difference between inrush and fault current.pdf

Iranica Journal of Energy & Environment 2 (4): 302-312, 2011ISSN 2079-2115 IJEE an Official Peer Reviewed Journal of Babol Noshirvani University of TechnologyDOI: 10.5829/idosi.ijee.2011.02.04.3139

BUT

Corresponding Authro: M. Rasoulpoor, Faculty of Electrical and Engineering Systems,Shahrood University of Technology, Shahrood, Iran.

302

Discrimination Between Inrush and Short CircuitCurrents in Differential Protection of Power Transformer Based

on Correlation Method Using the Wavelet Transform

M. Rasoulpoor, M. Banejad and A. Ahmadyfard

Faculty of Electrical and Engineering Systems,Shahrood University of Technology, Shahrood, Iran

(Received: October 30, 2011; Accepted: December 8, 2011)Abstract: This paper presents a novel technique for transformer differential protection to prevent incorrectoperation due to inrush current. The proposed method in this paper is based on time-frequency transformknown as the Wavelet transform. The discrete Wavelet transform is used for analysis the differential currentsignals in time and frequency domains. The investigation on the energy distribution of the signal on thediscrete Wavelet transform components shows the difference distribution between inrush and internal faultcurrent signals of power transformer. The correlation factor which is a statistical parameter is used here toexpress the pattern of the energy distribution for different current signals. The proposed algorithm is based onthe correlation factors to distinguish between internal fault and inrush currents in the transformer differentialprotection. The proposed algorithm is tested and simulated for several cases by simulating inrush and internalfault currents. The simulation of inrush and internal fault currents are performed using electromagnetic transientprogram PSCAD/EMTDC software. Simulation results show that the proposed scheme accurately identifiesinrush and fault currents at the distance of the power transformer protection in a time period less than quarterof power frequency cycle. In addition the proposed method has high sensitivity and reliability. The method haslow computation work and not requires determining the threshold for each new power system.

Key words:Inrush current % Differential protection % Power transformer % Discrete wavelet transform (DWT)% Internal fault

INTRODUCTION A common way to detect the inrush current is use of

Differential relays are the initial and principal an inrush current is larger than an internal fault current [1].protective tools for power transformers protection. The ratio of the second harmonic component to originalThe essence of operation of differential protection is component of the current signal is compared with abased on comparing currents flowing into and out of the threshold value. This ratio is a measure to distinguish thetransformer protected area. There are some cases such as inrush current from fault current. However, by occurringinrush currents, saturation of current transformer and power system changes and improving in transformer coreoperation of the transformer tap changer that may lead to material and, the level of second harmonic component ofincorrect operation of the differential relay. Any sudden the inrush current has been reduced [2]. On the otherchange in transformer terminals voltage results in a hand, in some severe faults, the level of second harmonictransient current into the power transformer that is called component in fault currents is higher than inrush currents.inrush current. One of the important cases of inrush Drawbacks of the method in [2] indicate that there is acurrents is the case that transformer is switching to the need for expressing new tools for detecting inrushpower system. Inrush current results in needless action currents. In one approach namely, differential powerfor differential relays. method, sum of active power flowing into transformers

the second harmonic component of current that usually in

,1

( ) ( )a bt b

taa

ψ ψ−

=

Iranica J. Energy & Environ., 2 (4): 302-312, 2011

303

from each terminal is used [3]. The proposed method in [4] This paper is organized as follows: the proposedis based on the modal transform of voltage and currentwaveforms. This method needs measuring of the voltageas well as current and, also computations and recognitiontime are lengthy. In [5], the dead angle method is usedwhich is based on the time at which the current waveformis zero, but it has also long time delay in its detectionalgorithm. In recent years, some methods such as fuzzylogic and neural network have been proposed foridentifying inrush current [6-8]. The fuzzy logic methodshave not good performances when there are rapidchanges in the power system. Also, these methods use alot of rules to make a decision and to create these rules,more work should be done. On the other hand, the neuralnetwork based techniques require large training patternswhich need a lot of computation time. Also, a separatedesign should be performed for each new network in thepower system.

