多变量输入和多步长的APU排气温度(exhaust gas temperature，EGT)预测中均取得了很好的效果，预测精度大幅提高。

Abstract:

In order to solve the handicap that the performance parameters of auxiliary power unit (APU) are hard to accurately forecast, the study proposed a hybrid model of dual-stage attention mechanism (DAM) and CNN (convolutional neural network)-LSTM (long short-term memory) based on characteristic and temporal. The method introduced a channel attention mechanism in the characteristic extraction phase and a temporal attention mechanism in the output phase. Enhanced the CNN"s ability to extract important characteristics. The influence of key information from historical moments on the prediction output was enhanced simultaneously. For the purpose of improving the prediction accuracy, the improved pso algorithm was used to optimize the hyperparameters of the LSTM network. The proposed new method has extremely low errors in the prediction of APU exhaust gas temperature (EGT) with multi-variable input and multi-step length, and the experimental results show that the prediction accuracy is significantly improved.