A novel model for optimizing antenna design is proposed in this paper, combining the artificial neural network (ANN) with the simulated annealing (SA). The data set to train the ANN is obtained from the full-wave simulation software. The trained ANN can quickly predict the electromagnetic parameters of the antenna, and then the SA is applied to optimize the antenna geometrical parameters to obtain the target value. In order to verify the feasibility and effectiveness of this model, a broadband patch antenna with slots is designed. The sizes and positions of the slots are the input of the ANN, while the return loss of the antenna is the output. By applying the proposed model, the broadband patch antenna with 17.4% relative bandwidth is achieved which proves that the proposed model is feasible.