在2022年世界燃料電池的市場規模估值8,643百萬美元，相對於2016年2,894百萬美元，預期成長17.36%，再生能源的利用成為世界能源主流。因此，具有低噪音、低操作溫度、零CO2、NOX排放、能源使用效率高與快速啟動的燃料電池就成為最具潛能的能源選擇。然而，膜電極組(Membrane Electrode Assembly, MEA)使用後會產生衰退現象，除了會影響燃料電池的表現性能和效率，亦會影響燃料電池的使用壽命。由於長時間老化衰退的測試非常耗時，本研究將致力於提出快速評估對膜電極組老化的預測方法，此方法將有利綠能產業蓬勃發展。
本研究將著重於透過機器學習來預測燃料電池的衰退對其壽命和性能的，透過機器學習的壽命預測方式，以較少的時間和有限燃料有效的預測燃料電池壽命。在本研究中，配合質子交換膜最佳化的操作參數，導入以LSTM等演算法，藉此達到預測燃料電池壽命之驗證。本研究初步的模型對測試資料有良好的預測效果，目前階段針對五級石墨燃料電池堆在氫氣流量6 L/min，加濕溫度於攝氏80度，經過性能測試，於機器學習中導入重要老化特徵，誤差率(Mean Absolute Error, MAE)可以在 2％以內。本研究結果驗證機器學習可有效的應用於燃料電池壽命預測。

The use of renewable energy has conceptually become the world's mainstream. Therefore, fuel cells with low noise, low operating temperature, zero CO2, NOx emissions, high energy efficiency and fast start-up have become the most potential energy options. However, membrane electrode assembly will have a degradation phenomenon after use. In addition to affecting the performance and efficiency of the fuel cell, it will also affect the lifetime of the fuel cell. Since the long-term test is time-consuming, this study proposes a rapid prediction method for the aging of the membrane electrode assembly. This method will be a vigorous development of the green energy industry.

This study focuses on predicting the degradation of fuel cell lifetime and performance through machine learning. Machine learning can effectively predict the lifetime of fuel cell with less time and limited fuel. The optimized operating parameters of the PEMFC and LSTM algorithms are introduced to achieve verification of predicted fuel cell lifetime. The preliminary model of this study has a good prediction effect on the test data. Five-stage graphite fuel cell stack is used in this study, the hydrogen flow rate is 7 L/min, the humidification temperature is 80 oC, after long-term testing, degradation data is introduced in machine learning. Error rate (mean absolute error, MAE) can be within 2%. The results of this study verify that machine learning can be effectively applied to fuel cell lifetime prediction.

