Probabilistic neural-network-based protection of power transformer M. Tripathy, R.P. Maheshwari and H.K. Verma Abstract: An optimal probabilistic neural network (PNN) as a core classifier for fault detection and status indication of a power transformer has been presented. In this scheme, various operating con- ditions of a transformer are distinguished using signatures of the differential currents. The proposed differential protection scheme is implemented through two different structures of PNN, that is, one having one output and the other having five outputs. The developed algorithm is found to be stable against external fault, magnetising inrush, sympathetic inrush and over-excitation conditions for which relay operation is not required. For the test data of fault, it is found to operate successfully. The performance of proposed PNN and classical artificial neural network (ANN) has been com- pared. For evaluation of the developed algorithm, relaying signals for various operating conditions of a transformer are obtained by modelling the transformer in PSCAD/EMTDC. The algorithms are implemented using MATLAB. The results show the capability of PNN in terms of classification accuracy and speed in comparison to classical ANNs. 1 Introduction Power transformers are vital links in the chain of com- ponents constituting a power system. They are very expens- ive and are an important component of power system which facilitates the transmission of electric power at higher voltage over long distances. The continuous monitoring of power transformers can provide early warning of electrical failure and could prevent catastrophic losses as well as unscheduled outages of power supply. In view of this, avoiding damage to power transformers is vital; otherwise, continuity of power supply may be seriously disrupted. Furthermore, the repairing or replacing cost of a power transformer may be very high. Therefore, providing proper protection to power transformers is a crucial task. Accordingly, high demands are imposed on power-transformer-protective relays, that is (i) dependabil- ity (no missing operation), (ii) stability (no false tripping) and (iii) speed of operation (short fault clearing). Differential protection scheme is generally used as the primary protection of medium- and large-sized power trans- formers, in which the value of differential current greater than no-load value indicates an internal fault. The magnetis- ing inrush occurs in transformers whenever polarity and magnitude of residual flux do not agree with polarity and magnitude of instantaneous value of flux. Whenever there is a large and sudden change in the input terminal voltage of a transformer (either due to switching-in or due to recov- ery from external fault), large current is drawn by the trans- former from supply. Similar condition is encountered when a transformer is energised in parallel with another transformer already in service, and this situation is known as ‘sympath- etic inrush’. This large current from the source results in the saturation of the transformer core. Peaks of magnetising inrush current some time may rise very high to be of the order of 10 times that of full load current [1]. This large current from the source results in large differential current, which in turn causes the relay to operate undesirably. Owing to this reason, conventional differential relays are blocked for few initial cycles of energisation which makes the relay operation delayed on switching-in of the transformer on faults. Therefore, discrimination between magnetising inrush and internal fault condition is the key to improve the security of the differential protection scheme. Traditionally, two types of approaches are used for this purpose, that is, harmonic restraint (HR) and waveform identification (WI) concepts [2]. The HR is based on the fact that the second harmonic (sometimes the fifth) component of the magnetising inrush current is considerably larger than that in a typical fault current [1]. The literature reveals the extensive use of the HR method [3–6]. However, the HR-based method fails to prevent false tripping of relays because high second harmonic components during internal faults and low second harmonic components are generated during magnetising inrush for transformers having modern core material [7–10]. Therefore, the techniques based on detection of second/ fifth harmonic component are not useful to discriminate between the magnetising inrush and internal fault condition of modern power transformers. The second method consists of distinguishing magnetis- ing inrush and over-excitation condition from internal fault condition on the basis of WI concept [11, 12]. The development of advanced digital signal-processing tech- niques and recently introduced artificial neural network (ANN) provide an opportunity to improve the conventional WI technique and facilitate faster, secured and dependable protection for power transformers. As reported in literature, in recent years, different types of ANNs were used for power transformer protection because of its good generalisation ability and learning stab- ility with different topologies. Most of the ANNs were feed forward back-propagation neural network (FFBPNN) type # The Institution of Engineering and Technology 2007 doi:10.1049/iet-epa:20070009 Paper first received 4th January and in revised form 4th April 2007 The authors are with the Department of Electrical Engineering, Indian Institute of Technology Roorkee, Roorkee 247667, Uttarakhand, India E-mail: rudrafee@iitr.ernet.in IET Electr. Power Appl., 2007, 1, (5), pp. 793–798 793