IDENTIFYING FAULTS IN 3 PHASE TRANSFORMER USING THE SIGNALS OF CURRENTS, LINE VOLTAGES AND MECHANICAL VIBRATION

Abstract

The electrical system is a complex system in both structure and operation, where any fault occurs, any element in the system affects power supply reliability, power quality and causes great economic damage. Therefore, the identification of the transformer state in the working process can help us to early diagnose fault patterns in 3-phase transformers, thereby reducing economic losses and improving reliability. The quality of electricity supplied to consumers is essential. This paper deals with identifying faults in 22/0.4kV distribution 3 phase transformers by using the ANSYS software to simulate samples of electrical data and mechanical displacement. The Levenberg - Marquadrt algorithm combined with the MLP neural network was used by the author to identify the MBA states. Neural network learning results have been successful and identified 05 fault states of the MBA, including the short 2 turns of phase B high voltage winding, short 5%, 10% of the total number of phase B high voltage winding, relax phase B windings and loose bolts that attach the MBA coils to the support beam). The identification results have an accuracy of 98.9%.