A modified approach to both conventional and ANN based SVPWM controllers for voltage fed inverter in sensorless vector control IM drive

Abstract

In this paper, space vector pulse width modulation (SVPWM) scheme for voltage fed inverter (VFI) using conventional method and artificial neural network (ANN) based approach are presented separately. In the conventional method, the difficulty of explicitly expressing cross-over and holding-angle as a function of modulation factor in overmodulation mode-I and mode-II respectively are overcome by introducing Newton's Forward Interpolation (NFI). This greatly simplifies the implementation of conventional SVPWM technique without compromising the accuracy issue. The SVPWM is further implemented by ANN based approach built with three subnets to account for three regions of inverter operation distinctly. In comparison to a single ANN taking care of all the three regions, this apparent redundancy of subnets markedly reduces the error in calculating turn-on time for inverter switches. The performances of these two schemes are quantitatively expressed by total harmonic distortion in motor line current. Finally, the schemes are also tested using VFI for sensorless vector control of induction motor drives. The results lead to conclude that the ANN based approach can successfully take over the conventional approach of SVPWM for such applications. The proposed method is validated through computer simulation using MATLAB/SEVIULINK.