(749c) Data-Driven Incipient Fault Management for Proton Exchange Membrane Fuel Cell

In the efforts of establishing cleaner energy systems, hydrogen is receiving increasing attention for transportation and energy storage applications. Among various hydrogen systems of interest, fuel cell vehicles (FCVs) are well studied as the next-gen energy vehicles. Specifically, proton exchange membrane fuel cell (PEMFC) is considered the leading technology for handling hydrogen as fuel. It converts chemical energy from hydrogen to electrical energy via an electrochemical reaction and leaves water as the only byproduct. Therefore, PEMFC is attracting the interest of both the fuel cell community and the automobile industry. However, there are challenges associated with large-scale commercialization of PEMFC for electric vehicles such as operational safety, reliability, performance, and cost [1]. Though several prognostics and health management (PHM) [2, 3] and fault diagnosis [4, 5] studies have been performed to overcome these challenges, the problem of incipient fault diagnosis in PEMFC has not been fully examined. Incipient faults are usually masked by disturbances or other system uncertainties due to their small magnitudes and thus, are unnoticeable and difficult to detect [6, 7]. If no corrective action is taken, incipient faults can evolve into functional faults and lead to catastrophic failures. Therefore, early detection and isolation of incipient faults is highly valuable for the on-board safety of automobiles.

Motivated by the aforementioned considerations, we propose an incipient fault management framework that uses an adaptive data-driven modeling method, OASIS [8] (to detect faults), along with contribution analysis (to isolate faults) through a moving window strategy [9]. First, we obtain a data-driven process model using OASIS for the current operating condition. The model is used to predict the real-time dynamics. Based on this prediction, we continuously evaluate and monitor process risk associated with the fuel cell operation. Here, we incorporate moving window approach for risk assessment to improve the detection accuracy of incipient faults. Also, the adaptive model obtained through OASIS can cope up with any changes in the system, which further helps with decoupling incipient faults from process uncertainties. Next, after the detection step, we initiate contribution analysis to identify the location of the detected incipient fault. Finally, we isolate the process variable with the maximum contribution to the process risk from the remaining components of PEMFC. In this work, we investigate a scenario of hydrogen leakage with a small flowrate. As hydrogen is highly inflammable, it is important to diagnose any small traces of hydrogen leakage to facilitate safe operation of the PEMFC system. To conclude, the proposed framework has the ability to 1) predict the fuel cell operation under varying operating conditions, 2) detect and isolate an incipient fault before it propagates as a functional fault, and 3) assess the process risk to ensure safe operation of the PEMFC system.

Literature cited

[1] Zhang, X., Zhang, T., Chen, H. and Cao, Y., 2021. A review of online electrochemical diagnostic methods of on-board proton exchange membrane fuel cells. Applied Energy, 286, 116481.

[2] Jouin, M., Gouriveau, R., Hissel, D., Péra, M.C. and Zerhouni, N., 2013. Prognostics and Health Management of PEMFC–State of the art and remaining challenges. International Journal of Hydrogen Energy, 38(35), 15307-15317.

[3] Meraghni, S., Terrissa, L.S., Yue, M., Ma, J., Jemei, S. and Zerhouni, N., 2021. A data-driven digital-twin prognostics method for proton exchange membrane fuel cell remaining useful life prediction. International Journal of Hydrogen Energy, 46(2), 2555-2564.

[4] Tian, Y., Zou, Q. and Lin, Z., 2020. Hydrogen Leakage Diagnosis for Proton Exchange Membrane Fuel Cell Systems: Methods and Suggestions on its Application in Fuel Cell Vehicles. IEEE Access.

[5] Li, Z., Outbib, R., Giurgea, S. and Hissel, D., 2018. Fault diagnosis for PEMFC systems in consideration of dynamic behaviors and spatial inhomogeneity. IEEE Transactions on Energy Conversion, 34(1), 3-11.

[6] Safaeipour, H., Forouzanfar, M. and Casavola, A., 2021. A survey and classification of incipient fault diagnosis approaches. Journal of Process Control, 97, 1-16.

[7] Chen, H., Jiang, B., Zhang, T. and Lu, N., 2020. Data-driven and deep learning-based detection and diagnosis of incipient faults with application to electrical traction systems. Neurocomputing, 396, 429-437.

[8] Bhadriraju, B., Bangi, M.S.F., Narasingam, A. and Kwon, J.S.I., 2020. Operable adaptive sparse identification of systems: Application to chemical processes. AIChE Journal, 66(11), p.e16980.

[9] Ji, H., He, X., Shang, J. and Zhou, D., 2016. Incipient sensor fault diagnosis using moving window reconstruction-based contribution. Industrial & Engineering Chemistry Research, 55(10), 2746-2759.