Research On Improved Neural Network Control Of Permanent Magnet Synchronous Motor

PMSM(Permanent Magnet Synchronous Motor) for its advantages of small size, high power density, strong overload capacity and excellent control performance has been widely applied and developed in precision control applications. In recent years, due to the energy shortage and low-carbon fossil background, very optimistic about prospects for the development of electric vehicles. Permanent magnet synchronous motor is a high priority in terms of electric vehicle drive. It makes the study of permanent magnet synchronous motor control has become an important content and focus.Permanent magnet synchronous motor having a strongly nonlinear, time-varying and strong coupling characteristics, it is difficult to establish accurate mathematical model, the traditional PID controller can not get a good result. With the development of intelligent control technology, neural network control becomes a new type of control scheme. Artificial neural network’s self-learning, adaptive and self-organizing capacity, which combined with PID controller, make up the shortcomings that traditional PID controller can not adjust parameters in real time and robustness is not strong.In this paper, a hyperballs cerebellar model(HCMAC) neural network based of permanent magnet synchronous motor adaptive speed control tactics,the network structure is simple, without repeated learning, learning fast convergence. When in MATLAB / SIMULINK Environment Simulation results were compared with the PID.It is not difficult to find that using HCMAC neural network online real-time control, speed error is smaller, more precise control, static and dynamic response faster.The paper made a further study of the genetic algorithm optimization neural network. The neural network is chosen cerebellar model(CMAC) neural network, and PID controller is also used as a reference.It turns out that it verify the validity and rationality of the method.So, for high precision and high real-time requirements of the case, it has a strong academic value and predictable application.