Adaptive surrogate model-based optimization framework applied to battery pack design

The efficiency represented by the computational costs for solving complex engineering optimization problems can be improved based on surrogate models. Adaptive sampling methods can effectively reduce the number of samples and enhance the accuracy of the surrogate model. However, from the perspective of optimization, most adaptive sampling methods ignore the inevitable connection between the optimization process and the establishment of a surrogate model. The efficiency and accuracy of existing adaptive sampling methods are poor. In this paper, a novel adaptive sampling method is proposed based on the complex method (CM). In the proposed method, the samples are used not only to establish the surrogate model but also to form complex shapes to guide the optimization search. An iterative process is added to collect new samples to improve the accuracy of the surrogate model. The proposed method is compared with three classic adaptive methods with several benchmark functions. The comparison results indicate that the proposed method can obtain a global optimum with low computational burden and a small sample. Finally, the optimum design of an electric vehicle battery pack is presented to illustrate the feasibility and validity of the proposed method.