Recent Techniques Used in Transmission Line Protection

ABSTRACT

This article presents a survey of the developments in digital relays for protection of transmission lines. For a modern power system, selective high speed clearance of faults on high voltage transmission lines is critical and this survey indicates the efficient and promising implementations for fault detection, classification and fault location in power transmission line protection. The work done in this area favor computerized relays, digital communication technologies and other technical developments, to avoid cascading failures and facilitate safer, secure and reliable power systems. Efforts have been made to include almost all the techniques and philosophies of transmission line protection reported in the literature up to October 2010.

The focus of this article is on the most recent techniques, like artificial neural network based and phasor measurement unit-based concepts as well as other conventional methods used in transmission line protection.

1

CHAPTER-1

INTRODUCTION

1.1 General

Transmission lines are among the power system components with the highest fault incidence rate, since they are exposed to the environment. Line faults due to lightning, storms, vegetation fall, fog and salt spray on dirty insulators are beyond the control of man. The balanced faults in a transmission line are three phase shunt and three phases to ground circuits. Single line-to-ground,line-to-line and double line-to-ground faults are unbalanced in nature. On a transmission system the protective relaying system is incorporated to detect the abnormal signals indicating faults and isolate the faulted part from the rest of the system with minimal disturbance and equipment damage. This survey attempts to cover the various developments in digital relays for transmission line protection reported in the literature up to October 2010 and point to some of the references showing promising directions. Rockefeller first presented the implementation of digital relaying in 1969 (Rockefeller, 1969). The advances in the very large scale integrated (VLSI) technology and software techniques led to the development of microprocessor based relays that were first offered as commercial devices in 1979 (Sachdev, 1979). Selective, high speed clearance of faults on high voltage transmission lines is critical to the stability of the highly complex, modern power system. In this respect, lot of work has been done to improve the performance of digital protective relays and in the use of intelligent techniques for analysis of faults and protective relay operations. Distance relaying principle, due to their high speed fault clearance compared with the over current relays is a widely used protective scheme for the protection of high and extra high voltage (EHV) transmission and sub-transmission lines.

A distance relay estimates the electrical distance to the fault and compares the result with a given threshold, which determines the protection zone. In terms of hardware, distance relays have evolved from electromechanical relays to static relays and to microprocessorbased (digital) relays. When a fault occurs in an electrical transmission line, the distance relays detect the faulty line and type of fault but they may under reach/over reach depending upon pre-fault loading, fault resistance and remote end in-feeds. The impedance estimated by a digital distance relay reduces with the increase in the speed at which the estimate is obtained. Hence an impedance relay with a specified reach setting cannot operate at arbitrarily high speeds (Thorp et al., 1979). The first installation of digital

2

computer for relaying began in 1960’s which made it possible to store information so that the relay engineer can control the reach characteristics of a distance relay to suit the application and develop fault location algorithms (Gilcrest et al., 1972;Rockefeller et al., 1972; IEEE Std.37.114-2004, 2005). Such digital fault locators calculate the reactance of a faulty line estimated from the computation of voltage and current phasors at the line terminals (Sachdev et al., 1988; Adu, 2004). But these fault location methods need some simplifying hypothesis to allow the fault distance calculation, affecting the accuracy of the results. The one terminal approach is simple and easy to implement (Takagi et al., 1982; Guobing et al., 2009; Xianyning Lin et al., 2009; Eduardo De et al., 2004) although the two-end algorithms which process signals from both terminals of the line are superior in comparison to the one-end approaches (Girgis et al., 1992).

In the 70’s research was concentrated on “ultra high speed protection” based on the travelling wave. The post fault wave forms in the first one or two cycles after the occurrence of a fault contain high frequency transient wave fronts. Based on the analysis of these transient state signals the fault location can be calculated within a few milliseconds of the fault initiation. Different algorithms proposed for implementation of travelling wave distance protection are reported by Desikachar et al., (1984), Shehab- Eldin et al., (1988), Ancell et al., (1994), Ernesto Vazquez-Martinez, (2003), Dong et al., (2009). The “positional protection” utilizes the transit times of the high frequency fault generated transients to identify the faulted line section (Bo et al., 2000). It has been noted by most of the researchers that the travelling wave based method does not perform well for faults close to the relaying point and for faults with small fault inception angle, besides they require a very high sampling rate and their implementations aremore costly than implementation of impedance techniques. As the complexity of the power network increases, the transmission line protection and control must be based on real time power system changes and it must be at high speeds to ensure that the power system will not run into transient stability problems. Several papers have considered the real time power system changes and have reported about accurate, fast faulty phase selection and fault location (Horowitz et al., 1988; Dadash Zadeh et al., 2009; Dash,1987; Wang et al., 1997; Bo, 1988; Bo et al., 2003; Essay et al., 2001; El-Arroudi et al., 2004; Pathirana et al., 2005; Bockarjova et al., 2006; Cook, 1986; kang et al., 2009; Johns et al., 1990; Zamora et al., 1996; Gopalakrishnan et al., 2000; He etal., 2006; Parikh et al., 2007; Abdelaziz et al., 2005; Girgis et al., 1998; Samantaray, 2009; Xu et al., 2008; Samantaray et al., 2008; Salat et al., 2004; Ravikumar et al., 2008; Shrivastava et al., 2007; Sanjay Dambhare et al., 2009; Law et al., 2008). In the late 80’s synchronized measurement technology emerged as a promising prospect in achieving real time protection. With global positioning system (GPS), digital measurement at different line terminals can be performed synchronously (Crossley et al., 1998; Bo et al., 2000). They are more accurate than distance relaying algorithms which are affected with inadequate modeling of transmission lines and parameter uncertainty due to line aging, line asymmetry and environmental factors. The Pharos Measurements Units (PMU) are the most widely used synchronized measurement devices for power system applications, whose measurements are synchronized with respect to a GPS clock and PMU-based fault locators are more accurate than the method based on unsynchronized phases (Jiang et al., 2000; Lin et al., 2002; Xu et al., 2008; Lin et al., 2004; Brahma et al., 2004).