Frequency analysis can be an effective technique toanalysis and get features of signals with complicatedcharacteristics. Such effective analysis is achieved byemploying new and efficient signal processing tools.The traditional signal processing tools such as Fouriertransform are based on the condition of stationarity andalternating of the signals. However, the disturbances inpower systems are non periodic, non-stationary, withshort duration and abrupt natures [9]. Recently, some newsignal processing techniques such as Wavelet analysishave been proposed to solve this. The Wavelet techniquecan be applied successfully to various signal and imageprocessing methods, especially for the signals withtransient natures and variation with time such as somepower system disturbances. Useful information in signalusually can be extracted completely when a simultaneousresolution is performed both in time and frequencyregions using the Wavelet transform. Therefore, thistransform have been used in many study cases fordiscriminating between fault current and inrush current inpower transformer differential protection [10-12].

This paper proposes a new algorithm foridentification inrush currents from fault currents.The Wavelet transform is used to decompose the signalinto approximation and detail coefficients for thisapproach. Then, the detection is performed based ona criterion that is formed using these coefficients.The recognition criterion employs a correlation factor.This factor is used to determine the relationshipbetween wavelet coefficients energy at the differentlevels. The proposed algorithm is evaluated by severalsimulated inrush currents and internal fault currents onthe simulated power system.

method is presented in the Section II. In this section briefdescription about the Wavelet transform and correlationfactor is given and then, the use of the Wavelet transformand correlation factor is explained in the proposedmethod. The proposed algorithm in the differentialprotection is presented in Section III. Simulatingthe proposed method on a test system is presented inSection IV.

Proposed Approach Based on Wavelet Transform andCorrelation FactorDiscrete Wavelet Transform: The Wavelet Transformprovides an efficient tool for signal processing in timeand frequency domains. Investigation on the signalenergy components within the lower frequency levelsreveals that it is possible to use the information of thisenergy for discriminating inrush from short circuitcurrent. Also the sub-band filtering of the signalleads naturally to multi-resolution signal decomposition[13]. As the input signals in our study are nonstationary signals that their statistical propertieschange with time. For such signals, ordinary Fourieranalysis is not adequate for extracting proper features. Forthis purpose, it is necessary to express the signal both intime and frequency domains. The short-time Fouriertransform (STFT) and the Wavelet transform have arepotent for analysis the signals. The STFT has a fixed-duration window that makes a fixed frequency resolutionand thus allows only a fixed time-frequency resolution. Onthe other hand, the wavelet transform is obtained bydilation (or contraction) and translation of the primaryband pass function. The Wavelets are used to transformthe under investigation signals into anotherrepresentation which shows the information of the signalsin a more useful form. It is worthwhile to note that thewavelet transform is a convolution of the wavelet functionwith the original signal. In this paper, the wavelettransform is implemented in its digital format, DiscreteWavelet transform (DWT). The complexity of DWT isvery low in compared to the continuous Wavelettransform (CWT). Also, performing coefficients analysisof DWT is easier than many coefficients of CWT. Patternrecognition of the signal manner is more accurate usingDWT. The continuous wavelet function is defined asfollows [14]:

(1)

0 0,

00

1( ) ( )

m

m n mm

t b at

aaψ ψ

−=

2 1 2 1

, , , ,0 1 0

( ) ( ) ( )

M m M mM

m n m n m n m nn m n

f t S t T tφ ψ− −− −

= = =

= +∑ ∑ ∑


304

The dilation and contraction of the wavelet is Fig. 1 shows the decomposing the signal into detailestablished by the dilation parameter "a". The movement and approximation coefficients. In this figure, LP and HPof the wavelet along the time axis is established by the are high pass and low pass filters, respectively.translation parameter "b". The wavelet described by Eq. According to this figure, in discrete wavelet analysis, the1 is known as the mother wavelet or analyzing wavelet. signal is first divided into low and high frequencyFor DWT, wavelet function is defined by sampling from components at the first level. This first low frequencythese parameters (a, b). This wavelet has the form as: sub-band, containing most of the signal energy, is

These approximation coefficients are convolved with(2) the low pass filter and then down sampled to give the

Where, the integers 'm' and 'n' control the wavelet dilation other hand, the approximation coefficients are convolvedand translation, respectively. The parameter of 'a ' is a with the high pass filter and down sampled to give the0

specified fixed dilation step parameter and 'b ' is the detail signal coefficients at the next level. The detail0

location parameter. Common choices for discrete wavelet components are kept and the approximation componentsparameters 'a ' and 'b ' are 2 and 1, respectively which is are again passed through the low pass and high pass0 0

known as the Dyadic grid arrangement. The Dyadic grid filters to give coefficients at the subsequent level andis the simplest and most efficient choice for practical process is be continued. Eventually, the signal can bepurposes [14]. Dyadic wavelet (mother wavelet) can be represented by a low-pass at a certain scale, plus sum ofwritten as (n is an integer number): detail signals at different resolutions.

