EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Regime‐Aware Machine Learning–Based Voltage Prediction of PEM Fuel Cells Under Variable Operating Conditions

Hüseyin Kahraman

International Journal of Energy Research · 2026-01 · 2026 (1)

DOI: 10.1155/er/8667013 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Energy Research, online erschienen am 30.08.2026. Verfasst von Hüseyin Kahraman. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 48 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate voltage prediction of proton exchange membrane fuel cells (PEMFCs) under variable operating conditions remains challenging because of nonlinear interactions among current density, temperature, pressure, humidity, and mechanical assembly conditions. This study develops a regime‐aware machine–learning framework for PEMFC voltage prediction using 9930 time‐resolved experimental observations collected from 11 distinct operating regimes. Random forest (RF), least‐squares boosting (LSBoost), Gaussian process regression (GPR), artificial neural networks (ANNs), and support vector regression (SVR) were evaluated using grouped five‐fold cross‐validation, in which all observations from the same temperature–humidity–pressure–torque combination were kept within a single fold. This strategy enabled assessment of model generalization to previously unseen operating regimes. RF showed the most balanced performance, with a mean R 2 of 0.9215 ± 0.0784, root mean squared error (RMSE) of 0.0421 ± 0.0236 V, and mean absolute error (MAE) of 0.0287 ± 0.0156 V. LSBoost achieved comparable average performance, whereas GPR produced highly accurate predictions in selected folds but showed greater fold‐to‐fold variability. ANN and SVR exhibited limited generalization under strict regime–level validation. A raw‐versus‐engineered‐feature ablation analysis demonstrated that current‐density transformations and selected interaction terms improved RF predictive performance. Complementary intrinsic and permutation importance analyses identified current‐density–related variables and inlet pressure as influential predictors. The results demonstrate the importance of grouped validation for realistic assessment of PEMFC voltage–prediction models and provide a basis for future optimization, monitoring, and control‐oriented applications.

Zitieren

APA

Kahraman, H. (2026). Regime‐Aware Machine Learning–Based Voltage Prediction of PEM Fuel Cells Under Variable Operating Conditions. International Journal of Energy Research, 2026(1). https://doi.org/10.1155/er/8667013

Vancouver

Kahraman H. Regime‐Aware Machine Learning–Based Voltage Prediction of PEM Fuel Cells Under Variable Operating Conditions. International Journal of Energy Research. 2026;2026(1). doi:10.1155/er/8667013

BibTeX
@article{Kahraman2026,
  author = {Kahraman, Hüseyin},
  title = {Regime‐Aware Machine Learning–Based Voltage Prediction of PEM Fuel Cells Under Variable Operating Conditions},
  journal = {International Journal of Energy Research},
  year = {2026},
  volume = {2026},
  number = {1},
  doi = {10.1155/er/8667013}
}
Kostenlos starten →