Position Sensorless Control of 4S3P Inverter Fed PMSM Drive Based on Artificial Neural Network

Abstract

A position sensorless Permanent Magnet Synchronous Motor (PMSM) drive based on single layer Recurrent Neural Network (RNN) is presented in this research work. In the proposed control methodology, instead of a usual 6-Switch 3-Phase (6S3P) inverter a 4- Switch 3-Phase (453P) inverter is used. This reduces the cost of the inverter, the switching losses, and the complexity of the control board for generating six Pulse Width Modulated (PWM) signals. A simulation model of the drive system is developed and used in this study. The motor equations are written in rotor fixed d-q reference frame. A Proportional plus Integral (Pt) controller is used to process the speed error to generate the reference torque current. The RNN estimator is used to estimate stator flux components along the stator fixed stationary axes (a-fl). In this study, the Correlated Real Time Recurrent Learning (CRTRL) algorithm is used for training the neural network. The rotor angle is used in vector rotator to generate the reference phase currents. Hysteresis current controller block controls the switching of the three phase inverter to apply voltage to the motor stator. Numerical simulation is carried out in order to verify the effectiveness of the proposed control system. Simulation studies show that the proposed RNN estimator can be used to accurately measure the motor fluxes and rotor angle over a wide speed range. The robustness of the drive system is tested for different operating conditions, i.e., sudden load torque change, parameter deviation, speed reversal, ramp change of speed, load disturbance, presence of computational error, etc. The control system is found to work acceptably under these conditions. It is also simple and low cost to implement in a practical environment.