R (t) = 2 R(2G t–n) (3) by each level is directly related to the sampling frequencym,nGm/2 m

The orthonormal dyadic discrete wavelets are the wavelet analysis. According to the Nyquist criterion,associated with scaling function and their dilation the highest frequency component found in a signal isequation. The scaling function is given by: the half of the sampling frequency [15]. In this study,

N (t) = 2 N(2G t–n) (4) corresponds to 400 samples at the power frequency cyclem,nGm/2 m

Let f(1) be a discrete input signal and it is a signal It is worthwhile to note that there are differentwith a finite length of N = 2 . Thus, the range of scales mother wavelet functions for different applications.M

can be investigate is 0<m<M. The input signal can be On the other hand, choosing the mother waveletexpressed in terms of its discrete wavelet equation as plays an important role in the success and efficiency offollows: the analysis and detecting signal events. The mother

under study signal as well as the shape of signal(5) transient being detected. This paper includes analyzing

Where, T and S are known as detail and approximation frequency current signals. For these signals, Daubechiesm,n m,n

coefficients, respectively. Detail and approximation mother wavelet is an appropriate choice as reported incoefficients are correspond with filtering plus down [16]. In the proposed method, Daubechies with twosampling of the original signal f(t) as shown in Fig. 1. coefficients ‘db2’ is introduced as the mother wavelet.

the approximation coefficients in the first level.

approximation signal coefficients at the next level. On the

It should be noted that the frequency band presented

and it is used to determine the frequency band for

the chosen sampling frequency is 20 kHz which is

of 50 Hz.

wavelet should be chosen based on the nature of the

short duration, fast decaying and oscillation type of high

Fig. 1: m-level signal decomposition

1 2,x xρ

1 21 2

,1 2

cov ( , )x x

ariance x xρ

σ σ=

2 var , 1,2.iance iiσ = =

1 2,x xρ

1 2

1 1 2 2

, 2 21 1 2 2

( )( )

( ( ) )( ( ) )

mn mn

mn mn

m nx x

m n m n

x x x x

x x x xρ

− −

=− −

∑∑∑∑ ∑∑

( ), 1,2i ix mean x i= =

2 12

,0

( )M i

i i nn

Ed d− −

=

= ∑

12

0

( ( ))k

j jn

E I n−

=

= ∑

, 100ii j

j

EdEd

E= ×

, 1, 2, 3, 4, 5,[ , , , , ]i j j j j j jEd Ed Ed Ed Ed Ed=

11 12 13 1,

21 22 23 2,

31 32 33 3,

41 42 43 4,

51 52 53 5, 5,

j

j

j

j

j j

E E E E

E E E E

Ed E E E E

E E E E

E E E E

=

……………

1, 2, 3, 4, 5,[ , , , , ]Tj j j j j jEd Ed Ed Ed Ed Ed=


305

The next step is to determine the number of Where, d are the components of detail coefficients,decomposition levels. Since the data window has 64 'M' is the all number of Wavelet decomposition levels andsamples, utmost of decomposition level can be six. For our 'i' is a certain level that the detail components energy ispurpose five levels is sufficient and signal features are computed at its level.specified for using in the discriminative algorithm by Also, the total energy of signal at each slidingthese numbers of levels. A sliding window of 64 samples window is computed as:is used in this paper that results in five level ofresolution with five details plus an approximation.

Base of Proposed Method: The proposed method is basedon a statistical criterion that is used to extract At the above equation 'k' is the amount of samples atspecification of differential currents components. each sliding window and 'j' is the number of slidingTo illustrate the discriminating criterion, the covariance is window. From Eqs. 8 and 9, the energy percentage offirstly described. The covariance function tries to capture detail coefficients to the total energy of signal isa sense of joint dependence between two real valued computed:random variables. A correlation coefficient; however, isan appropriately scaled version of a covariance. (10)The correlation coefficient between two random variablesx and x , denotes by is defined as follows [17]: Therefore, a vector of energy percentage of DWT1 2

(6)

Where,

Two random variables, x and x are called negatively1 2

correlated, uncorrelated, or positively correlated,respectively if and only if is negative, zero orpositive, respectively.

