Analysis of Unexpected Fault Current Limiter
Analysis of Unexpected Fault Current Limiter
Analysis of Unexpected Fault Current Limiter
Abstract: The need for high security of electricity supply, Three common methods are used to avoid excessive fault
common to many large industrial process plants, often results in currents, namely: i) split the system into groups of two or three
the interconnection of a number of systems supplied by high transformers (depending on size), ii) replace switchgear with
capacity transformers. This can give rise to prospective fault
higher capacity equipment, and iii) utilise current limiting
currents which exceed existing circuit breaker ratings. One
means of reducing fault currents is to use Fault Current Limiters reactors. These options have disadvantages that in some cases
(FCLs). These devices are designed to separate parts of the are costly to overcome. For example: if a four transformer
network when a fault occurs. system with N-1 capacity is split into two pairs of
Unexpected operation of FCLs has been identified in the 11 kV transformers, each pair will be exposed to an N-1 load that is
distribution network of a mineral processing plant. To investigate 50% greater than the rating of one transformer. Hence splitting
this a full time domain model of the electrical system including the a system that is already loaded to capacity often also requires
FCLs has been developed using ATP. Independent verification of
the purchase of larger transformers to replace the now
the modelling approach was achieved by implementing an
equivalent model using the PSCAD-EMTDC software. Extensive undersized units. Replacement of switchgear also has a cost
studies identified that fault initiation transients caused high di/dt implication, but more often than not replacement requires a
values that, in conjunction with the complicated current prolonged shut down to part of the plant, and the cost of lost
summation tripping logic, resulted in unexpected tripping of one production can greatly exceed the capital cost of new plant.
of the FCLs under certain conditions. The use of current limiting reactors is a common method of
A number of alternative solutions were considered and reducing fault levels. However the voltage drop across the
simulated to determine if satisfactory operation could be
reactor can often result in unacceptable voltage levels under
expected. The chosen solution involved a network
reconfiguration, the relocation of two reactors, and the certain operating configurations, for example operation with a
installation of three additional FCLs. This enabled simpler transformer out of service where that transformer’s load is
tripping criteria to be adopted, which prevents the fault initiation supplied through a reactor.
transients from causing spurious trips. A less commonly used alternative method of reducing fault
The study demonstrates the benefits of using powerful levels is to employ FCLs.
simulation tools such as EMTP/ATP and PSCAD-EMTDC for
analysing complex power systems and identifying the causes of
II. FAULT CURRENT LIMITER OPERATION
problems that are generally impractical to find by normal
measurements due to the need to capture information during Fault current limiters are devices that are designed to
faults. separate parts of the network when a fault occurs. They are fast
operating devices that commence operation within the first few
Keywords: Fault current limiter (FCL), current limiting fuse, milliseconds after a fault, and limit the first peak of the fault
ATP, EMTP, PSCAD-EMTDC, fault level studies, transients, current to acceptable levels.
fuse modelling. Fault current limiters consist of current limiting fuses
I. INTRODUCTION operated in parallel with explosible links. The explosible link
may be constructed from a brass tube fitted with an explosive
Many large industrial processes have evolved from smaller
charge. The tube is appropriately fissured to control its
establishments via staged expansions.
bursting performance, and is employed to carry the normal
The distribution networks required to provide electricity to
load current of the device. The construction and operating
these plants are usually expanded by the addition of extra
characteristics of this type of FCL are described in [1]. An
transformers. This gives rise to prospective fault currents that
alternative form of explosible link consists of a segmented
can exceed circuit breaker fault current ratings. In industrial
applications it is usually the first peak of the fault current that copper bar, again appropriately fissured, which separates into
presents the most onerous condition due to the contribution a number of segments, again by the triggering of an explosive
from many connected motors. charge. This type of FCL is illustrated in [2].
