Classification of Disturbances in Electrical Signals Using Neural Networks

• Authors: Carlos León;Antonio López;Juan C. Montaño;Iñigo Monedero
• Venue: IWANN '01 Proceedings of the 6th International Work-Conference on Artificial and Natural Neural Networks: Bio-inspired Applications of Connectionism-Part II
• Year: 2001

This paper describes a currently project accomplished by the authors in the area of Power Quality (PQ) using artificial neural networks (ANN). The efforts are oriented to obtain a product (Power disturbances monitor for three-phase systems) that permits a real time detection, automatic classification, and record process of impulsive or oscillatory voltage transients, long term disturbances, and waveform distortions in electrical three-phase AC signals. To classify the electrical disturbances, we consider using a fully connected feedforward ANN with a backpropagation learning method based on Generalized Delta Rule. In order to select the best alternative more than 200 network architectures were tested. Long-term disturbances, like swells or long-duration interruptions, have been detected using a method based on the test of the RMS value of the signal. Short-term disturbances, like sags, are detected by sampling a cycle of the electrical signal, and waveform distortions are detected using the main harmonics of the signal. To train the ANN we have developed a three-phase virtual generator of electrical disturbances. In order to compress the ANN input data we use the Wavelet Transform.