Assume that the two random variables are twomatrices or vectors with the same size. Thus, the resultedequation for the correlation factor between x , x is1 2

as follows:

(7)

Where, m,n are dimension of vectors and .From above explanation, this statistic coefficient and

Wavelet transform is used to produce discriminativecriterion to distinguish between inrush and fault currentsin power transformer. In order to create the discriminativecriterion, at first energies of detail components fromwavelet transform in each level are computed as:

(8)

i,n

(9)

coefficient for each of the five levels and in any slidingwindow is obtained from:

(11)

Where, Ed = Wavelet detail coefficients energyi

percentage to total signal energy at the level,(i=1, 2, 3, 4, 5) and j is the number of sliding window.Therefore, for the input signal with N discrete samples,the energy vectors matrix of all samples is constituted asbelow:

(12)

As there are 64 samples at each sliding window, thenumber of all windows at the signal with N samples isj=N-64. Therefore, the Ed matrix has the dimension of5×N-64. It means that the resolutions are performed for thesignal with N samples at the sliding windows with 64samples at five levels of wavelet transform decomposition.After that the correlation factor between two continuousvectors is calculated as below (j is the number of thesliding window):

(13)

1 1, 1 2, 1 3, 1 4, 1 5, 1[ , , , , ]Tj j j j j jEd Ed Ed Ed Ed Ed+ + + + + +=

, 1 1( , )j j j jcorrelation Ed Edρ + +=

1,2 2,3 3,4 1,[ ]j jCf ρ ρ ρ ρ −= …


306

(14) I = (I – I )–(I – I ) (19)

(15) Where, I , I , I are the primary side currents and I , I , I

The D is the correlation factor between the energy and I , I , I are the differential currents and input currentsj,j+1

vectors at the j and j+1 numbers of sliding windows. for the differential relay. In the proposed algorithm I , I ,Thus, the correlation factors for the matrix of the (12) that I are the input signals which are employed by theare named Cf are as below: proposed method. Differential currents are then compared

(16) algorithm. The value of the threshold current is

In this study, the values of Cf factors are used in the power transformer and performance characteristics ofdiscriminative method. It is expected that the values of Cf differential relay, which is set such that to avoidto be near the value of "1" (highly correlated signals) for responding to heavy external fault and steady statethe steady state condition. It is because of fact that the conditions. If any of the differential currents exceeds thisnormal current has smooth condition and the energy of threshold, the program starts to apply the DWT analysisthe signal has not a noticeable change and energy to the currents signals. As a result, the signals arepercentage between two sliding windows are highly decomposed into detail and approximation coefficients.correlated. On the other hand, with a sudden change in Then, the energy percentage of detail coefficients isthe current magnitude and power system disturbances, calculated for any of the five levels. After this step, athe energy of signal has a rapid change in the DWT vector is organized from the calculated energycomponents and with change in the condition of the percentages for each sliding window (Eqs. 11 and 12).power system the signal energy should be greater than In the next step, the correlation factor (D ) is calculatedthe normal condition. In the case of inrush current between two consecutive vectors that are organizedbecause of oscillatory nature of the current and having a from two consecutive sliding windows (Eqs. 15 and 16).different energy spectrum at different frequency levels, The values of the correlation factors versus the samplesthe energy percentage at the high frequency level will be number for each of the three differential currents inchanged approximately with an oscillatory shape. the proposed algorithm is computed in the next step.

Proposed Discriminative Algorithm: The algorithm of the 2' and 'CF 3' respectively, for I , I and I in the simulationproposed approach is illustrated in this section. In order results part of the study.to clarify the proposed method, the diagram of the test Simulation results verify that the correlation factorssystem is depicted in Fig. 2. This test system will be used for inrush current have oscillation nature and for faultlater in the simulation section. current these factors have smooth nature after transient

At first, relay differential currents are calculated from state. By the use of the waveforms of Cf factors, athe currents that are measured at two sides of protected criterion is defined for the discriminative algorithm.transformer. These currents are calculated as follows [15]: The proposed criterion is based on the number of dips in

I = (I – I )–(I – I ) (17) shows later, the fault currents have only one dip and1 A a C c

I = (I – I )–(I – I ) (18) in the case of inrush current the number of dips are more2 B b A a

