Sie sind auf Seite 1von 6

Modeling and Real-Time Simulation of High

Impedance Faults for Protection Relay Testing and


Methods Validation
Aldair Wontroba
Adriano Peres de Morais João Paulo Abreu Vieira
Jean Pereira Rossini Electrical Engineering Graduate Program
Mairon Gallas Federal university of Pará
Belém-PA, Brazil
Ghendy Cardoso Junior jpavieira@ufpa.br
Federal University of Santa Maria
Santa Maria-RS, Brazil
aldair.wontroba@hotmail.com Marcelo Costa Santos
Distribution System Operation Management
CELPA –Electric Distribution Utility of Pará
Patrick Escalante Farias Belém-PA, Brazil
Federal Institute of Education, Science and Technology of Rio marcelo.costa@celpa.com.br
Grande do Sul - IFRS
Farroupilha, Brazil
pefarias@gmail.com

Abstract— This work presents a real-time model of high moisture in the surface. The arc restarts when the conductor
impedance fault. The model allows full relay validation reaches another region with enough moisture to reestablish an
including different types of ground surface. The model is electric path. Asymmetry is originated by the silica present in
incorporated into OPAL-RT simulator and implemented in the surface that causes a great drop voltage during the negative
Hypersim software, a comparison with ATPdraw is performed. semi cycle of the current so the positive semi cycle has a
The model and its theoretical basis are presented; the real-time greater magnitude [5]. The asymmetry is also related to the
implementation, its results and the issues are discussed. nonlinear V x I behavior at the fault point, which is a typical
HIF feature caused by the electric arc as well.
Keywords— High impedance fault, modeling, real time
simulation. Current

I. INTRODUCTION Shoulder
Asymmetry up
Before approving and commissioning feeder protection Bu
ild
relays, extensive testing is required, either as laboratory (or
field) experimental testing or HIL simulations, to validate
protection schemes and determine appropriate settings. The
first approach is time-consuming and costly while the latter is
limited by the simulation models' abilities to represent
Intermittence
adequately the feeder behavior during all types of faults.
High Impedance Faults (HIF) are one of the most common
type of fault in power distribution networks. HIF typically Time
occurs when the conductors in distribution network break and
touch the ground surface, i.e. asphalt, tree, concrete, etc. Due Fig. 1. Characteristics of HIF.
to the high impedance of these surfaces, the current at the fault
point has generally a low magnitude. In technical terms, the A correct fault model is indispensable for some power
IEEE PSRC [1] defines HIF as fault with current rates below system studies, essentially for HIF detection which most of the
the overcurrent relay’s threshold. Moreover, other features can methodologies are based on some of the inherent HIF features.
also define a HIF according to Fig. 1, such as, buildup and The majority of HIF detectors uses somehow the harmonic
shoulder stages, intermittence and asymmetry of the fault content present in the fault current [6]-[8]; each HIF feature is
current [2]-[4]. Buildup is the period in which the fault current responsible for a different content: low frequency
increases gradually due to accommodation of the cable in the (nonlinearity of V x I curve), even harmonics (asymmetry),
ground while the shoulder stage refers to the periods of high frequency (intermittence) and inter-harmonics (shoulder
constancy in the magnitude. Intermittence occurs when the and buildup stages)[9]. Methods based on the HIF current
HIF electric arc is extinct during some time due to the loss of shape are also widely applied to fault detection, which wavelet
[10]-[11] and mathematical morphologic [12]-[13] methods
The authors acknowledge the technical and financial support of National can be highlighted due to their capability of recognize
Institute of Science and Technology on Distributed Generation Power
Systems (INCTGD), Higher Level Personnel Training Coordination
transients and distortions present in the current signal.
(CAPES/PROEX) - Finance Code 001, National Council for Scientific and HIF detection is a research topic which has been
Technological Development (CNPq), Foundation for Research of the State of
Rio Grande do Sul (FAPERGS), Federal University of Santa Maria (UFSM)
investigated for many years. However, in terms of Intelligent
and Federal University of Pará (UFPA), Brazilian Institutions. Electronic Devices (IEDs) available commercially, there is
This paper was part of an R&D project with ANEEL, CELPA and only one solution able to detect it [14]. As the classical
CEMAR. overcurrent relays do not detect this type of fault, it is difficult

