Volume 41, Number 3, June 2023
|Page(s)||464 - 470|
|Published online||01 August 2023|
Degradation prediction method of PEMFC based on deep learning hybrid model integrating ARIMA and LSTM
Fuel cell involves many disciplines such as electricity, mechanics, electrochemistry, and thermodynamics, and its performance degradation process is complex, involving multi-physics, multi-scale, multi-parts, and multi-factors. Thus, it is difficult for a single model to capture all kinds of characteristics of fuel cell simultaneously in degradation prediction. To ensure the prediction accuracy while better fitting the data linearly and nonlinearly, a prediction model of ARIMA combined with LSTM neural network is proposed in this study. The prediction results with residuals are used as features for LSTM prediction work after first predicting the voltage decay data by ARIMA and LSTM. Comparing the hybrid model with the single ARIMA model and the NAR model with support vector regression learning, it is found that the hybrid model performs better in terms of prediction accuracy and prediction performance.
燃料电池涉及电学、机械、电化学、热力学等诸多学科, 其性能衰减过程复杂, 涉及多物理、多尺度、多部件、多因素, 单一模型在燃料电池老化预测中难以同时对其各类特征进行捕获。为在确保预测精度的同时更好地对数据进行线性和非线性拟合, 提出一种差分移动平均自回归结合长短期记忆神经网络的预测模型。通过ARIMA(autoregressive integrated moving average model)与LSTM(long short-term memory)对电压衰退数据线性及非线性部分进行预测后, 将预测结果与残差作为特征用于LSTM预测工作。将混合模型与单一ARIMA模型、NAR模型、支持向量回归学习对比发现, 混合模型在预测精确度和预测性能方面均有较好表现。
Key words: fuel cell / degradation prediction / LSTM / ARIMA
关键字 : 燃料电池 / 老化预测 / 长短期记忆神经网络 / 差分移动平均自回归
© 2023 Journal of Northwestern Polytechnical University. All rights reserved.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.