3 C c B b

A B C a b c

are secondary side currents of power transformer (Fig. 2)1 2 3

1 2

3

with a predetermined threshold to uses in the proposed

determined based on the specifications of power system,

j,j+1

The values of correlation factors are labeled by 'Cf 1', 'Cf1 2 3

the shape of the Cf factors. As the simulation results

sudden descent at the start instance of the fault. However

Fig. 2: Simulated power transformer

Selecting sliding window

Start

Three phase currents from CT’s

Calculating differential currents

Perform DWT for differential currents

Establish of energy percentage vector (Ed)

Calculation of correlation factor, between two reel vectors for three

differential currents (Cf 1, Cf 2, Cf 3)

The dips number of (Cf 1, Cf 2 or Cf 3)

> 1

No

Yes

Inrush current

Fault current

Trip signal

1I 2I 3I >

Yes

No

1, , 1 1, 2

, 1, 1 0.9

j j j j j j

j jj j

Onedip at Cfand

ρ ρ ρρ

ρ− + + +

++

> < ⇒ ↔ <

1Number of dips Inrush currentOtherwise Fault current

>


307

Fig. 3: Protection flowchart in the proposed method

than one and the dips repeat for next times until the inrushcurrent is settled. In counting the number of dips, thosedips are smaller than 0.9 are considered. The chosenmagnitude of 0.9 is because of the correlation factors thatare very near to the value of one and having morereliability at the algorithm and on the other handsimulation results show that the correlation factors forfault current are smaller than 0.9 after removing thetransient state. Thus, number of dips in the correlationfactors in the proposed algorithm is considered as:

(20)

The interpretation for the above equationwhich is in the calculation of number of dips,the values of correlation factors (D ) whichj,j+1

are smaller than the previous and next values andalso is smaller than the value of 0.9 are consideredas a dip.

Therefore, the number of dips at any of the 'Cf 1', 'Cf2' or 'Cf 3' exceeds one, it means that occurrence theinrush current otherwise it consider as a fault current.Reorganization criterion in the proposed algorithm isdefined as below:

(21)

If the recognition factors indicate a certaininternal fault in the differential current, a trip signal isissued to the power transformer circuit breakers to isolatethe transformer from the healthy system. The flowchart ofthis algorithm is shown in Fig. 3.

RESULTS

In this section effectiveness of the proposedalgorithm is evaluated for different types of faults, inrushcurrents and simultaneous occurrence of internal fault andinrush currents.

Power System Model: In order to generate the currentsignals for investigation of the performance of theproposed algorithm, a power system is simulated usingPSCAD/EMTDC simulation package. The single linediagram for the power system is shown in Fig. 4.It consists of a three phase 220 kV source connectedto a three phase, two winding, power transformer(300 MVA, YY 220/132kV) via a 20 km transmissionline. The neutral point of the Y connection is grounded.The transformer secondary is connected to a three phaseload (44S, 0.068H). Different cases of power systemcondition such as energizing condition, internalfaults (on 100% of transformer winding) andsimultaneous internal fault and inrush current are carriedout in this study.

Fig. 4: Simulated power system model

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -2 -1 0 1 2

KA

mp

Time(Sec)-(phase A)

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -10 -5 0 5

KA

mp

Time(Sec)-(phase B)

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -10

0

10

KA

mp

Time(Sec)-(phase C)

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -10 -5 0 5

10

KA

mp

Time(Sec)-(phase A)

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -10

0

10

KA

mp

Time(Sec)-(phase B)

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -20

0

20

KA

mp

Time(Sec)-(phase C)

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -20 -10

0 10 20

KA

mp

Time(Sec)-(phase A)

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -20

0

20

KA

mp

Time(Sec)-(phase B)

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 -20

0

20

KA

mp

Time(Sec)-(phase C)

0 0.1 0.2 0.3 0.4 0.5 0.9

0.95

1

Cf 1

0 0.1 0.2 0.3 0.4 0.5 0.8 0.85 0.9

0.95 1

Cf 2

0 0.1 0.2 0.3 0.4 0.5 0.8 0.85 0.9

0.95 1

Time(sec)

Cf 3


308

Fig. 5: Internal single line-to-ground fault current Fig. 7: Internal three line-to-ground fault current

Fig. 6: Internal double line-to-ground fault current Fig. 8: The Cf factors in case of internal single line-to-