978-1-5386-8218-0 / 19 / $ 31.00 © 2019 IEEE


to find oscillographic records of HIF to validate IEDs and new (2). The order of this polynomial, as well as its coefficients,
methods for detection of HIF. Basically, the proposed depends on the type of soil in which the HIF occurred.
methods use eletromagnetic softwares, to validate it and
comercial IEDs spend a long time installed in test mode. c k .t k + c k −1.t k −1 + ... + c1.t + c 0 , se t < Δt
Hence, this paper presents a HIF model [5] for real time R1 (t ) =  (2)
simulation. The model is incorporated into OPAL-RT  10 −5 , se t ≥ Δt
simulator and implemented in Hypersim software. The digital Where: k is the degree of the polynomial and index of each
model can be applied for validation of HIF detection methods coefficient; c are the coefficients of the polynomial; Δt is the
and real-time hardware-in-the-loop (HIL) testing of IEDs. The period of growth of the fault current.
results obtained are compared with HIF model implemented
on ATPDraw. B. R2 Control
The remainder of this paper is organized as follows. The second resistance (R2) models the asymmetric
Section II presents a literature review of HIF model. Section characteristic of the V x I curve due to the presence of the
III presents the HIF modeling for real-time simulation. The electric arc and is present in all cycles of the fault current. For
results obtained by the real-time model are presented and this, a routine checks if the voltage at the fault point is in the
compared with ATPDraw results on Section IV. Section V increasing branch (positive derivative) or decreasing (negative
includes the conclusions derived from this work. derivative) of the V x I curve of the surface in which the HIF
will be simulated. After verifying the branch, the routine uses
II. LITERATURE REVIEW (3) to obtain the respective fault current of the V x I curve.
There are two widely accepted models for HIFs simulation Finally, the corresponding resistance is calculated through the
in the literature, which are the models proposed in [15] and OHM Law.
[16]. The two models present similar results; however, unlike
 i Fi (m + 1) − i Fi (m)
model [15], model [16] is unable to represent all the HIF i Fi (m) +v (m + 1) − v (m) . Δv,
characteristics as buildup and shoulder stages. In [5] was  Fi Fi
proposed a modification in the model initially presented by  if v (m) < v (t ) < v (m + 1)
[15]. This modification makes it possible to simulate HIF on i F (t ) =  Fi F Fi (3)
different types of ground surfaces. For these reasons, the real- 
 i Fi (m),
time implementation presented in this paper is an

improved version of [5].  if v F (t ) = v Fi (m)
The model proposed in [15] consists of two variable Where: vF(t) is the voltage at the fault point; Δv = vF(t) - vFi(m);
resistances in series. The first one (R1) models the growing vFi(m) and iFi(m) are the voltage; the current of the
feature of the current during the settling time of the conductor characteristic curve V x I in sample m, respectively.
in the soil (buildup stage). The second one (R2), models the
typical asymmetry of the electric arc and is present in all fault In the first version of the model proposed by [15], R2 was
current cycles. Fig. 2 represents the model of variable calculated only with the data of the V x I curve for a stone
resistance. surface. However, in the modification proposed by [5], the
Fault Point
model became more complete, adding the V x I curves of six
vF1 other surfaces according to Fig. 3.

R1 III. HIF MODELING FOR REAL-T IME SIMULATION


iF1
Models of equipment and different conditions of operation
of the electric power system can be modeled for real-time
R2 simulations. Real-time simulation allows the evaluation of the
tested device's impact on the power grid and the identification
of adverse interactions with existing equipments and controls.
To achieve such a goal, a digital real-time simulator needs to
Fig. 2. Model of variable resistance. solve the model equations for one time-step within the same
In consequence, the voltage at the fault point (vF) is given time in real-world clock [17]-[18].
by (1). The HIF model was implemented and simulated with
v F (t) = v R1 (t)+ v R2 (t) (1) Opal-RT real time simulator and Hypersim software. Opal-
RT/Hypersim simulator performs real-time simulations and
Where: vF(t) is the instantaneous voltage at the fault point; non-real-time simulations. The simulation is considered to be
v R1 (t ) and v R2 (t ) are the instantaneous voltage on the real-time if the runtime, Te, for the system simulation is less
than or equal to the selected time, as shown Fig. 4 a, and case,
resistors R1 and R2, respectively. Te is greater than its time-step size for one or more time-steps,
A. R1 Control overruns occurs and the simulation is considered as non-real-
The R1 resistance emulates the buildup period, that is, the time, as shown Fig. 4 b.
growth characteristic of the fault current during the period of The Hypersim software can applied for the two categories
accommodation of the cable in the ground. For this, R1 has a of real time simulations which can include hardware, IED for
high value in the beginning of the fault and decreases its value instance, and software-in-the-loop processes, algorithm
gradually over time to a value close to zero. To represent this validation, for instance. In this work the test system was
behavior, [5] proposed the use of a polynomial according to
(a) (b) The control part performs functions using signals from UCM
1 1
Ibase=10A control inputs and internal signals in order to
Ibase=10A
produce signals to send to the UCM control outputs or to be
used internally. The Fig. 5 shows the UCM block used for HIF

vFi (pu)
v Fi (pu)

0 0 model.

