Neural network based active power filter for power quality improvement

In this paper, a neural network based reference current computation method for active power filter (APF) control under non-ideal mains voltage with different load condition is proposed. The neural network controller has been designed to extract fundamental frequency components from non-sinusoidal and unbalanced currents. These fundamental frequency components will be used as unit templates. The APF realized by a current controlled IGBT based PWM-VSI bridges with a common dc bus capacitor. Current controller based on pulse width modulation with suitable carrier frequency is used to generate the firing pulse. The proposed neural network reference generation approach has been evaluated in simulations. The results show excellent performance, as well as robustness and usefulness of the system.