Power transformer fault diagnosis based on DGA combined with cloud model

In the power system, power transformer takes the responsibility of voltage transformation, energy distribution and transfer, the normal operation of the power transformer is an important guarantee of the system´s security, reliability, quality, economic operation. Early detection and treatment of the potential transformer failure have a very important practical significance. In this article, the sample data of dissolved gas in transformer is transformed into multiple qualitative cloud concepts through the cloud model, which is consistent with human´s cognition better and can be used to build the state space of DGA. The method simplifies the database by determining if the dissolved gas sample data belong to their corresponding clouds clusters, and proposes a method of transformer fault diagnosis by the use of an association rule mining algorithm to explore the cloud inference mechanism between the dissolved gas and the fault type. When DGA values exceeds the warning level, it would analyse the combination rule generator of cloud reasoning, and obtain a series of failure prediction results with stable tendency and the appropriate credibility. Eventually the predicted results with credibility larger than the threshold are given to the user. The instance of research results collected in this work proves the effectiveness and reliability of this method.