A heat exchanger plate is the core component of a heat exchanger with wide channel, which has a direct impact on heat exchange effect. In order to improve its efficiency and decrease energy loss, an improved multi-objective brain storm optimization algorithm is introduced to seek the optimal structure of a heat exchange plate. According to its shape and layout, six structural parameters about size and distribution of trapezoidal convex units are defined as decision variables. Following that, 25 groups of samples are obtained using the orthogonal method and numerical simulation based on Fluent. Two surrogate models for evaluating Nusselt index and pressure loss are constructed by regression. Taking maximum heat exchange efficiency and minimum energy loss as optimization objectives, the algorithm of grid-based multi-objective brain storm optimization is employed to find the optimal structure of a heat exchange plate. Statistical experimental results indicate that surrogate model of heat exchange performance can effectively reduce the cost of evaluation, and the proposed optimization method can obtain the optimal structure with the highest heat exchange efficiency and least energy loss.