Surrogate Model-Assisted and Load-Dependent Optimization of Prismatic LFP Battery Module Cooling
This study examines the offline selection of coolant flow rate and inlet temperature for a fixed eight-cell, 50 Ah prismatic lithium iron phosphate battery module under different loads. A dataset of 100 steady-state CFD cases was used to predict maximum battery temperature, inter-cell temperature difference, and cold-plate pressure drop. Response surface methodology, XGBoost, and Kriging were compared using five-fold cross-validation on 80 training cases and prediction errors on 20 separately generated cases. Kriging achieved the lowest errors for all three responses. The response-specific Kriging models were refitted using all 100 cases and coupled with NSGA-II to generate approximate Pareto candidates at eight loads from 0.5 to 4.0 C. After screening for maximum temperature ≤ 40 °C and inter-cell temperature difference ≤ 5 K, the candidate with the lowest predicted pressure drop was selected at each load. The selected flow rates ranged from 1.000 to 1.008 L/min/plate, with load-specific inlet temperatures. Additional CFD simulations confirmed that all eight combinations met both adopted thermal targets under the original CFD model. Prediction of all three responses required 2.486 ± 0.788 ms per condition, supporting repeated evaluation of operating alternatives.
Authors
- 安继业
- Seongjae Won (ORCID: https://orcid.org/0009-0001-8006-8704)
- Jinwook Lee (ORCID: https://orcid.org/0000-0002-3645-2573)
- Gayeon Kim (ORCID: https://orcid.org/0009-0000-3490-4161)
- Taegyeol Lee
Institutions
- Soongsil University (KR)
Publication Details
- Journal
- Applied Sciences
- Published
- 2026-11-18
- DOI
- https://doi.org/10.3390/app162211426
- Primary Topic
- Battery Modeling and Management
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Korea Institute of Energy Technology Evaluation and Planning