Artificial neural network based space vector PWM for a five-level diode-clamped inverter

This paper reports the neural network based space vector pulse width modulation for a five-level three-phase diode-clamped inverter. The ANN-based structure generates control signals for the five-level inverter. A multilevel three-phase inverter offers several advantages over conventional inverters such as lower-voltage stresses on power electronic switches, better electromagnetic compatibility and smaller ratings for the switches. Space vector pulse width modulation (SVPWM) enhances the output features of this inverter by properly utilizing the DC link voltage. An artificial neural network (ANN) makes the implementation of SVPWM easier as now any non-linear function of an arbitrary degree can be approximated. The ANN for SVPWM was trained in Matlab Simulink and implemented with TMS320F2812 using Matlab embedded encoder. For comparisons, sinusoidal PWM-based control technique was also implemented. The results are presented that illustrate the merits of the ANN-based SVPWM for a five-level diode-clamped inverter.