On detection of a fault meeting the tripping criteria, the
M. S. Hibbert is with Connell Wagner Energy Division, L1, 433 Boundary St,
charge is detonated separating the explosible link, thus leaving
Spring Hill, Qld 4004, Australia (hibbertm@conwag.com) only the current limiting fuse in the circuit. The fuse then
Presented at the International Conference on Power System operates separating the two parts of the network, effectively
Transients (IPST05) in Montreal Canada on June 19-23, 2005 limiting the first peak of the fault current to that generated by
Paper No. IPST 05-038 the source on the faulted part of the network plus the let
through current of the fuse. The principle of operation of corroborates this theory, and hence a time domain model of the
current limiting fuses is well known and will not be expanded system was required to analyse the system in the first half
on here. It is understood, however, that the detailed design cycle after fault initiation. References [3] and [4] provided
aspects of current limiting fuses for this application are some useful information on modelling of current limiting
somewhat different to those used for typical applications. In fuses, and after reviewing these it was decided to model the
particular, for this application the rating of the fuse element is FCL using the Alternative Transients Program (ATP) software.
significantly smaller than the normal load current expected to
be carried by the FCL, thus ensuring rapid operation. TABLE I CONFIGURATION OF FAULT CURRENT LIMITERS
Faulted Bus Fault Current Limiter Operation
Bus Name/Rating FCL 1 FCL 2 FCL 3
III. APPLICATION OF FAULT CURRENT LIMITERS
Bus A 26 kA No No Yes
Fault current limiters were used in an industrial plant to Bus B 26 kA Yes No No
manage fault levels after an expansion added two new Bus C 40 kA No No Yes
Bus D 26 kA No Yes No
47 MVA 132/11 kV transformers to an existing four
Bus E 40 kA No No Yes
transformer system as shown in Fig.1. The addition of extra Bus F 40 kA No No Yes
reactors to the system, which already had one reactor Bus G 40 kA No No Yes
connecting each transformer to a common bus, was not Bus H 40 kA Yes Yes Yes
practical due to the size of reactors that would be required, and Bus I 40 kA Yes Yes Yes
the resulting voltage drops under transformer out conditions. Bus J 26 kA No No Yes
Bus K 26 kA No No Yes
Instead, two reactors were replaced with FCLs and a third FCL Bus L 26 kA No No Yes
was installed in the tie cable to the new switchboard (busbars
E, F and G).
A. Basic FCL Model
T1 T2 T3 T4 T5 T6
The explosible link was modelled using a type 12 switch
(refer to [5] for details) which is controlled by a logical
bus A bus B bus C bus D bus F variable, and allows current chopping at any point in the
bus E bus G waveform. The current limiting fuse element was modelled
using a type 91 piecewise linear time varying resistor with a
FCL1 FCL2 FCL3
non-linear characteristic, which started its characteristic after
the voltage across it exceeded a nominated value. As it was
Legend
Circuit Breaker
known that the design of the current limiting fuse element of
bus H bus I the FCL was different to that of a standard current limiting
Fault Current Limiter
fuse, it was decided that the characteristic should be derived
from test data of these fuse elements. Oscillogram traces from
Current Limiting Reactor
[6] were studied and transposed to a spreadsheet where the
bus J bus K bus L Transformer
non-linear characteristic was derived by simply dividing the
voltage characteristic by the current characteristic. The
G1 G2 Generator
resulting characteristic was similar to that shown in [4],
although more approximate in the first two milliseconds, and is
Fig. 1 Application of Fault Current Limiters given below in Fig. 2, with the characteristic values following
in Table II.
