Fault prediction based on dissolved gas concentration from insulating oil in power transformer using Neural Network

Abstract

Reliable and continued performance of power Transformer is the key to profitable generation and transmission of electric power. Failure of a large power transformer not only results in the loss of very expensive equipment, but it can cause significant guarantied damage as well. Replacement of that transformer can take up to a year if the failure is not disastrous and can result in tremendous revenue losses and fines. A Power Transformer in operation is subjected to various stresses like thermal stress and electrical stress, resulting in liberation of gases from the hydrocarbon mineral oil which is used as insulant and coolant. Dissolved Gas Analysis is a technique used to assess incipient faults of the transformer by analyzing specific dissolved gas concentrations arising from the deterioration of the transformer. DGA is used not only as a diagnostic tool but also to track apparatus failure. In this research work the dissolved gas values measured in PPM for a 230KV/110KV Power Transformer which are obtained from Electricity Board are used as references to the developed Neural Network. The Neural Network was trained and the gas concentration values for forthcoming years were predicted. Using the interpretation result of Key gas method, Rogers method and IS: 10593 method, the predicted gas concentration values compared and the fault of the Power Transformer were predicted. The trained Neural Network shows the good performance for the prediction of fault in a 230KV / 110KV Power Transformer.