An Overview of Transmission Line Protection by Artificial Neural Network : Fault Detection , Fault Classification , Fault Location , and Fault Direction Discrimination

Contemporary power systems are associated with serious issues of faults on high voltage transmission lines. Instant isolation of fault is necessary to maintain the system stability. Protective relay utilizes current and voltage signals to detect, classify, and locate the fault in transmission line. A trip signal will be sent by the relay to a circuit breaker with the purpose of disconnecting the faulted line from the rest of the system in case of a disturbance for maintaining the stability of the remaining healthy system. This paper focuses on the studies of fault detection, fault classification, fault location, fault phase selection, and fault direction discrimination by using artificial neural networks approach. Artificial neural networks are valuable for power system applications as they can be trained with offline data. Efforts have been made in this study to incorporate and review approximately all important techniques and philosophies of transmission line protection reported in the literature till June 2014.This comprehensive and exhaustive survey will reduce the difficulty of new researchers to evaluate different ANN based techniques with a set of references of all concerned contributions.


Introduction
There is no fault-free system and it is neither practical nor economical to build a fault-free system.The various cases of abnormal circumstances such as natural events, physical accidents, equipment failure, and misoperation generate faults in the power system.The consequences of faults are traumatic amplification of current flow, increasing heat produced in the conductors leading to the major cause of damage.The actual magnitude of fault depends on resistance to flow and varied impedance between the fault and the source of power supply.Total impedance comprises of fault resistance, resistance and reactance of line conductors, impedance of transformer, reactance of the circuit, and impedance of generating station.The conventional distance relay settings are based on a predetermined network configuration with worst fault outcomes [1][2][3][4][5][6].As the neural network based algorithm has more adaptability and is likely to be more accurate, various researchers used it for power system protection which is the main focus of this study.A number of prime purposes and applications of ANN are accessible in the literatures; those will assist to recognize the perception of accepting it as a tool for fault detection, classification, and localization on transmission line of the power systems.Various journals, conference papers, books, online libraries, and databases were researched and reviewed for gathering proper information to develop a broad insight and comprehension of the subject being studied.Both scholarly and nonscholarly articles were surveyed and considered from databases like IEEE, Scopus, Google Scholar, Academia Search Premier, Pro-Quest, EBSCO, and other relevant websites.
The paper is organized as follows.In Sections 2 and 3, a brief introduction of power system faults and artificial neural networks is provided, Section 4 is about distance protection by ANN method; in Section 5, ANN and its application for protecting transmission line are illustrated.Section 6 deals with the conclusions drawn from this survey followed by acknowledgments and references.

Faults in Power System
Fault is an unwanted short circuit condition that occurs either between two phases of wires or between a phase of wire and ground.Short circuit is the most risky fault type as flow of heavy currents can cause overheating or create mechanical forces which may damage equipments and other elements of power system [1][2][3][4][5][6].
2.1.Categories of Faults.Faults also can be classified into three types, that is, symmetrical faults, unsymmetrical faults, and open circuit faults.

Symmetrical Faults.
The fault that results in symmetrical fault currents (i.e., equal currents with 120 displacements) is known as a symmetrical fault.Three-phase fault is an example of symmetrical fault where all three phases are short circuited with or without involving the ground.

Unsymmetrical Faults.
Examples of different unsymmetrical faults are single phase to ground, two phases to ground, and phase to phase short circuits.The details of these shunt fault types that can occur in transmission line are described as follows.
(1) Single Phase to Ground (L-G) Fault.L-G is a short circuit between any one of phase conductors and earth (prevalence is 70%-80%).It may be caused either by insulation failure between a phase conductor and earth or breaking and falling of phase conductor to the ground.
(2) Two Phases to Ground (L-L-G) Fault.L-L-G is a short circuit between any two phases and earth (prevalence is 10%-17%).
(3) Phase to Phase (L-L) Fault.L-L is a short circuit between any two phases of the system (prevalence is 8%-10%).
(4) Three-Phase (L-L-L) Fault.L-L-L is a short circuit between any two phases of the system (prevalence is 2%-3%).

