Diagonal Recurrent Neural Network Sensorless Control Of Permanent Magnet Synchronous Motor

The intelligent control based on neural network offers an effective control method for high-performance AC drive systems. Therefore the combination of intelligent control and vector control has become a hotspot. Thus, the study of the sensorless control method based on neural network for Permanent Magnet Synchronous Motor(PMSM)has important theory and practical meanings.In this thesis, A diagonal recurrent neural networks(DRNN)controller is introduced on the basis of model observer. The controller solves model observer approach relying on mechanical parameters of machine excessively. A sensorless vector control strategy of PMSM based on DRNN is proposed. The mathematics model of PMSM which is analyzed in chapter 2 is built up, the simulation model of vector control system with space voltage pulse width modulate(SVPWM)is built up. First, the thesis dissertates on the principles of vector control system to sensorless operation of PMSM, then DRNN observer is analyzed in details. DRNN controller is integrated into sensorless operation of PMSM. According to the mathematical models of PMSM, the d- and q-axis current components are estimated to feed back, which can adjust the weight of DRNN to realize earning and adjusting on line. The simulation in the environment of MATLAB shows that this method has good motor rotor position tracking and speed tracking performances. Moreover, the ability of anti-disturbance of load is the remarkable feature of this system, and the control performance is pretty good. It is shown that the proposed system can satisfy the fast dynamic response and accurate speed or torque control and strong robustness.