Real Time Selective Harmonic Control - PWM Based on Artificial Neural Networks

Selective harmonic elimination-pulse width modulation (SHE-PWM) is a widely used low switching frequency modulation technique for medium-voltage high-power converters. This approach is able to adjust the converter fundamental component while eliminating low-order harmonics. However, some applications such as active power filters (APFs) require regulating simultaneously, both the fundamental and low-order harmonics in amplitude and phase. This article presents a novel selective harmonic control-PWM (SHC-PWM) modulator, valid for APFs, based on artificial neural networks (ANNs) and sequential quadratic programming (SQP). A new offline search methodology, based on a hybrid metaheuristic-numerical algorithm, is defined to calculate the solution space when both the fundamental and a low-order harmonic are controlled in phase and amplitude. The solutions obtained are used to train the ANNs offline. Afterwards, the ANN + SQP calculation method is used to solve the SHC-PWM problem in real-time (RT). Experimental results are provided for a three-level converter to verify the effectiveness of the proposed RT control method.