Fault Current: To obtain the simulation data for internalfault, different faults are simulated inside protective zone other waveforms, because it depends on the currents ofof the transformer. These faults consist of phase to the healthy phases. Therefore, its value is alwaysground fault (single phase B to ground), double phase to approximately "1" similar to the normal condition andground fault (fault between two phases B, C to ground) there is no change in the Cf factor.and three phase to ground fault on the secondary side of Fig. 9 shows the case of an internal fault (doublethe power transformer. The starting time of faults is at phase to ground). In Fig. 9 the values of Cf factors are0.2 s and duration time of fault is 0.3 s. Figs. 5-7 show approximately "1" for normal condition. After occurrencethe differential fault current that are simulated at the the fault a rapid change is seen in the factors. Then,PSCAD software. similar to the previous case, the values of Cf reach close

Fig. 8 shows the Cf factors for three differential to the value of "1". In this figure, all differential currentscurrents (Cf 1, Cf 2, Cf 3) for single phase internal fault depend on the faulty phases, so all of plots have almost(fault B-G). As can be seen from this figure, the values of the same shape.Cf are approximately "1" for steady state until the time Another case of an internal fault (three phasethat the fault is occurs, then a rapid change and descent to ground) is shown in Fig. 10. In this figure allis appeared in these Cf values, then again come back to differential currents are related to the faulty phases;the value of "1". The waveform for I is not the similar to therefore, the values of Cf factors are approximately "1"1

ground fault

0 0.1 0.2 0.3 0.4 0.5 0.5 0.6 0.7 0.8 0.9

Cf 1

0 0.1 0.2 0.3 0.4 0.5 0.8 0.85 0.9

0.95 1

Cf 2

0 0.1 0.2 0.3 0.4 0.5 0.8 0.85 0.9

0.95 1

Time(sec)

Cf 3

0 0.1 0.2 0.3 0.4 0.5 0.5 0.6 0.7 0.8 0.9

Cf 1

0 0.1 0.2 0.3 0.4 0.5 0.7 0.8 0.9

1

Cf 2

0 0.1 0.2 0.3 0.4 0.5 0.7 0.8 0.9

1

Time(sec)

Cf 3

0 0.5 1 1.5 -0.5 0

0.5 1

1.5

KA

mp

Time(Sec)-(phase A)

0 0.5 1 1.5 -1.5 -1

-0.5 0

KA

mp

Time(Sec)-(phase B)

0 0.5 1 1.5 -0.1

0

0.1

KA

mp

Time(Sec)-(phase C)

0.2 0.4 0.6 0.8 1 1.2 1.4 0

0.5 1

1.5

KA

mp

Time(Sec)-(phase A)

0.2 0.4 0.6 0.8 1 1.2 1.4 -1.5

-1 -0.5

0

KA

mp

Time(Sec)-(phase B)

0.2 0.4 0.6 0.8 1 1.2 1.4 -0.1

0

0.1

KA

mp

Time(Sec)-(phase C)


309

Fig. 9: The Cf factors in case of internal double Fig. 11: Inrush current without residual magnetismline-to-ground fault

Fig. 10: The Cf factors in case of internal three

line-to-ground fault which the circuit breaker is closed and transformer is

for the normal condition. After occurrence the fault a fast amount of the residual magnetism in the transformer core,change in the factors is observed and then each of the Cf which may exists due to previous switching operations offactors returns to "1" value. It comes back to previous the transformer. If the polarity of the residual flux value bevalue shows occurrence an internal fault current. opposite of the steady state condition polarity, the dip

At all cases for fault currents the correlation factor value of the inrush current can increases. On the otherbecomes smaller than 0.9 at the start time of fault and hand, the residual magnetism can be considered bytransient state, after that it becomes approximately "1" switching the transformer for several times or the desiredand there is only one dip with the magnitude smaller than remnant magnetism can be set in unenergized transformer0.9 in the Cf values. with controlled dc current sources in PSCAD/EMTDC

Inrush Current: In the case of magnetizing inrush current current is collected while the transformer is unloaded andthat is created from transformer switching; two major the transformer is switched to the power system with thefactors affect the dip value of the magnetizing inrush circuit breakers on the primary side of the transformer.current [15]. The first factor is the energizing time instant, Several cases of inrush current with variable timewhich can be controlled through controlling the time at switching and different value of residual magnetism were

Fig. 12: Inrush current with residual magnetism

connected to the power system. The second factor is the

simulation model [18]. Differential magnetizing inrush

0 0.1 0.2 0.3 0.4 0.5 0.4 0.6 0.8

1

Cf 1

0 0.1 0.2 0.3 0.4 0.5 0.7 0.8 0.9

1

Cf 2

0 0.1 0.2 0.3 0.4 0.5 0.7 0.8 0.9

1

Time(sec)

