Comparison of State-Space and LSTM-Based Virtual Temperature modelling in PEM fuel cells.
DOI:
https://doi.org/10.17979/ja-cea.2026.47.13863Keywords:
PEM Fuel Cells, Virtual Temperature Sensing, State-Space Modelling, Long Short-Term Memory (LSTM), System Identification, Fault-Tolerant MonitoringAbstract
This study compares model-based state-space prediction and Long Short-Term Memory (LSTM) approaches for virtual outlet-temperature sensing in proton exchange membrane fuel cells (PEMFCs). Experimental data consisting of stack current,
inlet cooling-water temperature, and outlet cooling-water temperature were collected under dynamic operating conditions. The state-space model was developed using system identification techniques, while the LSTM model was trained using a data-driven
sequence-learning approach. Different input combinations were evaluated using root mean square error (RMSE) and correlation analysis. Results showed that both methods could track thermal behavior, although their performance depended strongly on input selection and thermal observability. The LSTM demonstrated strong capability in learning nonlinear thermal dynamics, while the state-space approach provided improved physical interpretability.
References
Abiola, A., Manzano, F.S., Andújar, J.M., 2023. A Novel Deep Reinforcement Learning (DRL) Algorithm to Apply Artificial Intelligence-Based Maintenance in Electrolysers. Algorithms https://doi.org/10.3390/a16120541 16, 541.
Ariza, E., Correcher, A., Vargas-Salgado, C., 2023. PEMFCs Model-Based Fault Diagnosis: A Proposal Based on Virtual and Real Sensors Data Fusion. Sensors 23, 7383. https://doi.org/10.3390/s23177383
Guffanti, D., Pavon, W., 2025. MultivariateSystem Identification of Differential Drive Robot: Comparison Between State-Space and LSTM-Based Models. Sensors 25, https://doi.org/10.3390/s25185821 5821.
Ignatenko, V., Yudintsev, A., Lyapunov, D., 2019. Application of State-Space Method for Control System Analysis, in: 2019 International Siberian Conference on Control and Communications (SIBCON). Presented at the 2019 International Siberian Conference on Control and Communications (SIBCON), IEEE, Tomsk, Russia, pp. 1–5. https://doi.org/10.1109/SIBCON.2019.8729658
MathWorks, 2026. System Identification Toolbox User’s Guide. The MathWorks, Inc., Natick, Massachusetts, USA.
Vuong, V.-D., Nguyen, L.-H., Goulet, J.-A., 2025. Coupling LSTM neural networks and state-space models through analytically tractable inference. International Journal of Forecasting 41, 128–140. https://doi.org/10.1016/j.ijforecast.2024.04.002
Zheng, X., Zaheer, M., Ahmed, A., Wang, Y., Xing, E.P., Smola, A.J., 2017. State Space LSTM Models with Particle MCMC Inference. https://doi.org/https://doi.org/10.48550/arXiv.1711.11179
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Abiodun Abiola, Antonio Javier Barragán, José Manuel Andújar, Francisca Segura
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
