To solve the key technical problems of the laborious process and difficulty in achieving rapid structural optimization design using the traditional computational fluid dynamics (CFD) method to obtain the wind coefficient of the main beam section of port crane, a rapid prediction model for wind coefficient of crane main beam section based on convolutional neural network is proposed. The wind coefficient rapid prediction model proposed in this paper uses the free geometric deformation method to process the basic section shape to obtain the cross-section pattern set of crane main beam with rich geometric characteristics, CFD method is then used to calculate the wind coefficient corresponding to the cross-section pattern of each main girder to generate the dataset. On this basis, the prediction model is trained based on the dataset and its network structure is optimized, and the nonlinear mapping relationship between the main beam section and the wind coefficient is established. In addition, this paper further combines the prediction model with the genetic algorithm to establish an optimization design method for main beam cross-section. Taking the F₁₁ section in the dataset as an example, the accuracy and efficiency of the optimization method are tested by using the windproof performance as the optimization objective. The test results show that the wind coefficient rapid prediction model proposed in this paper achieves the average relative error of 1.87% when predicting the wind coefficient of each main beam section. The prediction time is in the order of milliseconds, which is significantly faster compared to the traditional CFD method, showcasing a significant improvement in efficiency. The optimized F₁₁ section of crane main beam section developed by applying the optimization design method developed in this paper is reduced by 15.89% compared with the wind coefficient before optimization. This significant reduction greatly improves the windproof performance of the main beam section, which proves the reliability of the optimization method proposed in this paper. It can be used as a new method for the optimization design and rapid selection of crane main beam section structure.