Harmonic elimination and reactive power compensation through a shunt active power filter by twin neural networks with predictive and adaptive properties

A method for controlling an active power filter using Artificial Neural Network(ANN) is presented in this paper. This paper applies ANN based predictive and adaptive reference generation technique. Predictive scheme extracts the information of the fundamental component through an ANN that replaces a low pass filter. This ANN based low pass-filter is trained offline with large number of training set to predict the fundamental magnitude of load current. This predictive reference generation technique works well for clean source voltage. However, the performance deteriorates in case of distortion in source voltage and also with noise. To overcome this, an Adaline based ANN is applied after the operation of the predictive algorithm. It has been shown that the combined predictive-adaptive approach offers better performance. Simulation results and experimental results are presented to confirm the usefulness of the proposed technique..