Neural-network-based space-vector PWM of a three-level inverter covering overmodulation region and performance evaluation on induction motor drive

Abstract

A multi-layer feedforward artificial neural-network (ANN) based implementation of space-vector pulse width modulation algorithm for a three-level voltage-fed inverter has been described in the paper that fully covers the undermodulation and overmodulation regions with linear transfer characteristics extending operation smoothly up to square-wave. The ANN, when implemented by dedicated application-specific IC chip, permits simple, fast and reliable operation far exceeding the capability of digital signal processor (DSP). The network receives the voltage magnitude and angular position of the command space vector at the input, and generates the digital words for switching times at the output that are then converted to symmetrical pulse widths for the three phases through simple logic circuits and a single timer. Several alternative ANN topologies have been proposed after successful training, and their structures and performances have been compared. The ANN-based modulator was then incorporated in induction motor drive systems with rotor flux-oriented indirect vector control, and the drive performances were evaluated extensively. In all the cases, performances were found to be excellent.