The application of compound method in on-line monitoring system to predict and judge transformer fault

This paper put forward a compound method to predict and judge the faults of transformer from the gas that dissolved in oil. First, the database system was developed to store the data of gases that was timing collected from senior machine. Then, the improved grey model would read the data after it filtered, calculate and forecast an equal interval one. At the same time, the residual error and the corrected parameters was formed. The new data and the corrected parameter were read and computed by the self-learning neural networks, and the final predicted value was formed. And then another neural network judgment system would read the value, and locate the type of faults.