基于双向长短期记忆网络和注意力机制的质子交换膜燃料电池衰退行为预测

doi:10.19595/j.cnki.1000-6753.tces.250149

Abstract
Figure/Table
References
Related Citation (15)

Download: PDF (8951 KB) HTML (1 KB)
Export: BibTeX | EndNote (RIS)

Abstract The lifetime and performance degradation of proton exchange membrane fuel cells (PEMFCs) are critical factors limiting their widespread application. At the same time, this issue represents one of the core research topics in the field of fuel cell technology. Accurately predicting the voltage degradation behavior of fuel cells not only aids in a deeper understanding of their degradation mechanisms but also facilitates the prediction and analysis of their operational states, thereby significantly enhancing system reliability and performance stability. An analysis of PEMFC aging data reveals that the aging process exhibits high nonlinearity and time dependency, making it challenging to comprehensively characterize the degradation process using traditional modeling methods. To address the difficulty of predicting PEMFC voltage degradation, this paper proposes a prediction model for fuel cell degradation behavior that integrates a bidirectional long short-term memory network (Bi-LSTM) with the attention mechanism.
The proposed model first selects features that reflect the degradation behavior of fuel cells as inputs. It utilizes the Bi-LSTM to extract long-term dependencies from the data and predict temporal sequence information, fully capturing the dynamic characteristics of the degradation behavior. Subsequently, the attention mechanism enhances the weight allocation of key features, ultimately yielding PEMFC degradation prediction results. During model training, to improve training efficiency, optimize prediction performance, and effectively avoid overfitting and convergence issues caused by improper learning rate settings, the paper employs techniques such as early stopping, optimal model checkpointing, and adaptive learning rate adjustment. This study uses the PEMFC aging dataset from Tongji University, with the first 60% of the data serving as the training set and the remaining 40% as the test set, to conduct model training and testing. Systematic comparative analyses with other classical neural network models (e.g., convolutional neural networks) are also performed.
Experimental results demonstrate that the proposed Bi-LSTM-Attention (Bi-LSTM-At) model exhibits outstanding prediction performance, achieving a root mean square error (RMSE) of only 0.006 1 V. The prediction accuracy improves by 23% compared to traditional neural network models, enabling precise prediction of PEMFC degradation behavior. Through multi-step prediction experiments, the proposed model has been verified to capture the overall voltage degradation trend even over extended periods, demonstrating strong long-term predictive capability. This indicates that the model achieves excellent forecasting performance and accuracy in long-term predictions. Moreover, the proposed model can accurately predict degradation behavior in both the early and late stages of operation, reflecting robust performance. This robust predictive ability provides effective technical support for fuel cell health monitoring and remaining useful life assessment. Based on the experimental results and comparisons with existing traditional algorithms, the following conclusions are drawn: (1) The proposed method converges rapidly and exhibits excellent fitting performance. (2) Compared with conventional Transformer and LSTM models, this method shows significant advantages in terms of prediction accuracy and overall model performance. (3) The introduction of the attention mechanism significantly enhances the prediction performance and accuracy of most models; compared to the Bi-LSTM model, incorporating the attention mechanism substantially reduces training time. (4) The method features a simple structure and is easy to implement, with strong real-time performance and stability, thereby achieving superior prediction accuracy for time-dependent data.

Key words： Proton exchange membrane fuel cell degradation behavior prediction bidirectional long short-term memory attention mechanism time-series analysis

Received: 21 January 2025

PACS:

TM614

	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors

Cite this article:

URL:

https://dgjsxb.ces-transaction.com/EN/10.19595/j.cnki.1000-6753.tces.250149 OR https://dgjsxb.ces-transaction.com/EN/Y2026/V41/I2/701