3

The unsynchronized approaches are cheaper, since there is no need to use GPS and are not affected by errors due to different sampling rates or phase shifts introduced by the different recording devices and transducers. Such impedance based fault location methods have been presented by Novosel et al., (1996), Izykowski et al., (2006), Dascastagne et al., (2008), having negligible fault location error if pharos and transmission line parameters are accurate. Although the use of GPS, pharos measurement units (PMUs), digital communication technologies, high precision signal transducers have facilitated accurate protection of power system over a wide area, they are subjected to software insecurity and communications latency.

There is a need for the measuring algorithms that have the ability to adapt dynamically to the system operating conditions such as changes in the system configuration, source impedances and fault resistance. Keeping this in view the trends since 90’s, intelligent techniques are under investigation to increase reliability, speed and accuracy of existing digital relays based onArtificial Neural Network (ANN), Fuzzy Logic (FL), Fuzzy-Neuro and Fuzzy Logic-Wavelet based systems..

TRANSMISSION LINE PROTECTION PRINCIPLES

1.1 General

Transmission lines are a vital part of the electrical distribution system, as they provide the path to transfer power between generation and load. Transmission lines operate at voltage levels from 69kV to 765kV, and are ideally tightly interconnected for reliable operation. Factors like de-regulated market environment, economics, rightof- way clearance and environmental requirements have pushed utilities to operate transmission lines close to their operating limits. Any fault, if not detected and isolated quickly will cascade into a system wide disturbance causing widespread outages for a tightly interconnected system operating close to its limits. Transmission protection systems are designed to identify the location of faults and isolate only the faulted section . The key challenge to the transmission line protection lies in reliably detecting and isolating faults compromising the security of the system.

4

FACTORS INFLUENCING LINE PROTECTION

1.1 General

The high level factors influencing line protection include the criticality of the line (in terms of load transfer and system stability), fault clearing time requirements for system stability, line length, the system feeding the line, the configuration of the line (the number of terminals, the physical construction of the line, the presence of parallel lines), the line loading, the types of communications available, and failure modes of various protection equipment. The more detailed factors for transmission line protection directly address dependability and security for a specific application. The protection system selected should provide redundancy to limit the impact of device failure, and backup protection to ensure dependability. Reclosing may be applied to keep the line in service for temporary faults, such as lightning strikes. The maximum load current level will impact the sensitivity of protection functions, and may require adjustment to protection functions settings during certain operating circumstances. Single-pole tripping applications impact the performance requirements of distance elements, differential elements, and communications schemes.

The physical construction of the transmission line is also a factor in protection system application. The type of conductor, the size of conductor, and spacing of conductors determines the impedance of the line, and the physical response to short circuit conditions, as well as line charging current. In addition, the number of line terminals determines load and fault current flow, which must be accounted for by the protection system. Parallel lines also impact relaying, as mutual coupling influences the ground current measured by protective relays. The presence of tapped transformers on a line, or reactive compensation devices such as series capacitor banks or shunt reactors, also influences the choice of protection system, and the actual protection device settings

FIG-1

5

MULTILINE APPLICATION ADVANTAGES

1.1 General

Before considering using a GE Multiline relay for a specific transmission line protection application, it is important to understand how the relay meets some more general applicationrequirements for simplicity, security, and dependability. GE Multiline relays provide simplicity and security for single pole tripping, dependability for protection communications between lineterminals, security for dual-breaker line terminals, and simplicity and dependability of redundant protection schemes.

1.2 Single-Pole Tripping

Single pole tripping using distance protection is a challenging application. A distance relay must correctly identify a singlephase fault, and trip only the circuit breaker pole for the faulted phase. The relay also must initiate the recloser and breaker failure elements correctly on the fault event. The distance elements protecting the unfaulted phases must maintain security during the open-pole condition and any reclosing attempts.

The D90Plus Line Protection System and D60 Line Distance Relay use simple, dedicated control logic for single pole tripping applications. This control logic uses a Phase Selector, Trip Output and Open Pole Detector in conjunction with other elements as shown in the simplified block diagram. The Trip Output is the central logic of single pole tripping. The Trip Output combines information from the Open Pole Detector, Phase Selector, and protection elements to issue a single pole or three pole trip, and also to initiate automatic reclosing and breaker failure. The Phase Selector is the key element for maintaining the security of single pole tripping applications, quickly and accurately identifying the faulted phase or phases based on measured currents and voltages, by looking at the phase angles between the positive sequence, negative-sequence, and zero-sequence components.

The Open Pole Detector ensures the relay operates correctly during a single pole trip, placing the relay in an open pole condition when a single pole trip command is issued, or one pole of the circuitbreaker is open. The Open Pole Detector asserts on a single pole trip command, before the circuit breaker pole actually opens, to block protection elements that may misoperate under an open pole condition, such as negative sequence elements, under voltage

6

protection, and phase distance elements associated with the faulted phase (for example, AB and CA elements for an AG fault).The Open Pole Detector also resets and blocks the Phase Selector so the other distance elements may operate for evolving faults. The Open Pole Detector also accounts for line charging currentand for weak infeed conditions. Once the Open Pole Detector operates, a further trip will cause the Trip Output to declare a three pole fault, indicating either an evolving fault condition or a reclose onto a permanent phaseto- ground fault. This total logic simplifies the setting of the D60 for single pole tripping, and ensures dependable and secure operation when faced with single line-to-ground faults. The L90 Line Differential Relay and the L60 Line Phase Comparison Relay are both phase-segregated, current only relays. Single pole tripping on these relays does not present any unusual challenges, as each phase of the protection element operates independently of the other unsalted phases.

1.3 Communications

Often transmission lines are protected by using schemes that require communications with relays located at other line terminals. The reliability of the communications obviously impacts the reliability of the protection system. GE Multilin relays include features that maintain reliable operation of the protection communications during power line faults, communications channel delays, communications channel switching, and communications channel dropout.

