Implementation of Artificial Neural Network Controlled Shunt Active Power Filter for Current Harmonics Compensation

Abstract: A pure sinusoidal voltage is a conceptual quantity produced by an ideal AC generator. The nonlinear loads cause the current to vary disproportionately and the current taken by them has a non-sinusoidal waveform. The aim of this project is to present harmonic cancellation using shunt APF with an application of neural network (NN). Shunt active power filter (SAPF) is commonly used as an effective method in compensating harmonic components in non-linear loads. This active filter concept uses power electronic equipment to produce the current components that cancel the harmonic current components from the nonlinear loads. For the extraction of sequence currents and to improve the performance, the back propagation algorithm is used. This is a multi-layer feed-forward neural network. The back propagation method is for obtaining the source currents at the fundamental frequency. The Active Power Filter reference compensation currents are then determined by subtracting the fundamental components from the load currents. Thus proposed is an algorithm for maintaining ideal phase source currents when the source voltages are amplitude-imbalanced. Hence this method can be used when the loads may be balanced/unbalanced, linear/non-linear and any distortion the source current must be sinusoidal.

Keywords: sinusoidal supply, neural network, back propagation, compensation, Point of common coupling, load