Cf 3

0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.4 0.6

0.8 1

Cf 1

0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.7 0.8 0.9

1

Cf 2

0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.7 0.8 0.9

1

Time(sec)

Cf 3

0 0.5 1 1.5 -10

0 10 20

KA

mp

Time(Sec)-(phase A)

0 0.5 1 1.5 -20 -10

0 10

KA

mp

Time(Sec)-(phase B)

0 0.5 1 1.5 -10

0

10

KA

mp

Time(Sec)-(phase C)

0 0.1 0.2 0.3 0.4 0.5 0.4 0.6 0.8

1 C

f 1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

Cf 2

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

Time(sec)

Cf 3


310

Fig. 13: The Cf factors in case of inrush current without Fig. 15: Simultaneous internal fault and inrush current

residual magnetism

Fig. 14: The Cf factors in case of inrush current with fault and inrush currentresidual magnetism

simulated. In this paper switching time at 0.2 s, is shown. there are many dips for correlation factor and in theFig. 11 shows the differential inrush current without the proposed algorithm the number of dips for the correlationresidual magnetism and Fig. 12 shows the differential factors (Cf 1, Cf 2, Cf 3) becomes greater than one andinrush current for two times switching of transformer and differential relay is blocked.therefore existence of residual magnetism.

Figs. 13 and 14 illustrate the Cf factors in the case of Simultaneous Internal Fault and Inrush Current:an inrush current (without and with residual magnetism, After studying fault and inrush currents cases separately,respectively). According to Figs. 13 and 14, it is obvious some more complicated cases (simultaneous internal faultthat the Cf factors for two cases is approximately "1", and inrush current) are examined. Fig. 15 shows thesimilar to previous cases in about faults, for normal differential current signal for the case of simultaneouscondition. After switching the transformer a sudden internal fault and inrush current, whereas switching ischange in detection criterion (Cf) value is appeared, but at the primary side and fault (three phases to ground)waveforms do not come back to the previous values, is on the secondary side of the transformer andclose to the value of "1" and have an oscilatory nature. starting time of fault and transformer switching is at 0.2 s.Therefore, Cf factors go away from the value of "1" and Fig. 16 shows the Cf factor for three differential currents

Fig. 16: The Cf factors in case of simultaneous internal

become lower than the defined threshold (0.9). As a result,


311

(I ,I , I ) for simultaneous inrush and fault current. dips number is one and there is only at the start time of1 2 3

The values of Cf factors for these shapes are fault. The simulation results show fast and reliableapproximately "1" for normal operation until the time of capabilities of the proposed algorithm in identifying theoccurrence the simultaneous fault and inrush, that shows different types of currents flowing in a power transformera sudden change in the Cf factor value for the differential under various conditions.currents, then it comes back to "1" value, similar to faultcases. Therefore, there is only one dip at any of the Cf REFERENCESfactors and the fault is identified fast and reliably and donot make a mistake in detection algorithm. 1. Rahman, M.A. and B. Jeyasurya, 1988.

According to represented cases, the discrimination A State-of-the-art Review of Transformer Protectionbetween inrush and fault current is performed accurately Algorithms. IEEE Transactions on Power Delivery,with considering the Cf 1, Cf 2 and Cf 3 versus the time 3(2): 534-544.axis. After the desired signal decomposition in to five 2. Liu, P., O.P. Malik, D. Chen, G.S. Hope and Y. Guo,levels by using the DWT, the energy percentage 1992. Improved Operation of Differential Protection ofvectors are established, as it was explained before. Power Transformers for Internal Faults. IEEEThen, correlation factors are plotted for each current Transactions on Power Delivery, 7(4): 1912-1919.signal (Cf 1, Cf 2, Cf 3). The simulation results depict that 3. Yabe, K., 1997. Power Differential Method fora sudden change in waveform of Cf factor shows the Discrimination between Fault and Magnetizing

occurrence of a disturbance. Then, the algorithm Inrush Current in Transformers. IEEE Transactionscomputes the number of dips for Cf values for each three on Power Delivery, 12(3): 1109-1118.phase differential currents. The exceed of dips number 4. Sidhu, T.S. and M.S. Sachdev, 1992. On Lineshows inrush current; otherwise, it is considered as fault Identification of Magnetizing Inrush and Internalcurrent and a trip signal is sent to the power transformer Faults in Three Phase Transformers. IEEEcircuit breakers. Therefore, the fault and inrush currents Transactions on Power Delivery, 7(4): 1885-1890.can be detected less than time of 3 ms that is a good time 5. Zhang, H., J.F. Wen, P. Liu and O.P. Malik, 2002.for detection of fault. Discrimination between Fault and Magnetizing

