DANet demand power prediction model with cross-power level transfer learning for PEMFCs

Frequent power fluctuations under real driving conditions affect the dynamic performance and durability of proton exchange membrane fuel cells (PEMFCs), making accurate demand power prediction important for fuel cell vehicle (FCV) applications. PEMFCs are modular (30 kW/unit), enabling upgrades from 90 to 120 kW. However, demand power data are non-stationary, and target-domain data at new power levels are limited. Therefore, this study proposes the dual-path attention network (DANet) demand power prediction model. The DANet model combines a temporal convolutional network (TCN) with a bidirectional gated recurrent unit (BiGRU) and introduces channel and temporal attention mechanisms. Through a freeze-and-fine-tune transfer learning strategy, the DANet model trained at the 90 kW power level is adapted to the 120 kW power level. The results indicate that prior to transfer learning, the DANet model at the 90 kW power level exhibited an RMSE of 0.316 kW and an MAE of 0.228 kW, significantly outperforming long short-term memory model (LSTM), and Transformer model. After transfer learning, the mean error of the adjusted DANet model at the 120 kW power level decreased from 0.6 to 0.2 kW under 70% of the training sets. The proposed PEMFC demand power prediction model can effectively accelerate the development of new vehicles.

Authors

Institutions

Publication Details

Journal
Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering
Published
2026-11-16
DOI
https://doi.org/10.1177/09544070261506856
Primary Topic
Electric and Hybrid Vehicle Technologies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

DANet demand power prediction model with cross-power level transfer learning for PEMFCs

J.S. Wang, Biaoyi Liu, Hua Ding, Yamin Hu et al.
4 links
Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering
Electric and Hybrid Vehicle Technologies
article

DANet demand power prediction model with cross-power level transfer learning for PEMFCs

J.S. Wang, Biaoyi Liu, Hua Ding, Yamin Hu, Wenxuan Wei, Xiaoming Zou, Donghai Hu, Yan Sun
article en

Abstract

Frequent power fluctuations under real driving conditions affect the dynamic performance and durability of proton exchange membrane fuel cells (PEMFCs), making accurate demand power prediction important for fuel cell vehicle (FCV) applications. PEMFCs are modular (30 kW/unit), enabling upgrades from 90 to 120 kW. However, demand power data are non-stationary, and target-domain data at new power levels are limited. Therefore, this study proposes the dual-path attention network (DANet) demand power prediction model. The DANet model combines a temporal convolutional network (TCN) with a bidirectional gated recurrent unit (BiGRU) and introduces channel and temporal attention mechanisms. Through a freeze-and-fine-tune transfer learning strategy, the DANet model trained at the 90 kW power level is adapted to the 120 kW power level. The results indicate that prior to transfer learning, the DANet model at the 90 kW power level exhibited an RMSE of 0.316 kW and an MAE of 0.228 kW, significantly outperforming long short-term memory model (LSTM), and Transformer model. After transfer learning, the mean error of the adjusted DANet model at the 120 kW power level decreased from 0.6 to 0.2 kW under 70% of the training sets. The proposed PEMFC demand power prediction model can effectively accelerate the development of new vehicles.

Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering
Jiangsu University (CN)
Affordable and clean energy, Industry, innovation and infrastructure, Climate action
Electric and Hybrid Vehicle Technologies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon