Surrogate model‐based heat dissipation optimization of air‐cooling battery packs involving herringbone fins

Abstract

For lithium-ion batteries (LIBs) used in electric vehicles (EVs), their performance is greatly affected by temperature. To ensure the safety and reliability of EVs, a reliable and efficient battery thermal management system (BTMS) is essential. This paper mainly analyzes air cooling BTMS. First, we propose an air-cooling heat dissipation method for battery modules with herringbone fins and long sleeves, and prove the effectiveness of the program by comparing it with the finless and sleeveless schemes. Second, parametric modeling of the scheme is carried out, the response surface methodology (RSM), radial basis function (RBF), and Kriging surrogate model of the scheme are constructed through design experiments. Then, the Kriging model is determined by the model evaluation index to be well adapted to the problem. We build a multiobjective optimization model (MOOM) and optimize it by multi-objective genetic algorithm (MOGA). Finally, it is found that the maximum temperature (T max ) and the maximum temperature difference (ΔT max ) decrease by 1.5 K and 36.79% without considering the cooling cost. When considering the cooling cost, T max and ΔT max dropped by 0.69 K and 17.92%, respectively. The optimization result improves the heat dissipation effect of the battery module and controls the cooling cost within the required range. Besides, optimization analysis can be carried out according to different actual cost requirements, which provide corresponding guidance for large-scale air-cooling heat dissipation analysis of LIBs. K E Y W O R D S battery thermal management system (BTMS), electric vehicles (EVs), Kriging surrogate model, multi-objective optimization model (MOOM)

MECSBO: Multi‐strategy enhanced circulatory system based optimisation algorithm for global optimisation and reliability‐based design optimisation problems

“…Optimisation has long been a focal point in design methodologies, aiming to find the most efficient solutions [1,2]. However, inherent uncertainties prevail in material properties, loads, and geometrical parameters, potentially influencing the practical engineering structure derived from deterministic optimisation approaches [3][4][5][6][7].…”

A novel learning function for adaptive surrogate-model-based reliability evaluation

“…In order to address the problem of high computational cost of MCS, numerous reliability evaluation methods based on surrogate models have been extensively investigated. These surrogate models include artificial neural networks (ANNs) [31], Kriging models [32,33], support vector machines (SVMs) [34], radial basis function (RBF) [31] and polynomial chaos expression (PCE) [35,36]. Once the actual LSF is approximated by these surrogate models, the computational cost of the MCS method will be greatly reduced, although taking a sufficient number of sample points in the variable space can make the surrogate model have a high fitting accuracy.…”

A comparative study for adaptive surrogate-model-based reliability evaluation method of automobile components

“…In reliability assessment, multi-physics coupling, multi-variable, multi-source uncertainty information (Wang et al ., 2022; Liu et al ., 2020; Pan and Deng, 2018; Xue and Deng, 2021) are often involved, so these assessments are even more unacceptable (Liu et al ., 2021b; Gao et al ., 2022; Yang et al ., 2022; Meng et al ., 2022b). Surrogate model (such as Kriging model (Li et al ., 2021; Meng et al ., 2021b), Canonical Low Rank Approximation (CLRA) (Wang et al ., 2019), Polynomial Chaos Expansions (PCE) (Zhu et al ., 2023), Support Vector Regression (SVR) (Luo et al ., 2022a) and Polynomial Chaos-Kriging (PCK) (Meng et al ., 2017; Schöbi et al ., 2017) etc.) is an approximation technique.…”