Pilot protection: Pilot protection schemes, such as directional comparison blocking and permissive over-reaching transfer trip, use simple on/off communications between relays. There are many methods to send this signal. The most common method is to use contact closure to an external communication circuit, such as power line carrier, microwave, radio, or fiber optic communications. GE Multilin relays simplify fiber optic communications method by using internal fiber optic communications via Direct I/O, eliminating the need for external communications devices. Direct I/O is a reliable mechanism that is simple to configure, securelytransmits digital status points such as tripping or blocking commands between relays via directly-connected or multiplexed fiber optic channels. Direct I/O operates within 2ms for

7

high speed communications to the remote line end. Direct I/O is available in any of the transmission line relays by adding an internal communications card. The output of the card can be IEEE C37.94, RS422 or G.703 communications to interface with fiber optic multiplexers, or may be a direct fiber connection to other relays. The communications card can be single-channel or dual-channel, to support point-to-point communications, dual point-to-point communications, or ring communications between up to 16 relays.

Line Current Differential: Communications is an integral piece of a line differential relay, as the currents from one line terminal must be sent to relays at other line terminals to perform the differential calculation. This requires the use of a digital communications channel, which is commonly a multiplexed channel where channel switching may occur. The analog information must be precisely time synchronized between the line ends for the differential calculation to be correct. Synchronization errors show up as phase angle offset, where identical currents produce phasors with different phase angles, and transient errors, where changes in current are seen at different times at different measurement points. For example, on a 60 Hz system, every 1ms of time shift between terminals introduces a 21.6° phase shift into the measured currents. There are two methods to account for the phase shift between line terminals due to the communications channel delay. One method is to measure the round-trip channel delay, and shift the local current phase by an angle equal to ½ of the round-trip delay time. This method is simple to implement, but creates a transient error when the communications channel is switched. In addition, the differential element will be temporarily blocked when the communications channel switches, or noise in the communications channel causes communications packet loss. The L90 Line Differential Relay employs a different method, using synchronous sampling by internally synchronizing the clocks on each L90. This method achieves high reliability, as the round-trip channel delay is not vitally important. The differential element successfully operates during channel switching or after packet loss, because the communications packets are preciselysynchronized. In the L90, synchronization is accomplished by synchronizing the clocks to each other rather than to a master clock. Each relay compares the phase of its clock to the phase of the other clocks and compares the frequency of its clock to the power system frequency and makes appropriate adjustments. The frequency and phase tracking algorithm keeps the measurements at all relays within a plus or minus 25 microsecond error during normal conditions for a 2 or 3 terminal system. In all cases, an estimate of phase error is computed and used to automatically adapt the restraint region of the differential element. The time synchronization algorithm can also use a GPS satellite clock to compensate for channel asymmetry. The use of a GPS clock is not normally required, except in applications such as a SONET ring where the communications channel delay may be asymmetric. This method produces synchronization accurate to within 125 microseconds between the relays on each end of the protected line. By using internally synchronized sampling, the L90 can accommodate 4 consecutive cycles of communications channel loss before needing to block the differential element. If the communications channel is restored within 5 seconds of channel loss, the L90 differential element will restart on the first received packet, without any time synchronization delay, due to the inertia of the internal clocks of the relays.

Line Phase Comparison: As with line differential, communications is an integral part of phase comparison relaying. Simple binary communications, such as power line carrier or

8

microwave, is used to send a pulse to the remote end when the phase angle of the measured current is positive. Coordination between the pulses from the remote end, and the phase angle measured at the local end, must be maThe L60 Line Phase Comparison Relay directly solves two common challenges with the carrier signal. The first issue is channel delay. The channel delay is measured during commissioning and is entered as a setting in the phase comparison element. The remote phase angle measurements are buffered and delayed by this value to match the incoming pulses from the remote relays. The L60 has two communications channels, and two independent channel time delays, to support three-terminal lines. The other common issue is pulse asymmetry of the carrier signal. Carrier sets may extend, either the mark (on) or space (off) signals at the receiving end compared with the originally sent signal. This difference is measured during commissioning by using oscillography data, and simply entered as a setting in the phase comparison element. In addition, the L60 supports some other methods to improve the reliability of protection communications. For short lines with negligible charging current, the channel delay measurement can be automated by running a loop-back test during normal system conditions and measuring the difference between the sent and received pulses. The L60 also supports automated check-back of the carrier system. Under normal conditions, the relay can initiate transmission of and modulate the analog signal to exchange small amounts of information. This automatic loop-back can replace the carrier guard signal, and more importantly, verifies the entire communications path, including the relays on both ends.

1.4 Security for Dual-Breaker TerminalsDual-breaker terminal line terminals, such as breaker-and-a-half and ring bus terminals, are a common design for transmission lines. The standard practice is to sum the currents from each circuit breaker externally by paralleling the CTs, and using this external sum as the line current for protection relays. This practiceintained.

9

works well during actual line faults. However, for some external fault events, poor CT performance may lead to improper operation of line protection relays. When current flows through a dual-breaker line terminal, the line current measured by a relay using external summation matches the actual line current only if the two CTs are accurate. The most significant relaying problem is CT saturation in either CT. The current measured by the relay may contain a large error current, which can result in the relay operating due to an incorrect magnitude ordirection decision. This incorrect operation may also occur if the linear error current of the CTs due to accuracy class is close to the through current level. These errors appear in the measured phase currents. As a result, relays that calculate the negative sequence and zero sequence currents from the measured phase currents may also see errors.

