VOL. 5, NO. 9, SEPTEMBER 2010 ISSN 1819-6608 ARPN Journal of Engineering and Applied Sciences © 2006-2010 Asian Research Publishing Network (ARPN). All rights reserved. www.arpnjournals.com 1 FAST COMPUTING NEURAL NETWORK MODELING FOR FAULT DIAGNOSIS IN POWER SYSTEMS P. Chandra Sekhar 1 , B. V. Sanker Ram 2 and K. S. Sarma 3 1 Department of Electrical and Electronics Engineering, MGIT, Hyderabad, India 2 Department of Electrical and Electronics Engineering, JNTUH College of Engineering, Hyderabad, India 3 Department of Electrical and Electronics Engineering, Vardhaman College of Engineering, Hyderabad, India E-Mail: patsachandrasekhar@gmail.com pcs_76@rediffmail.com bvsram4321@yahoo.com ABSTRACT In this paper an approach for fault location based on online neural network is designed. The approach of learning the neural network based on the running fault values are trained for the suggested neural network. This approach result in running fault diagnosis based on the fault observation parameter based on the diagnosis tool. The approach is designed to run on running values of the distributed system so as to overcome the level of fault happening in a run time environment, which is not observed in case of the conventional neural controlling method. Keywords: adaptive learning, neural network, fault diagnosis, distributed power system 1. INTRODUCTION An overhead transmission line is one of the main components in every electric power system. The transmission line is exposed to the environment and the possibility of experiencing faults on the transmission line is generally higher than that on other main components. Line faults are the most common faults. They may be triggered by lightning strokes, trees falling across lines. Fog and salt spray on dirty insulators may cause the insulator strings to flash over, and ice and snow loadings may cause insulator strings to fail mechanically. When a fault occurs on an electrical transmission line, it is very important to detect it and to find it’s location in order to make necessary repairs and to restore power as soon as possible. The time needed to determine the fault point along the line will affect the quality of the power delivery. Therefore, an accurate fault location on the line is an important requirement for a permanent fault. Pointing to a weak spot, it is also helpful for a transient fault, which may result from a marginally contaminated insulator, or a swaying or growing tree under the line. Fault location in transmission lines has been a subject of interest for many years. During the last decade a number of fault location algorithms have been developed, including the steady state phasor approach, the differential equation approach and the traveling wave approach [4], as well as two-end [13] and one-end [14] algorithms. In the last category, synchronized [5] and non-synchronized [9] sampling techniques are used. However, two-terminal data are not widely available. From a practical viewpoint, it is desirable for equipment to use only one-terminal data. The one-end algorithms, in turn, utilize different assumptions to replace the remote end measurements. Most of fault locators are only based on local measurements. Currently, the most widely used method of overhead line fault location is to determine the apparent reactance of the line during the time the fault current is flowing and to convert the ohmic result into a distance based on the parameters of the line. It is widely recognized that this method is subject to errors when the fault resistance is high and the line is fed from both ends, and when parallel circuits exist over only parts of the length of the faulty line. Many successful applications of artificial neural networks (ANNs) to power systems have been demonstrated, including security assessment, load forecasting, control, etc. Recent applications in protection have covered fault diagnosis for electric power systems [8], transformer protection [2] and generator protection [7]. However, almost all of these applications in protection merely use the ANN ability of classification, that is, ANNs only output 1 or 0. Various approaches have been published describing applications of ANNs to fault detection and location in transmission lines [10],[11],[12]. In this paper, a single-end fault detector and three fault locators are proposed for on-line applications using ANN. A feed forward neural network based on the supervised back propagation learning algorithm was used to implement the fault detector and locators. The neural fault detector and locators were trained and tested with a number of simulation cases by considering various fault conditions (fault types, fault locations, fault resistances and fault inception angles) and various power system data (source capacities, source voltages, source angles, time constants of the sources) in a selected network model. 2. NEURAL NETWORK In this paper, the fully connected multilayer Feed Forward Neural Network (FFNN) was used and trained with a supervised learning algorithm called Back Propagation Algorithm (BPA). The FFNN consists of an input layer representing the input data to the network, some hidden layers and an output layer representing the response of the network. Each layer consists of a certain number of neurons; each neuron is connected to other neurons of the previous layer through adaptable synaptic weights w and biases b as shown in Figure-1.