A fast and reliable transformer protection system based on the transformer magnetizing characteristics and artificial neural networks

Abstract

Transformers are one of the most important elements of power systems. Most transformers are equipped with protection systems to avoid damage to the transformers [1]. As any outage of the transformer have severed technical and economical consequences for the network, so implementing fast relaying algorithms for transformer protection devices to satisfy high reliability of the whole power systems is very important. In designing high speed protection systems, the fast discrimination of magnetizing inrush current is very important to prevent the false tripping of relays. The conventional method of inrush current detection [2] for transformer protection based on the value of the second harmonic component of the differential current which requires a long time exceeding one cycle. This paper discriminates between the inrush current state and the internal fault condition based on the transformer magnetizing characteristics within one short cycle. Simulation using EMTP-ATP program on a three phase transformer are carried out. Study cases of inrush current at different switching instants and for all types of internal faults for different phases and on different percentages of transformer windings are simulated. The results obtained from EMTP-ATP package for all simulated conditions used as a training data to artificial neural network (ANN). This helps the differential relay to recognize inrush and internal fault and give the trip signal in case of internal fault only.