[1] 吴字强, 王伟, 黄亮, 等. 燃料电池稳态输出功率与效率的双目标优化[J]. 电源学报, 2019, 17(2): 32-39.
Wu Ziqiang, Wang Wei, Huang Liang, et al.Bi- objective optimization for steady-state output power and efficiency of fuel cell[J]. Journal of Power Supply, 2019, 17(2): 32-39.
[2] 王博斐, 肖浩哲, 李国豪, 等. 基于控制目标的氢-电混动系统能量管理策略综述[J]. 发电技术, 2023, 44(4): 452-464.
Wang Bofei, Xiao Haozhe, Li Guohao, et al.A review of energy management strategy for hydrogen- electricity hybrid power system based on control target[J]. Power Generation Technology, 2023, 44(4): 452-464.
[3] YangDuo, Wang Siyu, Liao Yuefeng, et al. An online energy management strategy for fuel cell vehicles based on fuzzy Q-learning and road condition recognition[J]. IEEE Transactions on Intelligent Transportation Systems, 2024, 25(9): 12120-12130.
[4] 袁铁江, 计力, 田雪沁, 等. 考虑燃料电池汽车加氢负荷的电-氢系统协同优化运行[J]. 电力系统自动化, 2023, 47(5): 16-25.
Yuan Tiejiang, Ji Li, Tian Xueqin, et al.Synergistic optimal operation of electricity-hydrogen systems considering hydrogen refueling loads for fuel cell vehicles[J]. Automation of Electric Power Systems, 2023, 47(5): 16-25.
[5] 徐冰, 张敏, 杨文涛, 等. 质子交换膜燃料电池基于老化特征的剩余使用寿命混合预测方法[J]. 电工技术学报, 2025, 40(17): 5703-5714.
Xu Bing, Zhang Min, Yang Wentao, et al.PEMFC Hybrid prediction method for remaining useful life based on aging characteristics[J]. Transactions of China Electrotechnical Society, 2025, 40(17): 5703-5714.
[6] 任洲洋, 王皓, 李文沅, 等. 基于氢能设备多状态模型的电氢区域综合能源系统可靠性评估[J]. 电工技术学报, 2023, 38(24): 6744-6759.
Ren Zhouyang, Wang Hao, Li Wenyuan, et al.Reliability evaluation of electricity-hydrogen regional integrated energy systems based on the multi-state models of hydrogen energy equipment[J]. Trans- actions of China Electrotechnical Society, 2023, 38(24): 6744-6759.
[7] 赵洪山, 潘思潮, 马利波, 等. 基于分类回放双延迟贝叶斯深度确定性策略梯度的燃料电池温度控制[J]. 电工技术学报, 2024, 39(13): 4240-4256.
Zhao Hongshan, Pan Sichao, Ma Libo, et al.Control of fuel cell temperature based on classified replay twin delayed Bayesian deep deterministic policy gradient[J]. Transactions of China Electrotechnical Society, 2024, 39(13): 4240-4256.
[8] WangYujie, Yang Xingliang, Sun Zhengdong, et al. A systematic review of system modeling and control strategy of proton exchange membrane fuel cell[J]. Energy Reviews, 2024, 3(1): 100054.
[9] Flick S, Schwager M, McCarthy E, et al. Designed experiments to characterize PEMFC material proper- ties and performance[J]. Applied Energy, 2014, 129: 135-146.
[10] Chen Hong, Zhan Zhigang, Jiang Panxing, et al.Whole life cycle performance degradation test and RUL prediction research of fuel cell MEA[J]. Applied Energy, 2022, 310: 118556.
[11] Bressel M, Hilairet M, Hissel D, et al.Model-based aging tolerant control with power loss prediction of Proton Exchange Membrane Fuel Cell[J]. Interna- tional Journal of Hydrogen Energy, 2020, 45(19): 11242-11254.
[12] Liu Hao, Chen Jian, Zhu Chuyan, et al.Prognostics of proton exchange membrane fuel cells using a model- based method[J]. IFAC-PapersOnLine, 2017, 50(1): 4757-4762.
[13] Bressel M, Hilairet M, Hissel D, et al.Extended Kalman filter for prognostic of proton exchange memb- rane fuel cell[J]. Applied Energy, 2016, 164: 220-227.
[14] Ao Yunjin, Laghrouche S, Depernet D, et al.Proton exchange membrane fuel cell prognosis based on frequency-domain Kalman filter[J]. IEEE Trans- actions on Transportation Electrification, 2021, 7(4): 2332-2343.
