Design And Implementation Of FPGA-based Neural Network Decoupling Control Of Permanent Magnet Synchronous Motor

In recent years, AC drives has become a major research and application direction in industrial electric drive. The permanent magnet synchronous motor(PMSM) has the advantages of small size, light weight, high power density, low output torque, high efficiency and easy maintenance, so it is always popular in the theory and applied research of electrical transmission direction. For such a complex nonlinear control object, the traditional controlling methods can’t get excellent controlling and decoupling effect because of the exits of dynamic parameter changes and the internal coupling state. The neural network inverse system control is one viable option. Neural network has the ability to approximate nonlinear objects with arbitrary precision, decoupling control the PMSM with the neural network and the inverse system method has a good effect. However, the neural network does not fit the general processor because of the characteristics of parallel processing. The neural networks based on the FPGA can solve this problem.In this paper, based on the neural network decoupling control of the PMSM, the neural network inverse system is established and simulated, and the feasibility of neural network inverse system model is verified. The structure and parameters of the good performance neural network inverse system model is achieved. The hardware implementation method of the neural network inverse system module is further researched.A variety of incentives of the neural network function module is built and the characteristics of the different implementations are compared based on the FPGA.FPGA modules of the neural network is built, simulated and verified based on the previous studies.In this paper, in the process of building the neural network inverse model PMSM system, according to the mathematical model of the PMSM analysis, the reversibility of permanent magnet synchronous motor is proven and the inverse system model is established.And based on the neural network inverse system theory, the neural network model is established, the structures and parameters of the ideal neural network model is achieved through the Matlab training. The trained neural network module is brought back to the neural network inverse system to be verified and good performance results are reflected. And the implementation algorithm of the excitation function is identifiedand designed. The Simulation of the neural network implementation is conducted with Matlab and System Generator. According to the structure and parameters of neural networks which are achieved by the training of the neural network, the FPGA model of neural network is established by System Generator and verified by the simulation.