-1 HIF
-1 Calculation
-1 0 1 -1 0 1 Vfault R Rin
i Fi (pu) i Fi (pu)
(c) (d) Rcontrolled
1 1
Ibase=20A Fig. 5. HIF model in UCM block.
Ibase=40A
The test system used is presented in Fig. 6. Where the lines
vFi (pu)
vFi (pu)

0 have 10 km each. The parameters of the lines can be seen in


0
the TABLE I. A load of 24 kVA, PF = 1.0 is considered at the
end of the feeder. The nominal voltage of the system is 13.8
kV and 60 Hz. The HIF insertion occurs between the two lines
-1 -1 in only one phase.
-1 0 1 -1 0 1
i Fi (pu) i Fi (pu) Vsrc

(e) (f) +
1 1
Ibase=100A Ibase=100A
v Fi (pu)
v Fi (pu)

HIF
0 0
Fig. 6. Test system used for HIF
TABLE I. IMPEDANCE MATRIX OF LINES.
-1 -1
-1 0 1 -1 0 1 Zero Positive
i Fi (pu) i Fi (pu) R (Ω/km) 1.06 0.501
L (H/hm) 7.35E-3 2.31E-3
Increasing branch Decreasing branch C (F/km) 0.0 0.0

Fig. 3. V x I Curves at the HIF point corresponding to one cycle: a) sand, b) The HIF model corresponds to the one proposed by [5],
asphalt, c) gravel, d) cobblestones, e) grass, f) local soil. The user can simulate up to 6 different types of ground
implemented within the graphic interface of Hypersim, and surfaces. Each soil is identified using a pre-defined number
the HIF model was implemented in a UCM block (User Code and can be choose by user during the simulation. As described
Model), which not only have control interface but also is previously, the model is represented by two resistors. The first
equipped with electrical interfaces. The UCM is a Hypersim resistor R1 is modeled by a polynomial as a function of time,
utility that permits users to build their specific blocks in the where each soil has different coefficients (TABLE II). As the
model using a C-coded function while respecting the time increases after HIF occurrence, the resistance R1 tends to
Hypersim programming rules. stabilize. At this point the polynomial is no longer considered.
This is done because the polynomial cannot represent R1 for
Idle time a long time. Unlike [5], it is not adopted 10-5 after ≥ ∆ , but
Te(n-1) Te(n) Te(n+1)
rather the last calculated value (i.e. = ∆ ) for R1. TABLE
t III shows the times that R1 needs to stabilize.
t TABLE II. COEFFICIENTS OF POLYNOMIALS FOR EACH TYPE OF SOIL.
tn-1 tn tn+1 tn+2
Type of soil
(a) Coeff
icient Cobbles
Sand Asphalt Gravel Grass Local soil
Overrun tone
Te(n-1) Te(n)
c8 3.25E11 3.2E12 1.29E8 6.61E9 6.5E10 0
t c7 -2.9E11 -3.0E12 -2.92E8 -1.66E9 -5.7E10 0
c6 1.14E11 1.1E12 2.73E8 1.91E9 2.0E10 0
t c5 -2.3E10 -2.5E11 -1.37E8 -8.99E8 -4.00E9 -3.05E6
tn-1 tn tn+1 tn+2 c4 2.84E9 3.1E10 4.00E7 2.18E8 4.47E8 2.66E6
c3 -1.98E8 -2.3E9 -6.89E6 -2.87E7 -2.91E7 -8.08E5
(b)
c2 7.74E6 9.52E7 6.88E5 2.01E6 1.05E6 1.02E5
Fig. 4 Illustration of real-time and offline simulation. (a) Real-time c1 -1.55E5 -1.98E6 -3.79E4 -7.10E4 -2.03E4 -6.78E3
simulation. (b) Non-real-time simulation. c0 1.68E3 1.68E4 1.10E3 1.34E3 2.24E2 5.13E2

