为解决长短期记忆网络（LSTM）和门控循环单元神经网络（GRU）在捕捉长期依赖关系上的不足以及估算精度较低的问题,该文提出基于Informer算法的燃料电池寿命估算方法,旨在提高估算的准确性和效率。该方法采用加权平均法和皮尔逊系数法对数据进行平滑处理,以增强数据的趋势性并减少噪声影响。结合Informer模型的多尺度信息融合和长期依赖建模能力,设计了一个能够实现燃料电池寿命在线估算的寿命估算框架。随后设计3组实验与传统的LSTM和GRU模型进行比较,当训练集占比80%时,Informer模型UMAE、URMSE、UMAPE均最小,估算精度高于LSTM和GRU模型。说明Informer模型在长时间序列估算方面表现出色,为燃料电池寿命估算提供可靠的依据。
Abstract
In the present, common methods for estimating fuel cell life include long short-term memory （LSTM） and gated recurrent unit （GRU） neural networks. However, these methods have drawbacks in capturing long-term dependencies and ensuring accuracy. To address these issues, a fuel cell life estimation approach based on the Informer algorithm is proposed, aiming to enhance both accuracy and efficiency. Data smoothing techniques such as the weighted average method and the Pearson coefficient method are employed to reduce noise and improve data trend. By leveraging the multi-scale information fusion and long-term dependency modeling capabilities of the Informer model, a life estimation framework is crafted to enable online fuel cell life estimation. Subsequently, three sets of experiments are conducted to compare with traditional LSTM and GRU models. When the training set constitutes 80%, the Informer model exhibits the smallest mean absolute error （MAE）, root mean square error （RMSE）, and mean absolute percentage error （MAPE）, indicating higher estimation accuracy compared to LSTM and GRU models. These findings demonstrate the outstanding performance of the Informer model in long-term series estimation, thereby establishing a dependable foundation for fuel cell life estimation.
关键词
神经网络 /
燃料电池 /
并行处理 /
寿命估算 /
多头概率稀疏自注意力机制
Key words
neural networks /
fuel cells /
parallel processing /
life estimation /
ProbSparse self-attention mechanism
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 陈家城, 周苏. 大功率质子交换膜燃料电池建模及仿真[J]. 太阳能学报, 2024, 45(3): 290-297.
CHEN J C, ZHOU S.Modeling and simulation of high-power proton exchange membrane fuel cells[J]. Acta energiae solaris sinica, 2024, 45(3): 290-297.
[2] 刘阳, 陈奔. 车用PEMFC氢气系统建模及其排放特性研究[J]. 太阳能学报, 2023, 44(2): 260-268.
LIU Y, CHEN B.Modeling and emission characteristics study of pemfc hydrogen system for vehicles[J]. Acta energiae solaris sinica, 2023, 44(2): 260-268.
[3] JOUIN M, GOURIVEAU R, HISSEL D, et al.Prognostics and health management of PEMFC-state of the art and remaining challenges[J]. International journal of hydrogen energy, 2013, 38(35): 15307-15317.
[4] 马睿, 党翰斌, 张钰奇, 等. 质子交换膜燃料电池系统故障机理分析及诊断方法研究综述[J]. 中国电机工程学报, 2024, 44(1): 407-427.
MA R, DANG H B, ZHANG Y Q, et al.A review on failure mechanism analysis and diagnosis for proton exchange membrane fuel cell systems[J]. Proceedings of the CSEE, 2024, 44(1): 407-427.
[5] ZHU L, CHEN J.Prognostics of PEM fuel cells based on Gaussian process state space models[J]. Energy, 2018, 149: 63-73.
[6] WANG Y P, WANG K, WANG B W, et al.A data-driven approach to lifespan prediction for vehicle fuel cell systems[J]. IEEE transactions on transportation electrification, 2023, 9(4): 5049-5060.
[7] CHEN K, LAGHROUCHE S, DJERDIR A.Proton exchange membrane fuel cell degradation and remaining useful life prediction based on artificial neural network[C]//2018 7th International Conference on Renewable Energy Research and Applications (ICRERA). Paris, France, 2018: 407-411.
[8] CHENG Y J, ZERHOUNI N, LU C.A hybrid remaining useful life prognostic method for proton exchange membrane fuel cell[J]. International journal of hydrogen energy, 2018, 43(27): 12314-12327.
[9] 薛阳, 燕宇铖, 贾巍, 等. 基于改进灰狼算法优化长短期记忆网络的光伏功率预测[J]. 太阳能学报, 2023, 44(7): 207-213.
XUE Y, YAN Y C, JIA W, et al.Photovoltaic power prediction model based on IGWO-LSTM[J]. Acta energiae solaris sinica, 2023, 44(7): 207-213.
[10] LIU J W, LI Q, CHEN W R, et al.Remaining useful life prediction of PEMFC based on long short-term memory recurrent neural networks[J]. International journal of hydrogen energy, 2019, 44(11): 5470-5480.
[11] 莫易敏, 余自豪, 叶鹏, 等. 基于迁移学习与GRU神经网络结合的锂电池SOH估计[J]. 太阳能学报, 2024, 45(3): 233-239.
MO Y M, YU Z H, YE P, et al.Lithium battery SOH estimation method based on combination of transfer learning and GRU neural network[J]. Acta energiae solaris sinica, 2024, 45(3): 233-239.
[12] BENGIO Y, SIMARD P, FRASCONI P.Learning long-term dependencies with gradient descent is difficult[J]. IEEE transactions on neural networks, 1994, 5(2): 157-166.
[13] 何滢婕, 刘月峰, 边浩东, 等. 基于Informer的电池荷电状态估算及其稀疏优化方法[J]. 电子学报, 2023, 51(1): 50-56.
HE Y J, LIU Y F, BIAN H D, et al.State-of-charge estimation of lithium-ion battery based on Informer and its sparse optimization method[J]. Acta electronica sinica, 2023, 51(1): 50-56.
[14] LI Y Y, LIU Y C, WILLIAMSON D S.On loss functions for deep-learning based T60 estimation[C]//ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Toronto, ON, Canada, 2021: 486-490.
基金
安徽省自然科学基金（2308085ME180）; 广东恒翼能科技股份有限公司合作项目（W2023JSFW0479）; 高等学校学科创新引智计划（“111”计划）资助项目（BP0719039）
PDF(4506 KB)
