Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/276558026
READS
42
2 AUTHORS:
Dos Santos
University of Lisbon
9 PUBLICATIONS 5 CITATIONS
SEE PROFILE
SEE PROFILE
Instituto Superior Tcnico, Universidade de Lisboa, Av. Rovisco Pais 1, 1049-001, Lisboa, Portugal
Rede Elctrica Nacional S.A., Rua Cidade de Goa 4, 2685-038, Sacavm, Portugal
a r t i c l e
i n f o
Article history:
Received 25 August 2014
Received in revised form 2 March 2015
Accepted 26 April 2015
Keywords:
Power system faults
Fault rate
Fault type
Fault resistance
Stochastic modeling
Short-circuit computation
a b s t r a c t
Correct modeling of power system faults is a key issue in a diversity of power system studies, such as
in network planning, equipment specication and protection systems coordination. The present paper
addresses the probabilistic description of faults, based on available data collected by transmission system
operators for different voltage levels. Fault rate and individual fault characteristics are stochastically
modeled, namely fault location, type and resistance. A fault resistance model is suggested, based on
Weibull distribution, which parameters are set per voltage level. The proposed fault description is a
useful tool for power system planning and design, when a stochastic approach of the power system
faults characteristics is adopted. Time series of fault input data and simulation results are presented in a
common format, so to allow using the same statistical tools as used in power system monitoring and eld
data reporting. The model is able to reproduce atypical years, as happen in real transmission networks.
The developed fault model is used to generate stochastic short-circuit events, which are then used for
short-circuit current computation. The methodology is applied to the IEEE RTS and simulation results are
shown for the probability of amplitude and time constant values. These results are prone to be used to
specify network circuit breakers and current transformers using a probabilistic approach.
2015 Elsevier B.V. All rights reserved.
1. Introduction
Correct modeling of power system faults is a key issue in a
diversity of power system studies, such as in network planning,
equipment specication and protection systems coordination. Currently used methods, underlying international guidelines and
standards [13] are deterministic, based on worst case scenarios.
These are typically associated to extreme events leading to the
highest or lowest short circuit currents. Consequently, the deterministic approach results into over-dimensioning during most of
the operating time. Furthermore, as the likelihood of occurrence
is disregarded, no quantitative information is provided to support
investment decisions.
Contrary to deterministic methods, a probabilistic approach
quantitatively incorporates uncertainty, based on quantifying the
likelihood of each event, and eventually its associated uncertainty.
The approach allows quantifying the degree of exposure to a particular event and this information may be combined with associated
costs in a risk analysis methodology.
Corresponding author at: Rede Elctrica Nacional S.A., Av. Estados Unidos da
America,
55, 1749-061 Lisboa, Portugal. Tel.: +351 210013464.
E-mail addresses: andre.santos@ren.pt (A. dos Santos),
teresa.correiadebarros@ist.utl.pt (M.T.C.d. Barros).
http://dx.doi.org/10.1016/j.epsr.2015.04.015
0378-7796/ 2015 Elsevier B.V. All rights reserved.
30
A. dos Santos, M.T.C.d. Barros / Electric Power Systems Research 126 (2015) 2937
3. Fault modeling
3.1. Modeling approach
In the present paper, a fault simulation methodology based on
Monte Carlo simulation is presented. The occurrence of faults in
a network element, such as a transmission line or a busbar, is
simulated during a time span of representative signicance. The
simulation time span is divided into an equal number of time intervals (trials) of equal duration, such as a second, an hour, or a year.
The methodology output is a set of occurrences, each characterized
t0
PF (t0 ) = 1 exp
z () d
(1)
For each trial, a random number is generated according to a uniform distribution over the interval [0,1]. The generated number is
compared to the corresponding probability of fault occurrence PF
and, if higher, the corresponding network element is assumed to
fail during the trial.
Recorded network performance data usually refer to yearly line
faults per 100 km, l , and yearly busbar faults per 100 busses, b .
Accordingly, a constant failure rate zi (t) = i is assumed for each
network element. All busbars will have the same failure rate and
transmission lines will have different failures rates, according to
their length li :
bi =
b
100
(2)
li =
li l
100
(3)
A. dos Santos, M.T.C.d. Barros / Electric Power Systems Research 126 (2015) 2937
31
Fig. 1. Worst fault location inside a substation, corresponding to the highest interrupted current.
of each fault type. A random number is associated to each fault, generated according to a uniform distribution over the dened interval.
The segment corresponding to the generated number denes the
fault type.
3.5. Fault resistance
Only transmission line faults involving the ground are assumed
to be characterized by a fault resistance. The fault resistance associated to SLG and 2LG faults is described statistically. Field data have
been published corresponding to 230 kV and 138 kV [20], as well as
for 60 kV [21]. A study developed by the authors, briey outlined
in Appendix, shows that published data can be tted by Weibull
distributions:
f rf
r 1
f
rf
exp
(2)
Fault currents feed from network side A, IAB , and side B, IBA , when
both circuit breakers CBA and CBB are closed;
Fault current IA when CBB is open;
Fault current IB when CBA is open.
