The battery thermal management system effectively limits the temperature of each lithium-ion battery (LIB) to below 45°C and minimises the temperature difference between different LIBs to extend their service life. Given the volume constraints

the finite element method (FEM) was used to perform the structural optimisation calculation of battery thermal management systems (BTMS). However

owing to their high calculation costs

optimisation methods based on surrogate models are preferred. The k-means clustering strategy of the stochastic reduced-order model (SROM) method

as implemented within the domain of uncertainty analysis

was shown in this study to enhance the initial observation point sampling strategy of the Kriging optimisation method. The use of an active sampling strategy has been demonstrated to enhance the representativeness of observation points with respect to the overall grid points

which in turn accelerates the convergence rate of the Kriging optimisation method. In the multiphysics simulation example of an LIB liquid cooling system modelled in COMSOL software

the relative error of the improved Kriging method is reduced from 0.24% to 0.11% compared with the traditional Kriging method

and the calculation efficiency is increased by 86.7%. This provided a quantitative verification of the effectiveness of the proposed method.