A UCM is composed of a power part and a control part: TABLE III. TIME NEEDED TO R1 STABILIZE.
The power part has external nodes (graphically visible) and Aspha Cobble Local
internal nodes (invisible). This part is represented as an Sand Gravel Grass
lt stone soil
admittance matrix which will be added to the substation Time (s) 0.2 0.2 0.5 0.2 0.3 0.3
admittance matrix to solve the node equation.
The second resistor, R2, is given by the linear interpolation
of points in a V x I curve as previously seen in Fig. 3, which 20
represents the characteristic of the soil type. The intermittence ATP
is not considered in this model. The two resistors are HYPERSIM
calculated in each simulation time step within the UCM block, 0
and then summed and assigned to a controlled resistor to
emulate the HIF.
IV. RESULTS AND PERFORMANCES -20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (s)
In order to validate the results, the same test system and (a)
the HIF model were implemented in the ATPDraw software. 10
The results of the digital model were compared with
ATPDraw because it is already consolidated in other works ATP
[5] and [19]. Similar to Hypersim, a programming block, HYPERSIM
0
using the MODELS language, was used in the ATPDraw to
represent the HIF model. By comparing the results obtained in
the Hypersim with the ATPDraw, it was possible to verify the
reliability of the model for real time simulation. -10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Fig. 7 shows the current growth of the HIF, a process Time (s)
called buildup and shoulder, in which the resistance R1 has (b)
the function of representing. Note that in both simulators we 50
have a very similar response to the current growth. ATP
In order to measure the difference in response of the HYPERSIM
simulations, the equation (4) was used. 0

| |
(%) = 100% ∙ ∙ ∑ (4)
| ( )| -50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Where is the instantaneous current in the Hypersim and Time (s)
is the current in the ATPDraw. The sum of the percentage (c)
of the error divided by the number of samples is then made to 100
obtain the mean value. The average percentage difference in ATP
one cycle (shoulder period) was close to 1% for sand, asphalt, HYPERSIM
gravel and cobblestone. Grass and local soil have presented an 0
error above of 1%. The Table IV reports the values of the
average percentage error of Hypersim compared to the
ATPDraw. -100
TABLE IV. AVERAGE PERCENTAGE ERROR. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Sand Asphalt Gravel Grass Cobbles Local Time (s)
tone soil (d)
Error (%) 1.01 1.23 1.9 2.6 0.81 4.64 50
ATP
Fig. 8 shows the current of the HIF during shoulder period HYPERSIM
for all ground surface implemented. It is observed that the 0
waveforms of Hypersim and ATPDraw are very similar. It is
also noteworthy that the nonlinearity characteristics of the HIF
current are present in all soils, with asphalt, gravel and local
soils being more significant. -50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
For HIL testing a real-time simulation is required. Hence, Time (s)
the HIF model have been tested in real time mode. The real- (e)
time simulation was done with a 50-μs time step. The analog 100
signal of the voltage and current of the HIF were measured by ATP
an oscilloscope. The Fig. 9 represents the oscillography of HYPERSIM
0
voltage of fault point (green) and current (yellow) of the HIF
for gravel soil.
The tests were performed using two digital to analog ports -100
of the OPAL-RT, one for voltage and one for current. The 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
experimental setup is shown in Fig. 10. Time (s)
(f)
Fig. 7. Growth of the corrent for a) sand, b) asphalt, c) gravel, d) grass,
e) cobblestone, f) local soil.
Fig. 8. Current of HIF during a shoulder period for a) sand, b) asphalt, c) gravel, d) grass, e) cobblestone, f) local soil.

Fig. 9. Waveform of Voltage (green) and Current (yellow) of the HIF Fig. 10. Experimental setup using OPAL-RT for real-time simulation of
model for the gravel soil, seen in the oscilloscope. HIF model.