As regards the short circuit current time constant Ta , it is dened
for the positive sequence I+ , considering the network topology with
all circuit breakers closed, as suggested in [23]:
Ta =
Xeq
0 Req
(7)
where, 0 is the power frequency and, Xeq and Req are dened by:
Req + jXeq =
V f
I+
(8)
(5)
4. Application examplefault characterization
where,
F 1 (v) = 1 e(v/)
Transmission line fault clearance requires the operation of circuit breaker at both ends. As this operation depends on the tripping
command by the local line protection system, it is not reasonable to assume that all circuit breakers will interrupt the fault
current simultaneously. The opening sequence will determine the
short circuit current magnitude interrupted by each circuit breaker.
Considering a transmission line fault in a meshed network, which
equivalent is represented in Fig. 2, any of the circuit breakers CBA
or CBB may operate rst. The rst only interrupts its own fault current contribution, while the last extinguishes the short circuit as
the interrupted current is the total short circuit current. Since any
of the circuit breakers may be the last to operate, four short circuit
currents contributions have to be evaluate to identify the highest
interrupted current (Fig. 3):
(6)
32
A. dos Santos, M.T.C.d. Barros / Electric Power Systems Research 126 (2015) 2937
Table 1
Generating units modelassumptions.
Generator
Transformer
20
50 and 76
100
350
10
400
0
.064
2050
0
.095
50200
0
.143
200500
Fig. 4. Line fault rate in the Portuguese transmission network [26]220 kV (solid)
and 150 kV (dash).
in Fig. 5. The average value in the time span 20062012 was considered: 0.29 faults/100 busses/year (138 kV) and 0.46 faults/100
busses/year (230 kV).
Internal faults in the generating units and substation transformers were not considered.
4.1.6. Fault type
The fault type distribution for the 138 kV and 230 kV transmission lines was obtained from a statistical study conducted by
the authors for the Portuguese Transmission System, from 2004
to 2012, after analyzing all network faults. The fault type distribution was calculated, on a yearly basis, for the 150 kV and 220 kV
networks. Results are presented in Figs. 6 and 7, including the statistical values: average and standard deviation. The average values
have been considered as input data for the simulation.
Concerning busbar fault type distribution, no eld data was
available to the authors. Therefore this distribution was derived
from the fact that busbar faults are typically due to insulator
ashover or switching human error, resulting into SLG faults and
3PH faults, respectively. Since the probability of occurrence of
human error is much lower than insulator ashover, an 80%20%
distribution was considered.
Fig. 5. Bus fault rate in the Portuguese transmission network220 kV (solid) and
150 kV (dash).
A. dos Santos, M.T.C.d. Barros / Electric Power Systems Research 126 (2015) 2937
33
Fig. 6. Fault type in the 150 kV Portuguese transmission network from 2004 to 2012 (percentage), including computed average and standard deviation.
Fig. 7. Fault type in the 220 kV Portuguese transmission network from 2004 to 2012 (percentage), including computed average and standard deviation.
Table 2
Fault resistanceWeibull distribution parameters.
Voltage (kV)
()
138
220
33.1156
38.2712
1.4594
1.8406
Fig. 8. Line fault rate in the 138 kV networksimulation results for 80 years of
network operation.
These results are consistent with the pre-dened values. Simulation results reproduce atypical years as happen in real transmission
systems.
4.2.2. Fault type
Fault type distribution was computed separately for the 138 kV
and 230 kV networks, results being shown in Figs. 10 and 11.
Statistical analysis of results is also included. Results show the random nature of fault type distribution, respecting the predened
average value. Fault type distributions with the highest average values present the highest relative values of the standard deviations.
The smaller average value corresponds to LL faults in the 138 kV
Fig. 9. Line fault rate in the 230 kV networksimulation results for 80 years of
network operation.
34
A. dos Santos, M.T.C.d. Barros / Electric Power Systems Research 126 (2015) 2937
Fig. 10. Fault type in the 138 kV networksimulation results for 80 years of network operation.
Fig. 11. Fault type in the 230 kV networksimulation results for 80 years of network operation.
Fig. 12. Fault resistance in the 138 kV networksimulation results for 80 years of
network operation. Comparison with predened Weibull distribution.
Fig. 13. Fault resistance in the 230 kV networksimulation results for 80 years of
network operation. Comparison with predened Weibull distribution.
A. dos Santos, M.T.C.d. Barros / Electric Power Systems Research 126 (2015) 2937
Fig. 14. Short-circuit current distribution (density and cumulative values) in the
138 kV networksimulation results for 80 years of network operation.