[15] 方德宇, 楚潇, 刘涛, 等. 基于数据-模型驱动的锂离子电池健康状态估计[J]. 电气工程学报, 2022, 17(4): 20-31.
Fang Deyu, Chu Xiao, Liu Tao, et al.Research on health assessment method of lithium-ion battery based on data-model hybrid drive[J]. Journal of Electrical Engineering, 2022, 17(4): 20-31.
[16] Liu Hao, Chen Jian, Hissel D, et al.Remaining useful life estimation for proton exchange membrane fuel cells using a hybrid method[J]. Applied Energy, 2019, 237: 910-919.
[17] Wu Hangyu, Wang Wei, Li Yang, et al.Hybrid physics-based and data-driven prognostic for PEM fuel cells considering voltage recovery[J]. IEEE Trans- actions on Energy Conversion, 2024, 39(1): 601-612.
[18] Pan Rui, Yang Duo, Wang Yujie, et al.Performance degradation prediction of proton exchange membrane fuel cell using a hybrid prognostic approach[J]. International Journal of Hydrogen Energy, 2020, 45(55): 30994-31008.
[19] 丁恒, 黄凯, 田海建. 基于VMD和ISSA-ELM的锂离子电池剩余使用寿命预测[J]. 电源学报, 2024, 22(6): 188-198.
Ding Heng, Huang Kai, Tian Haijian.Prediction of remaining useful life of lithium-ion battery based on VMD and ISSA-ELM[J]. Journal of Power Supply. 2024, 22(6): 188-198.
[20] Liu Jiawei, Li Qi, Chen Weirong, 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.
[21] ZhouNing, Cui Benyu, Zhou Jianxin. Transformer- based prediction of the RUL of PEMFC[C]//2022 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS),Guangzhou, China, 2022: 345-348.
[22] Shin S, Kim J, Lee S, et al.Multimodal data-driven prediction of PEMFC performance and process condi- tions using deep learning[J]. IEEE Access, 2024, 12: 168030-168042.
[23] Yu Yang, Yu Qinghua, Luo Runsen, et al.A predictive framework for PEMFC dynamic load performance degradation based on feature parameter analysis[J]. International Journal of Hydrogen Energy, 2024, 71: 1090-1103.
[24] 汪建锋, 王荣杰, 林安辉, 等. 质子交换膜燃料电池退化预测方法[J]. 电工技术学报, 2024, 39(11): 3367-3378.
Wang Jianfeng, Wang Rongjie, Lin Anhui, etal. Degradation prediction method of proton exchange membrane fuel cell[J]. Transactions of China Electro- technical Society, 2024, 39(11): 3367-3378.
[25] Zuo Jian, Lv Hong, Zhou Daming, et al.Long-term dynamic durability test datasets for single proton exchange membrane fuel cell[J]. Data in Brief, 2021, 35: 106775.
[26] Bloom I, Walker L K, Basco J K, et al.A comparison of fuel cell testing protocols—a case study: protocols used by the U.S. Department of Energy, European Union, International Electrotechnical Commission/Fuel Cell Testing and Standardization Network, and Fuel Cell Technical Team[J]. Journal of Power Sources, 2013, 243: 451-457.
[27] Xie Wenlong, Guo Peng, Gao Xinlei.Elucidating the rate limitation of lithium-ion batteries under different charging conditions through polarization analysis[J]. Journal of Energy Storage, 2024, 82: 110554.
[28] Zuo Jian, Lü Hong, Zhou Daming, et al.Deep learning based prognostic framework towards proton exchange membrane fuel cell for automotive applica- tion[J]. Applied Energy, 2021, 281: 115937.
[29] 马飞越, 李澳, 吴诚威, 等. 基于LSTM-MIV神经网络的SF₆断路器触头电寿命预测[J]. 高压电器, 2024, 60(2): 69-77.
Ma Feiyue, Li Ao, Wu Chengwei, et al.Electrical life prediction for contact of SF₆ circuit breaker based on LSTM-MIV neural network[J]. High Voltage Apparatus, 2024, 60(2): 69-77.