Open Circuit
Faults.This type of fault is caused by breaking of conducting path.Such fault occurs when one or more phases of conductor break or a cable joint/jumper (at the tension tower location) on an overhead line fails.Such situations may also arise when circuit breakers or isolators open but fail to close in one or more phases.During the open circuit of one of the two phases, unbalanced current flows in the system, thereby heating rotating machines.Protective schemes must be provided to deal with such abnormal conditions.

Artificial Neural Network
Artificial neural network (ANN) has been equipped with distinctiveness of parallel processing, nonlinear mapping, associative memory, and offline and online learning abilities.The wide uses of ANN with its conquering outcomes make it  an effective diagnostic mean in electric power systems.Its versatility with multitude applicability can be seen in other areas of science and engineering research [7][8][9].It is a complex network of interconnected neurons where firing of electrical pulses via its connections leads to information propagation.ANN is trained by using prior chosen fault samples as input and set of fault information as output for fault diagnosis application.Neural networks are comprised of primarily three basic learning algorithms such as supervised learning, unsupervised learning, and reinforced learning.Among these supervised learning is most commonly used and is also referred to as learning with a teacher.This is applied when the target is having identified value and is associated with each input in the training set [7]. Figure 1 represents the supervised architecture of ANN.
Error back propagation (BP) neural network was applied by Chan [10] for diagnosis of fault in power system.However slow speed training and the shortcomings of local optima lead to the introduction of additional momentum factor for problem solving.Radial basis function (RBF) neural network has a faster learning speed and the ability of arbitrary function approximation.Bi et al. presented a novel RBF neural network for estimating section of fault.Their simulation results of 4bus test system shown that the capability of RBF neural network in grid fault diagnosis was better than the conventional BP neural net [11].
For solving improper problems, neural network topologies are to be altered and there is a need to retrain the network.Cardoso et al. [12] used the true capacity of multilayer perception (MLP) and generalized regression neural network (GRNN) for fault estimation in electric power system.GRNN is having the advantage of faster learning, global optimum, and lower requirement of comprehensive sample.They fed the failure information into MLP and the resultant outcome was given as output to GRNN.They also compared ANN fault diagnosis methods with expert system diagnostic methods and found that ANN based methods may evade the formation of expertise, expert heuristic knowledge, and expression and hence save tedious work.

