Aiming at the high performance control problem of the permanent magnet synchronous motor (PMSM), with the full consideration of its uncertainty, time-varying characteristics, measurement noise and other stochastic factors, a control scheme based on improved adaptive inverse control is proposed to compensate the controlled object to be a system with linear transfer relation by using the inverse system of the PMSM. A double closed-loop control structure under vector control is adopted to introduce the inverse control method of the multidimensional Taylor network (MTN) into the velocity loop. Firstly, the invertibility of the PMSM mathematical model is proved to solve the existence problem of the inverse model of the nonlinear system. A novel and dynamic network controller, called the MTN, is then established with the advantages of simple structure and low computational complexity. After that, in order to achieve the high precision of speed control, three MTNs are used respectively as an adaptive model identifier for system modeling, an adaptive inverse controller for inverse modeling and a nonlinear adaptive filter for eliminating the noise disturbance to separate the control of the object dynamic response and disturbance cancellation into relatively independent processes. An optimal control can be realized at no cost of the other. Finally, the simulation results show that the proposed control scheme can realize accurate speed control of the PMSM servo system, with good tracking performance and strong anti-interference ability.