Sensorless Control of PMSM Drive Based on Neural NetworksSensorless Control of PMSM Drive Based on Neural NetworksSensorless Control of PMSM Drive Based on Neural Networks

• Main Authors: Sergiu Gratian Berinde, 裴瑞
• Other Authors: Ming-Shyan Wang
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/10306358975653642905

碩士 === 南台科技大學 === 電機工程系 === 98 === The thesis is mainly concerned with designing a method for speed estimation for a sensorless permanent magnet synchronous motor (PMSM) drive. Traditionally, encoders or resolvers are used to get the position feedback in a servo drive, however, due to the drawbacks associated with these sensors, like system costs, increased rotor inertia, significant research has been done in the area of rotor position sensorless control. Most of the sensorless control methods require some knowledge of the motor parameters and structure, which are not always supplied by the manufacturer or are difficult to obtain. This thesis presents another way of estimating the speed without the knowledge of motor parameters. An artificial neural network (ANN) based observer is designed, taking on the role of a software transducer for speed feedback of the motor drive. The neural observer is based on the Elman neural network structure and takes as inputs only motor voltages and currents; the output is the estimated speed. The observer is first trained offline to learn the dynamics of the motor. After the offline training, the observer is incorporated into a TMS320F2812 DSP based drive and sensorless control is achieved. Finally, some advantages over other sensorless control methods are discussed and experimental results of the proposed method are presented.