CONCLUSION Correlation Transform. Electrical Power and Energy

In this study a new algorithm for discrimination 6. Shin, M.C., C.W. Park and J.H. Kim, 2003. Fuzzy Logicbetween inrush and internal fault currents is proposed Based Relaying for Large Power Transformerusing the DWT. The differential currents of the power Protection. IEEE Transactions on Power Delivery,transformer are decomposed into five levels. 18(3): 718-724.The algorithm is based on the energy percentage of the 7. Perez, L.G., A.J. Flechsig, J.L. Meador andDWT coefficients. The correlation factor between energy Z. Obradovicc, 1994. Training an Artificial Neuralpercentage vectors of the detail coefficients is calculated Network to Discriminate Between Magnetizing Inrushat each sliding window. Detection process is then and Internal Faults. IEEE Transactions on Powerperformed by determining the Cf factors versus the Delivery, 9(1): 734-441.time and computing the dips number of them. 8. Tripathy, M., R.P. Maheshwari and H.K. Verma, 2010.The proposed algorithm is tested on a power system Power Transformer Differential Protection based onmodel. Several cases of the inrush currents, internal faults Optimal Probabilistic Neural Network. IEEEand also simultaneous inrush and fault currents are Transactions on Power Delivery, 25(1): 102-112.simulated and some of them are represented in this paper. 9. Pandy, S.K. and L. Satish, 1998. MultiresoulutionThe waveshapes from the simulated cases show that the Signal Decomposition: A New tool for fault detectionCf factors for inrush currents have oscillatory nature but in power transformers during impulse tests. IEEEfor internal fault currents they settle near the one value. Transactions on Power Delivery, 13(4): 1194-1200.By use of this property a threshold is determined that is 10. Youssef, O.A.S., 2003. A Wavelet Based Techniquethe dips number of Cf. For an inrush current, number of for Discrimination between Faults and Magnetizingdips for the correlation factors goes away from one dip Inrush Currents in Transformers. IEEE Transactionsand an internal fault current that has a smooth nature, the on Power Delivery, 18(1): 170-176.

Inrush Current in Transformers Using Short-Time

Systems, 24: 557-562.


312

11. Morante, M.G. and D.W. Vicoletti, 1993. 15. HossumEldin, A.A. and M.A. Refaey, 2011. A NovelA Wavelet Based Differential Transformer Algorithm for Discrimination between Inrush CurrentProtection. IEEE Transactions on Power Delivery, and Internal Faults in Power Transformer Differential8(4): 1351-1358. Protection Based on Discrete Wavelet Transform.

12. Sendilkumar, S., B.L. Mathur and J. Henry, Electric Power System Research, 81(1): 1-6.2010. Differential Protection for Power 16. Mao, P.L. and R.K. Aggarwal, 2001. A NovelTransformer Using Wavelet Transform and PNN. Approach to the Classification of the TransientInternational J. Electrical and Electronins Phenomena in Power Transformers Using CombinedEngineering, pp: 468-474. Wavelet Transform and Neural Network. IEEE

13. Akansu, A.N. and R.A. Haddad, 1992. Transactions on Power Delivery, 16(4): 654-660.Multiresolution Signal Decomposition (Transforms, 17. Mukhopadhyay, N., 2011. International EncyclopediaSub-bands and Wavelets). Second edition, New of Statistical Science (Chapter: CorrelationJersey Institute of Technology, Newark, Nj, Coefficient). International Encyclopedia of StatisticalAcademic press. Science, Springer-Verlag Berlin Heidelberg,

14. Addison, P.S., 2002. The Illustrated Wavelet University of Connecticut-Storrs, Storrs, CT, USA.Transform Handbook. Napier University, Edinburgh, 18. Woodford, D., 2001. Introduction to PSCAD V3.UK, IOP publishing Ltd. In Manitoba HVDC Research Centre Inc, Winnipeg,

MB, Canada.

difference between inrush and fault current.pdf

Documents

Transcript of difference between inrush and fault current.pdf