لخّصلي

The electrical power system consists of so many different complex dynamic and interacting elements; which are always prone to disturbance or an electrical fault. The use of high capacity electrical generating power plants and concept of grid; i.e. synchronized electrical power plants and geographical displaced grids; required fault detection and operation of protection equipment in minimum possible time so that the power system can remain in stable condition. The faults on electrical power system transmission lines are supposed to be first detected and then be classified correctly and should be cleared in least fast as possible time. The protection system used for a transmission line can also be used to initiate the other relays to protect the power system from outages. A good fault detection system provides an effective; reliable; fast and secure way of a relaying operation.

The application of a pattern recognition technique could be useful in discriminating the faulty and healthy electrical power system. It also enables us to differentiate among three phases which phase of a three phase power system is experiencing a fault. The artificial neural networks (ANNs) are very powerful in identifying the faulty pattern and classification of fault by pattern recognition. There are lot of algorithms based upon ANN have been developed; tested and implemented practically in electrical power systems (Dalstein and Kulicke 1995; Bouthiba 2004; Venkatesan and Balamurugan 2007; Lin et al. 2001). Whei_Min Lin et al. (2001) presents the method based on pattern recognition; but the method is very complex. Angel L. Orille Fernandez et al. (2002) presented the finite impulse response (FIRANN) method to detect and classify the fault. The author uses the impulse response of voltages and currents; which limits its applications. M. S. Abdel Aziz et al. (2012); presents adaptive neuro_fuzzy inference system (ANFIS) in power distribution system. The Fourier transform is used with ANFIS; which has its inherent disadvantages. Jayabharta Reddy and Mohanta (2007) proposes and wavelet transform and fuzzy logic based algorithm for fault classification; but the fuzzy logic gives poor performance at boundary line cases. Alanzi et al. (2014) proposed the fault detection by unconventional a synchronized method; but decision making is left untouched. An efficient and reliable protection method should capable to perform more than satisfactory under various system operating conditions and different electrical network parameters. As far as ANNs are considered they exhibit excellent qualities such as normalization and generalization capability; immunity to noise; robustness and fault tolerance. Therefore; the declaration of fault made by ANN_based fault detection method should not be affected seriously by variations in various power system parameters. Therefore so many ANN_based techniques have been developed and employed in power system. The results obtained from these methods are encouraging (Kezunovic et al. 1996; Rizwan et al. 2013). Some algorithms based upon ANN for location of faults and relay architecture for protection of transmission line are also suggested by the researchers (Sanaye_Pasand and Kharashadi_Zadeh 2006; Lahiri et al. 2005). In this paper; a new algorithm based upon ANN is proposed for fast and reliable fault detection and classification. The various electrical transient system faults are modelled; simulated and an ANN based algorithm is developed for recognition of these faulty patterns. The performance of the proposed algorithm is evaluated by simulating the various types of fault and the results obtained are encouraging. It is observed that the algorithm developed is capable to perform fast and correct classification for different combinations of faulty conditions; e.g. fault type; fault resistance; fault location and short circuit MVA of the system.

The paper is divided into four categories. The section one is background; which discusses the vital points of fault detection. Second section gives over view of the artificial neural network and its training methods adopted. Third section gives the details of transmission line model and its simulation; the last fourth section is conclusion.

Artificial neural network
Artificial neural network (ANN) can be applied to fault detection and classification effectively because it is a programming technique; capable to solve the non linear problems easily. The problems in which the information available is and in massive form can be dealt with. Also; the ANNs are able to learn with experiences; i.e. by the examples (Chaturvedi 2008). They are widely accepted and used in the problem of fault detection and fault classification because of the following features:

Number of transmission line configuration are possible as there can be any possibility from short length; long length; single circuit transmission line to double_circuit transmission lines; etc.

There are several methods to simulate the network with different power system conditions in a fast and reliable manner;

The conditions of the electrical power system change after each and every disturbance. Hence a neural network is capable to incorporate the dynamic changes in the power systems.

The ANN output is very fast; reliable and accurate depending on the training; because its working depends upon a series of very simple operations.

The algorithm which employed ANNs programming offers many advantages; but it also suffers with many disadvantages; which are very complex in nature. Some of the important factors are the selection of type of network; architecture of the network (which includes the selection of number of layers; number of neurons in each layer; selection of activation functions; learning algorithms parameters etc.); termination criteria etc. There are various parameters like values of the pre fault and post fault voltages and currents of the respective three phases in steady state required for precise fault detection and classification.

The values of the pre fault and post fault voltage and current of respective three phases are very different and are governed by the type of fault. Thus; the fault classification method required a neural network that allows it to determine the type of fault from the patterns of pre fault and post fault voltages and currents; which are generated from the values measured from a three phase transmission line of an electrical power system at one terminal. The neural network is based upon the total six number of inputs; i.e. the voltages and currents of respective three phases. The neural network is trained by using these six inputs. The total number of outputs of the neural network is four in numbers; i.e. three phases A; B; C and fourth is ground of three phase transmission line.

Back propagation neural network (BPNN)
In the Back propagation neural network (BPNN) the output is feedback to the input to calculate the change in the values of weights. One of the major reasons for taking the back propagation algorithm is to eliminate the one of the constraints on two layers ANNs; i.e. similar inputs lead to the similar output. The error for each iteration and for each point is calculated by initiating from the last step and by sending calculated the error backwards. The weights of the back_error_propagation algorithm for the neural network are chosen randomly; feeds back in an input pair and then obtain the result. After each step; the weights are updated with the new ones and the process is repeated for entire set of inputs_outputs combinations available in the training data set provided by developer. This process is repeated until the network converges for the given values of the targets for a pre defined value of error tolerance. The entire process of back propagation can be understood by Figure 1. The back_error_propagation algorithm is effectively used for several purposes including its application to error functions (other than the sum of squared errors) and for the calculation of Jacobian and Hessian matrices. This entire process is adopted by each and every layer in the entire the network in the backward direction (Haykin 1994). The proposed algorithm uses the Mean Square Error (MSE) technique for calculating the error in each iteration.