Distance Protection by ANN
The fundamental principle of distance protection is that the apparent impedance seen by the relay reduces considerably in case of line fault.A fault is indicated if the ratio of apparent impedance to the positive sequence impedance is less than unity.This scheme of protection is inherently directional and used by impedance and Mho relays.This paper focuses upon the studies of distance protection scheme applying ANN approach.
Adaptive relaying was introduced for widespread applications including incorrect or fault operations measurement.The learning capacity of ANN from input and output patterns extended its applicability in several adaptive protection schemes.Khaparde et al. [13] applied adaline neural network model in offline mode for protective relaying operation of transmission lines.They also proposed adaptive distance protection by using ANN [14].They have applied MLP model to reduce misoperation of a relay.Girgis et al. [15] presented a method for the computation of fault location in twoand three-terminal high voltage lines which is based on digital computation of the three-phase current and voltage 60/50 Hz phasors at the line terminals.For evaluation of the convergence and distinctive solution, this method was tested by electromagnetic transient (EMPT) generated transient data from a steady state fault analysis.Qi et al. [16] proposed ANN approach for distance protection of power system by taking trained data from simulation of a simple power system under load and fault conditions.According to them conventional distance relays might not function properly under certain conditions such as nonlinear arc resistance, high impedance fault, and variable source impedance.However if such relays are implemented with ANN, such issues can be addressed.Khaparde [17] again proposed an adaptive scheme of distance protection using an artificial neural network.Lai [18] implemented an adaptive protection scheme by ANN approach for classification purpose.They have considered conditions of high impedance fault (hard detection because of minute fault current) and variable source impedance.Coury and Jorge [19] proposed distance protection using ANN for transmission lines utilizing the magnitudes of three-phase voltage and current phasors as inputs.ANN based approach for improving the speed of a differential equation based distance relaying algorithm was developed by Cho et al. [20].Several researchers illustrated various methodologies for improvements in fault distance computation [21][22][23][24][25].
Venkatesan and Balamurugan [26] developed neural network simulator for identifying the optimum ANN structure necessary to train the data and implement the ANN in hardware.However there is no precise rule for selection of the number of hidden layers and neurons per hidden layer.So it is not certain whether or not the ANN based relay gives the optimum output, for maintaining the integrity of the boundaries of the relay characteristics.Pradhan et al. [27] proposed a high speed distance relaying scheme based on RBF neural network due to its capability of distinguishing faults with data falling outside the training pattern.A sequential procedure for distance protection using a minimal RBF neural network for determining the optimum number of neurons in the hidden layer without resorting to trial and error was illustrated by Dash et al. [28].Authors [29] trained multilayer feed-forward architecture with two inputs and three-trip or no-trip output signals based approach and used BP technique for three-zone distance protection of transmission lines.The first output was used for main protection of the transmission line section, whereas the other two outputs provide backup protection for the adjacent line sections.The input features extracted by discrete-Fourier transform from the fundamental frequency voltage and current magnitudes.
Santos and Senger [30] developed and implemented of a unique ANN based algorithm for transmission lines distance protection.Their algorithm can be used in any transmission line despite of its configuration or voltage level and also does not require any topology adaptation or parameters adjustment when applied to varied electrical systems.Vaidya and Venikar [31] illustrated an ANN based distance protection scheme for long transmission lines by considering the effect of fault resistance of single line to ground fault type.They have utilized the magnitudes of resistance and reactance as inputs for classifying unknown patterns.A novel distance protection approach for detection and classification stages based on cumulants and neural networks was developed by Carvalho et al. [32].

Application of ANN on Transmission Line Protection
This section presents the studies on application of ANN for fault detection, classification, location, direction discrimination, and faulty phase selection on transmission line.

Studies on "Fault Detection and Classification".
It is necessary to identify the fault and classify its type with the aim of establishing safety and stability of the power system.Lim and Shoureshi [33] developed ANN based monitoring system for health assessment of electric transmission lines.Their system showed satisfactory performance in fault classification by using both MLP (multilayer perceptron) and ART (adaptive resonance theory) classifiers.A comparative study of different ANN based fault detection and classification schemes  is given in Table 1 highlighting the methods used, their response time, and ANN features along with its accuracy.

