An efficient and accurate method for optimizing capsule-type plate heat exchangers is proposed in this paper. This method combines computational fluid dynamics simulation, a backpropagation algorithm and multi-objective optimization to obtain better heat transfer performance of heat exchanger structures. For plate heat exchangers, the heat transfer coefficient j and friction coefficient f are a pair of contradictory objectives. The optimization of capsule-type plate heat exchangers is a multi-objective optimization problem. In this paper, a backpropagation neural network was used to construct an approximate model. The plate shape was optimized by a multi-objective genetic algorithm. The optimized capsule-type plate heat exchanger has lower flow resistance and higher heat exchange efficiency. After optimization, the heat transfer coefficient is increased by 8.3% and the friction coefficient is decreased by 14.3%, and the heat transfer effect is obviously improved. Further, analysis of flow field characteristics through field co-ordination theory provides guidance for the further optimization of plates.