Distance: Distance relays applied at dual-breaker line terminals are vulnerable to mis-operation on external faults. During a closein reverse external fault, the voltage is depressed to a very lowlevel, and the security of the relay is maintained by directional supervision. If one of the line CTs saturates, the current measured by the relay may increase in magnitude, and be in the oppositedirection of the actual fault current, leading to an incorrect operation of the forward distance element for an external fault. The D90Plus Line Protection System and the D60 Line DistanceRelay handles the challenge of dual-breaker line terminals by supporting two three-phase current inputs to support breaker failure, over current protection, and metering for each circuit breaker. The relays then mathematically add these currents together to form the total line current used for distance and directional over current relaying. Directly measuring the currents from both circuit breakers allows the use of supervisory logic to prevent the distance element and directional overcorrect elements from operating incorrectly for reverse faults due to CT error. This supervisory logic does

not impact the speed or sensitivity of the protection elements, operates during all load conditions, and correctly allows tripping during an evolving external-to-internal fault condition. The dual-breaker line terminal supervisory logic essentially determines if the

10

current flow through each breaker is either forward or reverse. Both currents should be forward for an internal fault, and one current should be forward and one reverse for an external line fault. The supervisory logic uses, on a per-phase basis, a high-set fault detector (FDH), typically set at 2-3 times the nominal rating of the CT, and a directional element for each CT input to declare a forward fault, for each breaker. The logic also uses, on a per-phase basis, a low-set fault detector (FDL), typically set at 1.5-2 times the nominal rating of the CT, and a directional element to declare a reverse fault, for each breaker. Tripping is permitted during all forward faults, even with weak indeed at the dual-breaker terminal. Tripping is blocked for all reverse faults when one breaker sees forward current and one breaker sees reverse current. During an evolving external-to internal fault, tripping is initially blocked, but when the second fault appears in the forward direction, the block is lifted to permit tripping.

Line Differential: Line differential protection is prone to tripping due to poor CT performance on dual-breaker terminals, as the error current from the CTs is directly translated into a differential current. The only possible solution for traditional line differential relays is to decrease the sensitivity of the differential element, which limits the ability of the differential element to detect low magnitude faults, such as highly resistive faults. The L90 Line Differential Relay supports up to four three-phase current inputs for breaker failure, overcurrent protection, and metering for each circuit breaker. The relay then uses these individual currents to form the differential and restraint currents for the differential protection element.

The L90 differential element design explicitly accounts for the performance of the CTs for dual-breaker line terminals. Each L90 protecting a transmission line calculates differential

11

and restraint quantities based on local information directly measured by the relay, and information received from relays located at the remote line ends. Tripping decisions are made locally be each relay. The information sent by one L90 to the other L90s on the line is the local differential and restraint currents. The local differential current is the sum of all the local currents on a per-phase basis. One L90 can accept up to 4 current measurements, but only 2 currents are used for a dual-breaker application. The local restraint current is defined by the following equation for each phase.

LOC _ RESTRAINT LOC _ REST _ TRAD LOC ADA I = I + MULT • IThe starting point for the restraint is the locally measured current with the largest magnitude. This ensures the restraint is based on one of the measured currents for all fault events, andincreases the level of restraint as the fault magnitude increases. ILOC_REST_TRAD is this maximum current magnitude applied against the actual differential characteristic settings. ILOC_ADA is the sum of the squares estimate of the measurement error in the current, and is used to increase the restraint as the uncertainty of actual measurement increases, such as during high magnitude fault events and CT saturation. MULT is an additional factor that increases the error adjustment of the restraint current based on the severity of the fault event and the likelihood the fault is an external fault, when CT saturation is most likely to cause an incorrect operation. The values of ILOC and ILOC_RESTRAINT are transmitted to the L90 relays located at the other line ends. The differential and restraint values used in the actual tripping decision combine both the local differential and restraint current, and the differential and restraint currents from the remote line ends. These calculations are performed individually on each phase.REST LOC RESTRAINT REM RESTRAINT REM RESTRAINT I = I + I + IConsidering the worst case external fault with CT saturation, the differential current IDIFF will increase due to the CT error that appears in ILOC. However, the restraint current IREST will increase more significantly, as the ILOC_RESTRAINT uses the maximum of the local currents, that is increased based on the estimation of CT errors and presence of CT saturation. The end result is a correct restraining of the differential element.

Phase Comparison: The L60 Line Phase Comparison Relay supports two three-phase current inputs for breaker failure, overcurrent protection, and metering for each circuit breaker. Therelay then uses these individual currents to form the local phase angle information for use in the phase comparison scheme. A phase comparison relay operates by comparing the relative phase angles of the current from each end of the transmission line. When the measured current exceeds the level of a fault detector, and the phase angles from each end of the line are in phase, thephase comparison relay operates. For a dual-breaker application using an external sum, the saturation of one CT may cause the relay current to increase high enough to operate the fault detector. Because the current from the unsaturated CT predominates in this waveform, the phase angle of the relay current may change. If the phase angle of the relay current is in phase with the relay current at the remote end of the line, the relay will trip.

12

The L60 in dual-breaker applications selects the appropriate phase angle, based on the information measured from the current flow through both circuit breakers. The relay uses fault detectors on each current input, and develops the phase angle for each current input, and then special dual breaker logic consolidates the fault detector flags and the phase angle pulses for the line terminal. The fault detector flag is set for a line terminal if either fault detector from the two breakers is picked up. The type of phase comparison protection scheme, tripping or blocking, controls the pulse combination logic. For a tripping scheme, a positive polarity is declared for the terminal if one breaker displays positive polarity with its respective fault detector picked up, while the other breaker either does not show negative polarity or its fault detector is not picked up.