Studies on "Fault Detection and Classification and Location".
It is extremely essential to identify and locate the transmission line faults for maintaining the proficient and trustworthy operation of power systems.For estimation of the fault location, there are a number of mathematical and intelligent methods accessible in the literature.However, the broad variations in operating conditions such as system loading level, fault inception instance, fault resistance and dc offset, and harmonics contents in the transient signal of the faulty transmission line give rise to unsatisfactory results.Amjady [67] diagnosed on line power systems fault by a new expert system.Their diagnostic system can be applicable for single or multiple faults and practically examined with real events on a model power system.Several researches have been carried out to detect, classify, and locate the fault on transmission lines by using neural networks by Oleskovicz et al. [68], Coury et al. [69], Othman et al. [70], Mahanty and Dutta Gupta [71], Gracia et al. [72], Lin et al. [73], Jain et al. [74], Othman and Amari [75], Gayathri and Kumarappan [76], Tayeb and Rhim [77], Jiang et al. [78,79], Warlyani et al.Step 1: RBF based SVMestimates the initial distance of fault using the positive sequence voltages and currents of faulty phases.
Step 2: Improving the final estimation of thisdistance using SCALCG based neural network with the high frequency range characteristics.
The maximum error of fault location was limited to 1. 93   (ii) The proposed scheme allows the protection engineers to increase the reach setting (i.e., a greater portion of line length can be protected)  (10,5) needs 30% less number of epochs to reduce the error to 5% of its initial value [88] and to 15% in [89] (i) Authors found that the new 20-input network performed better than the earlier 30-input network (ii) Here, real world fault data had been recorded by Alberta Power Ltd. on the 240 kV transmission systems [89] Sanaye-Pasand and Malik (1998) [90] Elman network for fault direction estimation 0 ms-12 ms (i) Elman network is a two-layer feed-forward network with the addition of a recurrent connection from the output of the hidden layer to its input (ii) The network outputs which fall above 0.5 and below −0.5 are interpreted as forward and backward faults, respectively (iii) 12 inputs, 12 hidden neurons, and one output neuron (iv) For both hidden and output layers: tansig function 40 different forward and backward faults at the relay location were applied to the system and the network's performance was investigated; in all cases except one 3 phases to ground fault, the directional module performed correctly  Advances in Artificial Neural Systems [80], Yadav et al. [81], Teklic et al. [82], and Jamil et al. [83].A comparative study of different ANN based fault detection, classification, and location schemes is given in Table 2 highlighting the methods used, their response time, and ANN features along with remarks.
5.3.Studies on "Fault Direction Discrimination".Fault direction estimation on transmission line is very crucial for enhancing the performance of power system.Advancement of huge generating stations and highly interconnected power systems entails less fault clearing times.The approach of ANN has been positively utilized for the improvement of many of the standard functions that are operated in transmission lines.The accuracy of an electromechanical, static, or a microprocessor based distance relay is affected by different fault conditions and network configuration changes.Hence the direction of the fault should be discriminated to maintain the normal operation of the power system.Dalstein et al. have used ANN method to estimate the fault location process by means of directional discrimination.They have proposed a neural network to estimate the direction of the fault [84,85].Authors [86] employed neural network for designing two different fault direction discrimination modules for high speed transmission line and found that fault direction can be identified quickly and accurately from their results.Table 3 highlights the different schemes [87][88][89][90][91][92][93][94][95][96] used for fault direction estimation with its response time and features of ANN along with remarks.

Studies on "Faulty
Phase Selection".Fault phase selection, an imperative part of fault diagnosis, is carried out by measuring faulty line parameters.Different power system faults such as LG, LL, LLG, LLL, and LLLG on a protected transmission line should be detected, classified, and located and faulty phase should be selected swiftly for performing the normal system operation.The summarized study of different ANN based fault phase selection schemes is given in Table 4 highlighting the methods used, their response time, and ANN features along with remarks [97-105].

Conclusion
There are widespread applications of ANN in power system protection, but this paper intensively analyzed few of them.Novel tools and techniques are preferred to maintain power system reliability and security within a satisfactory level for improvement of the performance of digital protective relays, renovation of power industry, and stability of the transmission lines.ANN is found to be robust, accurate, and efficient approach for transmission line fault detection, classification, localization, direction discrimination, and faulty phase selection.A comparative study of different schemes for fault detection, fault classification, fault location, fault direction estimation, and faulty phase selection has been discussed in detail.An extensive survey of the published studies on the subject of ANN application to transmission line protection is specified in this paper which will be beneficial for researchers for further research and development in this field.

Table 1 :
Comparative study of ANN based "fault detection and classification" schemes.

Table 2 :
Comparative study of ANN based "fault detection and classification and location" schemes.

Table 3 :
Comparative study of ANN based "fault direction discrimination" schemes.

Table 4 :
Comparative study of ANN based "faulty phase selection" schemes.SVM1 for fault phase selection: Postfault current and voltage samples for one-fourth cycle (five samples) as input (ii) SVM2 for ground detection: zero sequence components of fundamental, third and fifth harmonic components of the postfault current signal (iii) Sampling frequency: 1.0 KHz Data collection: PCL-208 data acquisition card, which uses 12-bit successive approximation technique for A/D conversion; installed on a PC (P-4) with a driver software routine written in C++ having six I/O channels with input voltage range of +5 V (i) The test results are compared with those of the radial basis function network (RBF) and were found to be superior with respect to efficiency and speed (ii) The classification test results from SVMs are accurate for simulation model and experimental setup and thus provide fast and robust protection scheme for distance relaying in transmission line