Artificial neural network based fault prediction framework for transformers in power systems

Artificial neural networks are primitive learning systems that may primarily be used for classification and pattern recognition. Power distribution networks in India provide last mile connectivity between the utility and the consumers; and thereby reliability of these systems is of utmost importance for continuous power supply. In India, power distribution networks are typically old and periodic upgradation of these is economically not viable. Timely maintenance of the distribution components, specifically transformers is insufficient for absolute reliability. Also, faults cause temporary interruption in power supply during the time of repair and replacement. The best solution to avoid this involves predicting the temporal probability of faults, heuristically. This paper discusses the prediction of faults on transformers using artificial neural networks.