1.5 Redundancy Considerations to Enhance

ReliabilityThe reliability of transmission system protection is dependent on the reliability of the protection scheme used and the individual components of the protection scheme. Transmission protectionsystems typically use redundancy to increase the dependability of the system. There are two general methods of implementing redundancy. One method is to use multiple sets of protectionusing the same protection scheme. The other method is to use multiple sets of protection using different protection principles. Depending on the voltage class, either method of redundancymay involve using 2 or 3 sets of protection. In both cases, the goal is to increase dependability, by ensuring the protection operates for a fault event. Security may be

13

improved through the use of socalled voting schemes (e.g. 2-out-of-3), potentially at the expense of dependability. Multiple sets of protection using the same protection scheme involves using multiple relays and communications channels. This is a method to overcome individual element failure. The simplest method is to use two protection relays of the same type, using the same scheme and communications channel. This only protects against the failure of one relay. In some instances, relays of different manufacturers are used, to protect against common mode failures. It is also common to use redundant communications channels, in case of failure on one communications channel. Often, the communications channels use different methods, such as power line carrier and fiber optic. This is especially true due to the concerns of power line carrier operation during internal fault events. An alternative way to increase reliability through redundancy is to use multiple protection methods on the same line such as phase comparison and permissive over-reaching transfer trip, using different communications channels. This method protects againstindividual element failure of both relays and communications channels. More importantly, it protects against the failure of one of the protection methods. For example, a VT circuit fuse failure blocks a distance relay from operating, while a line differential system or phase comparison system will continue to operate. For this reason, often at least one current-only scheme, such as phase comparison or line differential, and then one pilot protection scheme based on distance relays are employed. A second advantage of using multiple protection methods to protect one line is the ability to increase the security of the line. It is possible to implement a “voting” scheme, where at least 2 protection methods must operate before the line can be actually tripped. Such a voting scheme may be applied permanently on lines where security is an issue, such as major inter-tie lines. A voting scheme may also be applied only when the system is at risk, such as during wide-area disturbances, either automatically based on system conditions, or by command from system operators. GE Multiline simplifies solutions when multiple protection schemes are used by providing both protective relays that only use current and protective relays that use both current and voltage. The L60 Line Phase Comparison Relay and the L90 Line Differential Relay are both current-only protection relays with different operatingprinciples. The D90Plus, D60 and D30 Line distance protection systems are full-featured distance relays. These relays are on a common hardware and software platform, simplifying engineering, design, installation, and operations issues. All of these relays support multiple communications options, including power line carrier, microwave, and fiber optic communications. The relays are also designed to communicate with each other, to implementvoting schemes, reclosing control, and other applications.

14

TYPICAL APPLICATIONS

1.1 General

This section highlights some typical application of GE Multiline line protection relays. This section is not intended as a comprehensive list of possible applications. For questions about the correct relay for a specific application, visit www.GEMultilin.com to review thebrochure for a specific relay model, or contact GE Multiline

15

16

17

CHAPTER - 2

DEVELOPMENTS OF TECHNIQUES USED IN TRANSMISSION LINE PROTECTION

1.1Artificial Neural Network Approach

An artificial neural network, often just named a neural network, is a mathematical

model inspired by biological neural networks. A neural network consists of an

interconnected group ofartificial neurons, and it processes information using

a connectionist approach to computation. In most cases a neural network is an adaptive

system changing its structure during a learning phase. Neural networks are used for

modeling complex relationships between inputs and outputs or to find patterns in data.

An artificial neural network is an interconnected group of nodes, akin to the vast network

of neurons in a brain.

18

1.2 Background

The inspiration for neural networks came from examination of central nervous systems. In

an artificial neural network, simple artificial nodes, called "neurons", "neurodes",

"processing elements" or "units", are connected together to form a network which mimics a

biological neural network.

There is no single formal definition of what an artificial neural network is. Generally, it

involves a network of simple processing elements exhibiting complex global behavior

determined by the connections between the processing elements and element parameters.

Artificial neural networks are used with algorithms designed to alter the strength of the

connections in the network to produce a desired signal flow.

Neural networks are also similar to biological neural networks in performing functions

collectively and in parallel by the units, rather than there being a clear delineation of

subtasks to which various units are assigned. The term "neural network" usually refers to

models employed in statistics, cognitive psychology and artificial intelligence. Neural

network models which emulate the central nervous system are part of theoretical

neuroscience and computational neuroscience.

In modern software implementations of artificial neural networks, the approach inspired by

biology has been largely abandoned for a more practical approach based on statistics and

signal processing. In some of these systems, neural networks or parts of neural networks

(like artificial neurons) form components in larger systems that combine both adaptive and

non-adaptive elements. While the more general approach of such adaptive systems is more

suitable for real-world problem solving, it has far less to do with the traditional artificial

intelligence connectionist models. What they do have in common, however, is the principle

of non-linear, distributed, parallel and local processing and adaptation. Historically, the use

of neural networks models marked a paradigm shift in the late eighties from high-level

(symbolic) artificial intelligence, characterized by expert systems with knowledge embodied

in if-then rules, to low-level (sub-symbolic) machine learning, characterized by knowledge

embodied in the parameters of a dynamical system.

1.3 Models

Neural network models in artificial intelligence are usually referred to as artificial neural

networks (ANNs); these are essentially simple mathematical models defining a

function   or a distribution over   or both   and  , but sometimes models are also

intimately associated with a particular learning algorithm or learning rule. A common use of 19

the phrase ANN model really means the definition of a class of such functions (where

members of the class are obtained by varying parameters, connection weights, or specifics of

the architecture such as the number of neurons or their connectivity).

Network function

The word network in the term 'artificial neural network' refers to the inter–connections

between the neurons in the different layers of each system. An example system has three

layers. The first layer has input neurons, which send data via synapses to the second layer of

neurons, and then via more synapses to the third layer of output neurons. More complex

systems will have more layers of neurons with some having increased layers of input

neurons and output neurons. The synapses store parameters called "weights" that manipulate

the data in the calculations.

An ANN is typically defined by three types of parameters:

1. The interconnection pattern between different layers of neurons

2. The learning process for updating the weights of the interconnections

3. The activation function that converts a neuron's weighted input to its output

activation.

