In order to solve the problem of nonlinearity and parameter time-varying uncertainty in high-power AC servo system, a predictive model of adaptive mutation particle swarm optimization wavelet neural network for chaotic search is proposed. The mathematical model of AC servo motor is established. Different mutation methods are used to make the particles close to different search areas. Chaos optimization algorithm is introduced to improve the particle swarm. The AMPSO-WNN algorithm based on chaotic search is used to improve the probability and speed of global convergence. The simulation results show that the prediction accuracy of the optimized model is higher than before, and the improved algorithm has strong function approximation ability, the network performance is improved significantly, and the local minimum value problem is effectively solved.