Machine learning technique for low-frequency modulation techniques in power converters

In power systems, the main objective of active power filter (APF) is to control and reduce the harmonics of nonlinear loads. Also, the APF can compensate the reactive power (the AC power fundamental component) at the point of common coupling. This chapter investigates a technique for the modulation technique of the APFs. The real-time fundamental and harmonic compensations can be achieved using the low-frequency modulation techniques such as asymmetric selective harmonic elimination/mitigation-pulse width modulation (ASHE/ASHM-PWM) and asymmetric selective harmonic current mitigation-PWM (ASHCM-PWM) by using an artificial neural network (ANN) technique. This means that in real time by using the proposed technique, different phases and magnitudes of the fundamental and harmonics for the voltage of the converter can be obtained. Moreover, in this chapter, a guideline will be proposed for generating ANN training data for the ASHCM-PWM technique. Simulation and experimental results are conducted on a 7-level (3-cell) cascaded H-bridge APF to evaluate the advantages and effectiveness of the proposed ANN-based technique.