Mathematically, a neuron's network function   is defined as a composition of other

functions  , which can further be defined as a composition of other functions. This can

be conveniently represented as a network structure, with arrows depicting the dependencies

between variables. A widely used type of composition is the nonlinear weighted sum,

where  , where   (commonly referred to as the activation function [1] ) is

some predefined function, such as the hyperbolic tangent. It will be convenient for the

following to refer to a collection of functions  as simply a vector  .

ANN dependency graph

This figure depicts such a decomposition of  , with dependencies between variables

indicated by arrows. These can be interpreted in two ways.

20

The first view is the functional view: the input   is transformed into a 3-dimensional

vector  , which is then transformed into a 2-dimensional vector  , which is finally

transformed into  . This view is most commonly encountered in the context of optimization.

The second view is the probabilistic view: the random variable   depends upon the

random variable  , which depends upon  , which depends upon the random

variable  . This view is most commonly encountered in the context of graphical models.

The two views are largely equivalent. In either case, for this particular network architecture,

the components of individual layers are independent of each other (e.g., the components of   

are independent of each other given their input  ). This naturally enables a degree of

parallelism in the implementation.

Two separate depictions of the recurrent ANN dependency graph

Networks such as the previous one are commonly called feedforward, because their graph is

a directed acyclic graph. Networks with cycles are commonly calledrecurrent. Such

networks are commonly depicted in the manner shown at the top of the figure, where   is

shown as being dependent upon itself. However, an implied temporal dependence is not

shown.

Learning

What has attracted the most interest in neural networks is the possibility of learning. Given a

specific task to solve, and a class of functions  , learning means using a set

of observations to find   which solves the task in some optimal sense.

This entails defining a cost function   such that, for the optimal solution 

,     - i.e., no solution has a cost less than the cost of the optimal solution

(see Mathematical optimization).

The cost function   is an important concept in learning, as it is a measure of how far away a

particular solution is from an optimal solution to the problem to be solved. Learning

21

algorithms search through the solution space to find a function that has the smallest possible

cost.

For applications where the solution is dependent on some data, the cost must necessarily be

a function of the observations, otherwise we would not be modelling anything related to the

data. It is frequently defined as a statistic to which only approximations can be made. As a

simple example, consider the problem of finding the model  , which

minimizes  , for data pairs   drawn from some distribution  . In practical

situations we would only have   samples from   and thus, for the above example, we

would only minimize  . Thus, the cost is minimized over a sample of

the data rather than the entire data set.

When   some form of online machine learning must be used, where the cost is partially

minimized as each new example is seen. While online machine learning is often used

when   is fixed, it is most useful in the case where the distribution changes slowly over

time. In neural network methods, some form of online machine learning is frequently used

for finite datasets.

Choosing a cost function

While it is possible to define some arbitrary, ad hoc cost function, frequently a particular

cost will be used, either because it has desirable properties (such as convexity) or because it

arises naturally from a particular formulation of the problem (e.g., in a probabilistic

formulation the posterior probability of the model can be used as an inverse cost).

Ultimately, the cost function will depend on the desired task. An overview of the three main

categories of learning tasks is provided below.

Learning paradigms

There are three major learning paradigms, each corresponding to a particular abstract

learning task. These are supervised learning, unsupervised learning and reinforcement

learning.

Supervised learning

In supervised learning, we are given a set of example pairs   and the aim is to

find a function   in the allowed class of functions that matches the examples. In other

words, we wish to infer the mapping implied by the data; the cost function is related to the

mismatch between our mapping and the data and it implicitly contains prior knowledge

about the problem domain.

22

A commonly used cost is the mean-squared error, which tries to minimize the average

squared error between the network's output, f(x), and the target value y over all the example

pairs. When one tries to minimize this cost using gradient descent for the class of neural

networks called multilayer perceptrons, one obtains the common and well-

known backpropagation algorithm for training neural networks.

Tasks that fall within the paradigm of supervised learning are pattern recognition (also

known as classification) and regression (also known as function approximation). The

supervised learning paradigm is also applicable to sequential data (e.g., for speech and

gesture recognition). This can be thought of as learning with a "teacher," in the form of a

function that provides continuous feedback on the quality of solutions obtained thus far.

Unsupervised learning

In unsupervised learning, some data   is given and the cost function to be minimized, that

can be any function of the data   and the network's output,  .

The cost function is dependent on the task (what we are trying to model) and our a

priori assumptions (the implicit properties of our model, its parameters and the observed

variables).

As a trivial example, consider the model  , where   is a constant and the

cost  . Minimizing this cost will give us a value of   that is equal to the mean

of the data. The cost function can be much more complicated. Its form depends on the

application: for example, in compression it could be related to the mutual

information between   and  , whereas in statistical modeling, it could be related to

the posterior probability of the model given the data. (Note that in both of those examples

those quantities would be maximized rather than minimized).

Tasks that fall within the paradigm of unsupervised learning are in

general estimation problems; the applications include clustering, the estimation of statistical

distributions, compression andfiltering.

Reinforcement learning

In reinforcement learning, data   are usually not given, but generated by an agent's

interactions with the environment. At each point in time  , the agent performs an action   

and the environment generates an observation   and an instantaneous cost  , according to

some (usually unknown) dynamics. The aim is to discover a policy for selecting actions that

minimizes some measure of a long-term cost; i.e., the expected cumulative cost. The

23

environment's dynamics and the long-term cost for each policy are usually unknown, but can

be estimated.

More formally, the environment is modeled as a Markov decision process (MDP) with

states   and actions   with the following probability distributions: the

instantaneous cost distribution  , the observation distribution   and the

transition  , while a policy is defined as conditional distribution over actions

given the observations. Taken together, the two define a Markov chain (MC). The aim is to

discover the policy that minimizes the cost; i.e., the MC for which the cost is minimal.

ANNs are frequently used in reinforcement learning as part of the overall

algorithm. Dynamic programming has been coupled with ANNs (Neuro dynamic

programming) by Bertsekas and Tsitsiklis[2] and applied to multi-dimensional nonlinear

problems such as those involved in vehicle routing or natural resources management

because of the ability of ANNs to mitigate losses of accuracy even when reducing the

discretization grid density for numerically approximating the solution of the original control

problems.

