In view of the nonlinearity and time variability of industrial control systems, as well as the poor transient response in traditional adaptive control, this paper presents a neural network multi-model switching adaptive control method basing on particle swarm optimization. Firstly, the PSO algorithm was used to adjust the neural network weights to achieve the optimal value. Then an adaptive control strategy was designed basing on the BPNN and multiple models. The optimal controller can be selected to control the system through the constructed rational switching rules. The good approximation ability of neural network can improve the performance of adaptive control. The performance through PSO optimization are studied through simulation methods using Matlab, which verifies that the proposed method can significantly improve the overall performance of the system：fast convergence, high precision, good network generalization and approximation ability, and can precisely track the output of the control system.