[1] L. Xu, U. Reimer, J. Li, H. Huang, Z. Hu, H, Jiang, H, Janben, M. Ouyang, W. Lehnert,, "Design of durability test protocol for vehicular fuel cell systems operated in power-follow mode based on statistical results of on-road data," Journal of Power Sources, vol.377, pp.59-69, 2018, doi: 10.1016/j.jpowsour.2017.11.075.
[2] J. Chen and J. Li, "Fuzzy adaptive iterative learning coordination control of second-order multi-agent systems with imprecise communication topology structure," International Journal of Systems Science, vol.49, no.3, pp.546-556, 2018, doi: 10.1080/00207721.2017.1412533.
[3] J. Liu, S. Laghrouche, F.-S. Ahmed and M. Wack, "PEM fuel cell air-feed system observer design for automotive applications: An adaptive numerical differentiation approach," International Journal of Hydrogen Energy, vol.39, no.30, pp.17210-17221, 2014, doi: 10.1016/j.ijhydene.2014.08.013.
[4] Q. Li, W. Wang, B. Weaver and X. Shao, "Active rotordynamic stability control by use of a combined active magnetic bearing and hole pattern seal component for back-to-back centrifugal compressors," Mechanism and Machine Theory, vol.127, pp.1-12, 2018, doi: 10.1016/j.mechmachtheory.2018.04.018.
[5] L. Hong, J. Chen, Z. Liu, L. Huang and Z. Wu, "A nonlinear control strategy for fuel delivery in PEM fuel cells considering nitrogen permeation," International Journal of Hydrogen Energy, vol.42, no.2, pp.1565-1576, 2017, doi: 10.1016/j.ijhydene.2016.07.240.
[6] D. Zhou, F. Gao, E. Breaz, A. Ravey, A. Miraoui and K. Zhang, "Dynamic phenomena coupling analysis and modeling of proton exchange membrane fuel cells," IEEE Transactions on Energy Conversion, vol.31, no.4, pp.1399-1412, 2016, doi: 10.1109/TEC.2016.2587162.
[7] J. Liu, W. Luo, X. Yang and L. Wu, "Robust Model-Based Fault Diagnosis for PEM Fuel Cell Air-Feed System," IEEE Transactions on Industrial Electronics, vol.63, no.5, pp.3261-3270, 2016, doi: 10.1109/TIE.2016.2535118.
[8] E. Breaz, F. Gao, A. Miraoui and R. Tirnovan, "A short review of aging mechanism modeling of proton exchange membrane fuel cell in transportation applications," in IECON 2014-40th Annual Conference of the IEEE Industrial Electronics Society, 2014: IEEE, pp.3941-3947.
[9] B. Grondin-Perez, S. Roche, C. Lebreton, M. Benne, C. Damour, and J.-J. A. Kadjo, "Mechanistic Model versus Artificial Neural Network Model of a Single-Cell PEMFC," Engineering, vol.06, no.08, pp.418-426, 2014, doi: 10.4236/eng.2014.68044.
[10] K. Chen, S. Laghrouche, and A. Djerdir, "Proton Exchange Membrane Fuel Cell Degradation and Remaining Useful Life Prediction based on Artificial Neural Network," in 7th International IEEE Conference on Renewable Energy Research and Applications, ICRERA 2018, October 14, 2018 - October 17, 2018, Paris, France, 2018: Institute of Electrical and Electronics Engineers Inc., in 7th International IEEE Conference on Renewable Energy Research and Applications, ICRERA 2018, pp.407-411, doi: 10.1109/ICRERA.2018.8567023. [Online]. Available: http://dx.doi.org/10.1109/ICRERA.2018.8567023
[11] 黃鎮江, "燃料電池, 全華科技圖書股份有限公司," 2005.
[12] Y. Kawasoe,Y. Kawasoe, S. Tanaka, T. Kuroki, H. Kusaba, K. Ito, Y. Teraoka and K. Sasaki, "Preparation and electrochemical activities of Pt–Ti alloy PEFC electrocatalysts," Journal of The Electrochemical Society, vol.154, no.9, pp.B969-B975, 2007.
[13] J. M. R. D.A. Stevens, R.E. Mar, R.T. Atanasoski, A.K. Schmoeckel, M.K.Debe, J.R. Dahn, "Enhanced CO-Tolerance of Pt–Ru–Mo Hydrogen Oxidation Catalysts," Journal of The Electrochemical Society, vol.154(12), pp.B1211-B1219, 2007.
[14] S. Hochreiter and J. Schmidhuber, "Long short-term memory," Neural computation, vol.9, no.8, pp.1735-1780, 1997.
[15] J. Liu, Q. Li, W. Chen, Y. Yan, Y. Qiu, and T. Cao, "Remaining useful life prediction of PEMFC based on long short-term memory recurrent neural networks," International Journal of Hydrogen Energy, vol.44, no.11, pp.5470-5480, 2019, doi: 10.1016/j.ijhydene.2018.10.042.
[16] J. Bayer, C. Osendorfer, D. Korhammer, N. Chen, S. Urban, and P. van der Smagt, "On fast dropout and its applicability to recurrent networks," arXiv preprint arXiv:1311.0701, 2013.