Neural Network-Based Control Strategies For Permanent Magnet Synchronous Motor

In recent years, advances in rare-earth permanent magnetic material, semi- conductor, power devices, micro-processor, converter design technique and control theory have made permanent magnet synchronous motor (PMSM) drives play a vital role in motion-control applications in the low-to-medium power range. However, the control performance of the PMSM drive is still influenced by uncertainties, which usually features parameter variations, external load disturbances, unmodelled and nonlinear dynamics. To achieve high-performance PMSM drive, which has great ability of adaptation and better performance against disturbances, advanced control schemes have to be developed to deal with these uncertainties. On the basis of the research on structure and learning algorithm of neural network, this thesis make a thorough study for the speed, position and sensorless control schemes of PMSM drive, which combine the merits of PID control, adaptive control and sliding-mode control and the predominant capability of neural network in identification and control of uncertain nonlinear system. The proposed schemes, whose effectiveness is demonstrated by the theoretical analysis and experimental simulation results, overcome the bad influences of PMSM parameter variations and external load disturbances. The main contributions of this dissertation are as follows:1. The principle of i d=0 control method is analyzed systematically based on vector control technique of PMSM. This thesis points out that vector control is more a static decoupling than a full decoupling. In fact, the i d=0 control method of PMSM, which can realize torque linearization control, is a vector decoupling control.2. The learning algorithm of recurrent neural network is investigated. First, the dynamic BP algorithm is corrected for diagonal recurrent neural network(DRNN), that makes the algorithm simple. Secondly, based on the characteristics of connection...