35
Fig. 17. Fault time constant distribution (density and cumulative values) in the
230 kV networksimulation results for 80 years of network operation.
Fig. 15. Short-circuit current distribution (density and cumulative values) in the
230 kVsimulation results for 80 years of network operation.
Fig. 16. Fault time constant distribution (density and cumulative values) in the
138 kVsimulation results for 80 years of network operation.
2 kA. Results are in the range 1 kA10.5 kA for the 230 kV network,
showing a typical value around 3 kA.
The cumulative distribution gives important information to
establish circuit breaker and current transformer requirements,
considering a probabilistic approach. For instance, for the 138 kV
network, results show that the maximum interrupted current is
lower than 8 kA, for 98% of the faults. For the 230 kV network this
value is 11.5 kA.
36
A. dos Santos, M.T.C.d. Barros / Electric Power Systems Research 126 (2015) 2937
instance, for the 138 kV network, results show that the maximum
interrupted current is lower than 8 kA, for 98% of the faults. For the
230 kV network this value is 11.5 kA.
The proposed stochastic description of power system faults
applied to short-circuit calculation allows obtaining valuable information that enriches current knowledge of the system planner and
designer, which is usually based on maximum and minimum values. The typical values, its ranges and distribution, obtained with
the proposed model, broaden the overall knowledge the power
system at the planning and design stage, allowing solutions assessment based on risk analysis.
The Edison Electric Institute & Bell System study has published
the mode and average statistic values, Table A1. This information
was used to calculate the scale and shape parameters of a Weibull
distribution that best ts the collected data. Curve tting and the
cumulative distribution of the collected measurements are presented in Fig. A1.
The EDF study published a distribution of fault resistance observations in 60 kV networks, as presented in Table A2. A Weibull
distribution was tted to the data and is presented in Fig. A2. In
this gure, the area of the histogram bars is proportional to the
relative frequency of each bin and its value equals to observation
frequency/bin width.
Appendix A.
References
A study developed by the authors suggests that the random
nature of fault resistance can be described by a Weibull distribution. This study was based on the statistical analysis of collected
measurements, carried out by Edison Electric Institute & Bell System [20] for 220 kV and 132 kV networks and by EDF [21] for the
60 kV networks.
Table A1
Fault resistance after [19].
Network (kV)
Range ()
Mode ()
Average ()
132
220
868
5132
15
25
30
34
Fig. A1. Fault resistance density and cumulative distributionWeibull functions for
the 220 kV (solid) and 132 kV (dash) tted to published experimental data (circle).
Table A2
Distribution of fault resistance occurrences in the 60 kV network after [20].
Obs.
(%)
10
25
]10; 20]
15
]20; 40]
25
]40; 80]
24
]80;200]
8
>200
3
Fig. A2. Histogram of fault resistance observations in the 60 kV and tted Weibull
distribution function.
A. dos Santos, M.T.C.d. Barros / Electric Power Systems Research 126 (2015) 2937
Autotransformers, EHV capacitor banks and shunt reactors, and schemes based
on nonconventional instrument transformers. Was Protection Engineer Advisor
on several transmission level substation projects part of the Portuguese National
Transmission Grid. Secretary of CIGRE SC B5/WG 16: Modern Techniques For
Protecting Busbars In HV Networks and member of CIGRE WG B5.23 Short circuit protection of circuits with mixed conductor technologies in transmission
networks. Currently he is doing his PhD thesis in Instituto Superior Tcnico (IST),
the Engineering School of the University of Lisbon in the subject of Stochastic
Prediction of Voltages Sags Depth and Duration in Electric Power Transmission
Systems.
Professor Maria Teresa Correia de Barros has a long academic career at Instituto
Superior Tcnico (IST), the Engineering School of the University of Lisbon. Together
with teaching and research, she fullled different management positions, namely as
37
Vice President of the Scientic Council of IST and Pro-Rector of the university. From
2001 to 2009 she was working in Industry, at EDP -Energias de Portugal, namely
as Member of the Executive Board of Labelec, the company in charge of Studies,
Development and Testing.
Her main research interests are the elds of High Voltage Engineering, Electromagnetic Transients and Electrical Power Quality. Currently she is Vice-President of
the Power Systems Computation Conference, Chair of International Conference on
Power Systems Transients and Member of the Advisory Board of the International
Journal on Electrical Power & Energy Systems, published by Elsevier Science Ltd.
For more than 25 years she actively contributed to CIGR activities. Professor Correia de Barros become an IEEE Fellow (2000) in recognition of her contribution to
modeling and analysis of power systems transients. Since 2002, she is also a Fellow
of the Portuguese National Engineers Association (Membro Conselheiro da Ordem
dos Engenheiros). She was granted the CIGR Technical Committee Award 2008.