Neural Network Controlled Space Vector PWM for a High Power Factor Converter Under Perturbed Conditions

This work demonstrates the application of neural network to optimize space vector pulse width modulation (SVPWM) for a neutral-point clamped rectifier especially for supply and load perturbations. Neural network is used here to obtain the exact switching time for the converter. The advantage of using neural network is to avoid complex mathematical computations for generating fast, best path, and shape of the reference vector. Learning algorithm for offline training of neural network is based on a well-performed model of modified SVPWM used to synthesize a linear functional relationship between perturbed condition parameters and the control gains. The offline training of neural network is repeated until the control gains as input parameters for the neural network-based model produce optimal results for each perturbed condition. The proposed neural network-based controller scheme is modeled in MATLAB/Simulink software. Simulation results show that the proposed algorithm implementation is simple and displays better performance under any supply and load conditions.