Tasks that fall within the paradigm of reinforcement learning are control

problems, games and other sequential decision making tasks.

Learning algorithms

Training a neural network model essentially means selecting one model from the set of

allowed models (or, in a Bayesian framework, determining a distribution over the set of

allowed models) that minimizes the cost criterion. There are numerous algorithms available

for training neural network models; most of them can be viewed as a straightforward

application of optimization theory andstatistical estimation.

Most of the algorithms used in training artificial neural networks employ some form

of gradient descent. This is done by simply taking the derivative of the cost function with

respect to the network parameters and then changing those parameters in a gradient-

related direction.

Evolutionary methods,[3] gene expression programming,[4] simulated annealing,[5] expectation-maximization, non-parametric methods and particle swarm optimization [6]  are

some commonly used methods for training neural networks.

24

CHAPTER -3

IMPLEMENTATION OF SIMULINK IN PROTECTION OFTRANSMISSION LINES

1.1 General

Work on artificial neural networks commonly referred to as “neural networks”, has been motivated right from its inception by the recognition that the human brain computers in an entirely different way from the conventional digital computers. The brain is a highly complex, nonlinear and parallel computer. It has the capability to organize its structural constituents known as neurons, so as to perform certain computations (e.g. pattern recognition, perception and motor control) many times faster than the fastest digital computer in existence today.[1]A neural network is a massively parallel distributed processor made up of simple processing unit that has a natural propensity for storing experimental knowledge and making it available for use. It resembles the brain in two respects:Knowledge is acquired by the network from its environment through the learning process.Inter-neuron connection strengths known as synaptic weights are used to store the acquired knowledge

1.2 Artificial intelligence

Designing of intelligence computer system from characteristic associated with intelligence in human behavior Example: - 1. Neural Network 2. Fuzzy Logic 3. Expert system 4. Probabilistic reasoning. Types: 1. Hard computing 2. Soft computing Characteristics: - 1. Cognition 2. Logical Interface 3. Pattern RecognitionHuman brain has two properties: - Human brain is getting experienced to adapt themselves to their surrounding environments. So as a result the information processing capability of the brain is rendered, when this happen the brain becomes plastic.1. Plastic: - Capability to process information, capability of adding. Must preserve the information it has learn previously.2. Stable: - Remain stable when it is presented with irrelevant information, useless information.*. Synapses with large area are excitatory (+Ve weights) & with small area are inhibitory (-Ve weights).*. Synapses of the neuron are modulated as weights. (Strength of the connection)*. Biological neuron receive all inputs through dendrites, sum them & produces an output. If the sum is greater then the threshold value, then input signals are passed to the cell body.*. NN can mapped input patterns into output patterns.

25

*. NN’s are robust systems. They can recall full patterns from incomplete pattern or noise channel.*. Mc Culloh Pits neuron was centered on the idea that a neuron will fire an impulse only if the threshold value is exceed.Total input (I) receive by the soma

To generate the final output Y sum is passed through a nonlinear filter called activation filter.

1.3 Semolina Modeling Diagram

Modern power systems are increasingly complex and the use of computers to analyze and simulate their Performance is now commonplace. There are many computer packages on the market. Unfortunately, it is not all about entering data, but how the input data was arrived at. It is also important to distinguish between good and bad Data. Semolina figure; is made of generators, motors, transmission lines, transformers, high speed digital relays and circuit breakers. For each of these items data is required before meaningful results can be obtained, SIMULINK is mouse-driven. The modeling of any system is carried out in the model window through identification and Connection of the blocks that make up the system under consideration. Usually, it is important to model the system first on paper as to have a general knowledge of the various components that make up the system.

A three-phase, 60 Hz, 735 kV power system transmitting power from a power plant consisting of six 350 MVA generators to an equivalent network through a 600 km

26

transmission line. The transmission line is split in two 300 km lines connected between buses B1,B2, and B3. In order to increase the transmission capacity, each line is series compensated by capacitors representing 40% of the line reactance. Both lines are also shunt compensated by a 330 Mvar shunt reactance. The shunt and series compensation equipments are located at the B2 substation where a 300 MVA 735/230 kV transformer with a 25 kV tertiary winding feeds a 230 kV, 250 MW load. The series compensation subsystems are identical for the two lines. For each line, each phase of the series compensation module contains the series capacitor , a metal oxide arrestor (MOV) protecting the capacitor, and a parallel gap protecting the MOV. When the energy dissipated in the MOV exceeds a threshold level of 30 MJ, the gap simulated by a circuit breaker is fired. CB1 and CB2 are the two line circuit breakers .The generators are simulated with a Simplified Synchronous Machine block. Universal transformer blocks (two windings and three-windings) are used to model the two transformers. Saturation is implemented on the transformer connected at bus B2. Voltages and currents are measured in B1, B2, and B3 blocks. These blocks are Three-phase V-I Measurement blocks where voltage and current signals are sent to the Data Acquisition block through Got blocks. Fault and Line Switching You now study the transient performance of this circuit when a line-to-ground and three-phase-to-ground faults are applied on line 1. The fault and the two line circuit breakers CB1 and CB2 are simulated with blocks from the threephase library. Open the dialog boxes of CB1 and CB2. See how the initial breaker status and switching times are specified. A line-to-ground fault is applied on phase A at t = 1cycle. The two circuit breakers which are initially closed are then open at t = 5 cycles, simulating a fault detection and opening time of 4 cycles. The fault is eliminated at t = 6 cycles, one cycle after line opening.

