Possibilities of ANN Application for Condition Monitoring of Power Transformers Using DGA

The transformer is one of the most important elements of the power system. Monitoring the behavior of the transformer is necessary in order to avoid catastrophic failures and costly interruptions in operation. Dissolved Gas Analysis (DGA) is a conventional technique that is widely used for diagnosing initial failures in transformers. All classical approaches have limitations because they cannot accurately diagnose all faults, that is, a large number of transformer states remain undefined. These methods are characterized by lower accuracy. In order to improve the accuracy of diagnostics in conventional dissolved gas analysis (DGA), an artificial neural network (ANN) has been applied to the conventional IEC and Rogers method of relationships. The chosen ANN approach of DGA gives very satisfactory results because it is now possible to classify all transformer states. The ANN was simulated and tested in the MATLAB R2022b environment with an accuracy of over 85%. By comparing ANN to the IEC and Rogers methods of relationships, which have an accuracy of 5060%, this approach successfully addresses the limitations associated with conventional methods of relationships. ANN has thus developed into an excellent tool in the field of diagnostics of power transformer failures.