Theoretical Approach for Fault Prognosis in Electrical Power Transformers using High Frequency Signals and Artificial Intelligence Techniques

This paper proposes a prognostic method capable of identifying malfunctions in electrical power transformers using their high-frequency models and specific artificial intelligence techniques. From a general point of view, this approach is based on the recognition of parametric faults by processing frequency response measurements. The starting point of the work is to develop an equivalent lumped circuit in which parameter variations can be used to simulate different malfunctions and partial losses of functionality. To achieve this purpose, basic theoretical considerations on transformer models and specific high-frequency effects extracted from the literature are combined. The variations of parasitic components between primary and secondary windings and between phases in a three-phase power transformer are studied. The main objective of the paper is to propose a classification method based on the use of artificial intelligence algorithms capable of recognizing such variations. In this work, all the most important theoretical aspects are explored in depth, from the physical meaning of the lumped parameters to the possible real implementation of the prognostic method. Therefore, it can be used as a suitable basis for future field experiments with real measurements.