1.4 Demonstration

Notice that this system contains the Powergui block. In addition, when you start the system the 'power_3phseriescomp' model, the sampling time Ts = 50e-6 is automatically set in your workspace. The system is therefore be discretized using a 50 microseconds sample time.Line-to-Ground FaultDouble click the Data Acquisition block and open the three scopes. Start the simulation. As the system has already been initialized (1500 MW generation at the 13.8 kV bus) with the Lod Flow utility of the Powergui, the simulation starts in steady state. At t = 1 cycle a line-to-ground fault is applied and the fault current reaches 10 kA . During the fault, the MOV conducts at every half cycle and the energy dissipated in the MOV builds up to 13 MJ. At t = 5 cycles the line protection relays (not simulated) open breakers CB1 and CB2 and the energy stays constant at 13 MJ. As the maximum energy does not exceed the 30 MJ threshold level, the gap is not fired. After breaker opening the fault current drops to a small value and the line and series capacitance start to discharge through the fault and the shunt reactance. The fault current extinguishes at the first zero crossing after the opening order given to the fault breaker (t = 6 cycles). Then, the series capacitor stops discharging and its voltage oscillates around 220 kV . Three-Phase-to-Ground Fault Change the fault type to a three-phase-to-ground fault by checking Phases A, B, and C in the Fault Breaker block. Restart the simulation. Notice that during the fault the energy dissipated in the MOV builds up faster that in the case of a line-to-ground fault. The energy reaches the 30 MJ threshold

27

level after 3 cycles, one cycle before opening of the line breakers. As a result, the gap is fired and the capacitor voltage quickly discharges to zero through the damping circuit.

1.5 SIMULINK TEST RESULTS

(i) Results of VabcB1 & IabcB1 with respect to phase A

(ii) Results of VabcB2, IabcB2 & Iamb Fault with respect to phase A

28

(iii) Results of Va, Ia & Energy with respect to phase A

(iv) Results of VabcB1 & IabcB1 with respect to phase B

29

(v) Results of VabcB2, IabcB2 & IabcFault with respect to phase B

(vi) Results of Va, Ia & Energy with respect to phase B

30

(vii) Results of VabcB1 & IabcB1 with respect to phase C

(viii) Results of VabcB2, IabcB2 & IabcFault with respect to phase C

31

(ix) Results of Va, Ia & Energy with respect to phase C

32

CHAPTER 5

CONCLUSIONS

A survey of transmission line protection is done through this article. Since the implementation of digital relaying, a lot of work has been done to improve the performance of digital protective relays, but in the context of reformation in the power industry and operation of transmission lines close to the stability limits, new tools and algorithms are needed to maintain system reliability and security within an acceptable level. The ANN, fuzzy logic, genetic algorithm, SVM and wavelet based techniques have been quite successful but are not adequate for the present time varying network configurations, power system operating conditions and events. Therefore, it seems that there is a significant scope of research in AI techniques which can simplify the complex nonlinear systems, realize the cost effective hardware with proper modification in the learning methodology and preprocessing of input data and which are computationally much simpler. Also development of reliable software and communication system will pave the way for better relaying and fault location performance using multi terminal synchronized phasor measurement based on GPS. This article is an effort to present the most comprehensive set of references on the subject of recent techniques in transmission line protection.

33

REFERENCE

Vasilic S. and Kezunovic M., 2005. Fuzzy ART neural network algorithm for classifying the power system faults, IEEE Trans.Power Delivery, vol.20, no.2, pp.1306-1314.Venkatesan R. and Balamurugan B., 2001. A real-time hardware fault detector using an artificial neural network for distanceprotection, IEEE Trans. Power Delivery, vol.16, no.1, pp.75-82.Valsan, Swarup K.S., 2009. Wavelet transform based digital protection for transmission lines, Electrical Power and EnergySystems, vol.31, no.7-8, pp. 379-388.Wang F. and Tang J., 1997. Modeling of a transmission line protection relaying scheme using petri nets, IEEE Trans. Power Del.,vol.12, no.3, pp. 1055-1063.Wang H. and Keerthipala W.W.L., 1998. Fuzzy- Neuro approach to fault classification for transmission line protection, IEEETrans. Power Delivery, vol.13, no.4, pp. 1093-1104.Xiangning Lin, Hanli weng and Bin Wang, 2009. A generalized method to improve the location accuracy of the single-endedsampled data and lumped parameter model based fault locators, Electrical power and energy systems, vol.31, no.5, pp 201-205.Xu Z.Y., Huang S.F., Li Ran. Liu J.F., Qin Y.L., Yang Q.X. and He J.L., 2008. A distance protection relay for a 1000-kV UHVtransmission line, IEEE Trans. Power Del., vol.23, no.4, pp. 1795-1804.Xu Z.Y., Jiao S.H., Ran L., and Du Z.Q., 2008. An online fault-locating scheme for EHV/UHV transmission lines, IET Gener.Transm. Distrib., vol.2, no.6, pp.789-799.Yeo S.M., Kim C.H., Hong K.S., Lim Y.B., Aggarwal R.K., Johns A.T., and Choi M.S., 2003. A novel algorithm for faultclassification in transmission lines using a combined adaptive network and fuzzy inference system, Electrical Power and EnergySystems, vol.25, no.9, pp.747-758.Youssef O.A.S., 2004. Combined fuzzy-logic wavelet-based fault classification technique for power system relaying, IEEE Trans.Power Delivery, vol.19, No.2, pp.582-589.Zadeh L.A., 1965. Fuzzy sets, Inform. Control, vol.8, pp. 338-353.Zahra F., Jayasurya B. and Quaicoe J.E., 2000. High-speed transmission relaying using artificial neural networks, Electric Power

34

Systems Research, vol. 53, no.3, pp. 173-179.Zamora I., Minambres J.F., Mazon A.J., Alvarez-Isasi R., and Lazaro J., 1996. Fault location on two-terminal transmission linesbased on voltages, IEE Proc. – Gener. Transm. Distrib., vol.143, No.1, pp.1-6.Zhang N. and Kezunovic M., 2007. Transmission line boundary protection using wavelet transform and neural network, IEEETrans. Power Delivery, vol.22, no.2, pp. 859-869.

35