Taking full account of the complex mechanism of the continuous annealing furnace (CAF) production process and the difficulty in establishing an accurate mathematical model, this paper adopts a quantum particle swarm optimization (QPSO) algorithm to optimize the parameters of the radial basis function (RBF) neural network which used to predict the model of strip temperature in CAF. Firstly, to improve the accuracy of modeling, the input and output variables of RBF neural network model prediction are determined by analyzing the mechanism model of the heating section of the CAF and the factors affecting the strip temperature. Secondly, due to the trial and error method used for parameter selection in RBF neural network, which results in low work efficiency and difficulty in selecting the optimal value, the QPSO algorithm is introduced to search for the optimal solution. Furthermore, to avoid encountering local optimal issues and improve searching performance, an improved QPSO algorithm that combines the characteristics of generalized opposition-based learning and differential evolution is proposed. Finally, by collecting the production site data of a large-scale CAF, experiments are carried out to verify the effectiveness of the proposed methods.