The accuracy, simplicity and flexibility make this model


V. CONCLUSIONS reliable and convenient for the study of HIF in distribution
A flexible HIF model and its real-time implementation systems, which may be very useful to develop
were discussed and its usage demonstrated for HIF relay an accurate HIF detection methods and test IEDs.
testing and methods validation. This method allows The HIF model implemented in this work will be base for
representation of all the HIF characteristics as buildup stage, the future's works in real-time simulations using OPAL-RT
shoulder stage and asymmetry of the electric arc. Furthermore, simulator.
the presented model considers six types of soil. The HIF
model was implemented in a real-time simulator of large It is intended, with the model: to perform tests in relays of
power networks, called Hypersim. HIF detection; test HIF detection and localization techniques
in distribution systems; and enable HIL testing using OPAL-
The simulation results illustrate the behavior of the fault RT/Hypersim.
current is very similar to ATPDraw for all types of soil.
VI. REFERENCES [10] I. Baqui, I. Zamora, J. Mazn, G. Buigues, “High impedace fault
detection methodology using wavelet transform and artificial neural
[1] J. Tengdin, et al.; “High Impedance Fault Detection Technology”, networks”, Electric Power Systems Research, vol. 81, pp. 1325-1333,
Report of PSRC Working Group D15, March 1996. July 2011.
[2] Ramos, M.g; Bretas, A. S.; Bernardon, D. P.f; Pfitsher, L.f, [11] F. B. Costa, B. Souza, N. Brito, J. Silva, W. Santos, “Real-time
"Distribution Networks HIF Location: A Frequency Domain System detection of transients induced by high-impedance faults based on the
Model and WLS Parameter Estimation Approach". Electric Power boundary wavelet transform”, IEEE Transactions on Industry
Systems Research v. 146, p. 170-176, 2017. Applications, v. 51, pp. 5312-5323, May 2015.
[3] Nunes, J. U. N.g; Bretas, A. S.; Herrera-Orozco, A.g; Iurinic, L.g. [12] S. Gautam and S. M. Brahma, “Detection of High Impedance Fault in
Distribution Systems High Impedance Fault Location: A Spectral Power Distribution Systems Using Mathematical Morphology”, IEEE
Domain Model Considering Parametric Error Processing. International Transactions on Power Systems, vol. 28, no. 2, pp. 1226-1234, May
Journal of Electrical Power & Energy Systems, v. 109, p. 227-241, 2013.
2019. [13] M. Sarlak and S. Shahrtash, “High-impedance faulted branch
[4] Ferraz, R.g, Iurinic, L.g, Filomena, A.g, Gazzana, D.g, Bretas, A. S., identification using magnetic-field signature analysis”, IEEE
"Arc Fault Location: a Nonlinear Time Varying Fault Model and Transactions on Power Delivery, vol. 28, January 2013.
Frequency Domain Parameter Estimation Approach". International [14] A. Avendaño, “Testing the High-Impedance Fault Detector Element in
Journal of Electrical Power & Energy Systems, v. 80, p. 347-355, the SEL-751 and SEL-451”, SEL Application Guide, vol. 3, 2015.
2016. [15] S.R. Nan, J. K. Park, Y. C. Kang and T. H. Kim, “A modeling method
[5] W. C. dos Santos, B. A. de Souza, N. S. D. Brito, F. B. Costa and M. of a high impedance fault in a distribution system using two series time-
R. C. Paes Jr., “High Impedance Faults: From Field Tests to varying resistances in EMTP”, Power Engineering Society Summer
Modeling”, Journal of Control, Automation and Electrical Systems, pp. Meeting, pp. 1175–1180, July 2001.
885-896, September 2013. [16] A. E. Emanuel, D. Cyganski, J. A. Orr, S. Shiller, E. M. Gulachenski,
[6] S. M. Shahrtash, M. Sarlak, “High impedance fault detection using “High impedance fault arcing on sandy soil in 15 kV distribution
harmonics energy decision tree algorithm”, International Conference feeders: contributions to the evaluation of the low frequency
on Power System Technology, pp. 1–5, 2006. spectrum”, IEEE Transactions on Power Delivery, vol. 5, pp. 676-686,
[7] H. G. Yeh, D. H. Tran, R. Yinger. “High impedance fault detection April 1990.
using orthogonal transforms”, IEEE Green Energy and Systems [17] P. M. Menghal and A. J. Laxmi, ‘‘Real time simulation: Recent
Conference, November 2014. progress & challenges,’’ in Proc. Int. Conf. Power, Signals, Controls
[8] S. R. Samantaray, P. K. Dash, S. K. Upadhhyay, “Adaptive Kalman Comput. (EPSCICON), Thrissur, India, Jan. 2012, pp. 1–6.
Filter and Neural Network Based High Impedance Fault Detection in [18] T. Berry, A. R. Daniels, and R. W. Dunn, ‘‘Real time simulation of
Power Distribution Networks”, International Conference on Power power system transient behaviour,’’ in Proc. 3rd Int. Conf. Power Syst.
System Technology, October 2006. Monitor. Control, London, U.K., Jun. 1991, pp. 122–127.
[9] E. M. Lima, C. M. S. Junqueira, N. S. D. Brito, B. A. de Souza, R. A. [19] P. E. Farias, “Método para estimação da distância de faltas de alta
Coelho, H. G. M. S. de Medeiros, “High impedance fault detection impedância em redes de distribuição de energia elétrica considerando
method based on the short-time fourier transform”, IET Generation, diferentes tipos de solo”, Universidade Federal de Santa Maria, 2017.
Transmission & Distribution, vol. 12, pp. 2577-2584, June 2018.

Das könnte Ihnen auch gefallen