Condition monitoring and fault diagnosis of power transformer based on non-invasive measurement

In modern power systems, it is crucial to monitor and detect internal faults in power transformers promptly and accurately to ensure reliability and prevent disruptions. Failure to identify these faults promptly can reduce the transformer's lifespan, cause system disconnection, and compromise network stability. This paper introduces an innovative method for the discrimination, classification, and localization of internal short-circuit faults in power transformers, with a focus on three types of winding faults: turn-to-turn fault, series short circuits, and shunt short circuits. The proposed method introduces an online detection scheme utilizing the ΔV-I_in locus diagram, which leverages existing measurement devices without requiring additional hardware. A comprehensive winding model was developed in MATLAB to simulate insulation failures, and the method also analyzes the effects of faults and harmonic distortions on transformer performance. Features for fault discrimination and localization are derived from the ΔV-I_in locus and calculated using the practical design specifications of three power transformer models with capacities of 3 MVA, 5 MVA, and 7 MVA, operating at 50 Hz in a three-phase configuration. Experimental results on the 3 MVA transformer demonstrate that the formulated identifier efficiently detected all three types of insulation breakdown with an accuracy of 98.51%. Additionally, the fault localization algorithm achieved a fault location accuracy of approximately 93.28%. The findings indicate that the proposed approach is a robust and reliable tool for assessing the condition of power transformers.