Two 12.5 MW steam turbine driven 11 kV generators also
supply the system as shown, and fault current contribution
100
from approximately 90 MW of connected motor load also 90
added to the high prospective fault levels. 80
Resistance (Ohms)
70
60
IV. SPURIOUS OPERATIONS
50
The FCLs were configured to operate as shown in Table I. 40
Since commissioning of the FCLs incorrect operation for faults 30
20
on bus B and bus D had been experienced. In both cases 10
FCL 3 operated but was not required to operate. The cause of 0
these spurious operations was investigated. 0 1 2 3 4 5 6 7 8
Time (milliseconds)
V. MODELLING APPROACH
The postulated cause of the spurious operations was Fig. 2 Non-Linear Resistance Characteristic of Fault Current Limiter
transients interfering with the di/dt sensing. Review of [2] Fuse Element
TABLE II NON-LINEAR RESISTANCE CHARACTERISTIC R2
Time Resistance SW 2
Milliseconds Ohms
0 0.0025
R S2
0.5 2.94
1.0 4.71 R4 SM
1.5 7.06
SW 1
2.0 8.24
2.5 8.82
S1 R3
3.0 8.82
4.0 8.82 C2
4.5 9.01
5.0 11.18
R1 R1
5.5 13.53
6.0 25.00 C1 C1
7.0 125.00
8.0 250.00
Legend
8.5 375.00
R1 5 Ohm damping resistors
9.0 500.00 C1 4 nF damping capacitors
10.0 1,000.00 R2 10,000 Ohm damping resistor
C2 0.6 µF spike damping capacitor
A simple single phase network model was established to after transition through current zero
test the proposed FCL model, which in its simple form is R3 & R4 0.0001 Ohm resistors used for separation of switches
shown below in Fig. 3. SM permanently closed measuring switch
SM
4000
Current (Amps)
3500
-S42F (Volts)
(V)
10000 Voltage
-S42F
3000
S42G Voltage
2500
S
(Mag)
2000
5000 Current
Legend 1500
1000
Non-Linear Time Varying 500
R Resistance 0 0
Current Chopping Switch 50 5
10 15
10
TimeTime (ms)
(milliseconds)
Logical Signal controlling Switch
S opening
Fig. 5 Fault Current Limiter Simulation Result
B. Tripping Criteria
Fig. 3 Initial Fault Current Limiter Model The MODELS subprogram was used to simulate the
tripping criteria which were obtained from [7]. It should be
While this model worked, it suffered two drawbacks. Firstly noted that reference [7] is a confidential report to the client
the operation of the current chopping switch gave rise to a and is not available to the public, however similar information
significant voltage spike and subsequent numerical can be found among the references given in [1]. The tripping
oscillations. Secondly the non-linear resistor would allow criteria were fairly complex, and are given below for FCL 3:
current flow to continue past the current zero for certain i) the instantaneous value of the current through FCL must be
switching scenarios. Modifications were made to overcome between nominated high and low gateway settings
these problems and damp out the oscillations, resulting in the ii) the absolute value of the time derivative of the current
final model shown in Fig. 4. Switch SW1 represents the through the FCL must be above the nominated di/dt trip
explosible link and is triggered as soon as the tripping criteria setting for the FCL
are met. Switch SW2 is used to interrupt the current through iii) the absolute value of the time derivative of the sum of the
the time varying non-linear resistor as soon as it crosses the currents through all three FCLs must be above the
zero axis. This model resulted in acceptable results producing nominated di/dt trip setting for the summation value
characteristic voltage and current waveforms similar to those iv) all three of the above values must possess the same sign.
presented in [3], as shown below in Fig. 5. The satisfaction of the above four criteria will result in an
operation of the FCL. This was detected using logical PSCAD. Unlike the ATP model there was no need to use
comparison statements in MODELS, which then triggered the additional damping resistors in the PSCAD model. Numerical
operation of the current chopping switch SW1 in Fig. 4. The chatter is a time step to time step, symmetrical oscillation
piecewise non-linear time varying resistance characteristic phenomenon inherent in the trapezoidal integration method
commences as soon as the voltage across it reached 10 V used in the Dommel algorithm for transient simulation of
(which was an arbitrary choice), and the operation of the FCL electrical networks. It is usually initiated by the closing of a
concluded when switch SW2 in Fig. 4 opens following switch in a branch containing inductors. It does not matter if
transition of the current through a current zero. This transition the switching occurs at a natural current zero, or elsewhere in
was detected using simple logic in MODELS followed by the waveform. The PSCAD software includes a chatter
enabling of the opening signal S2 (refer Fig. 4). detection algorithm to continuously detect such spurious
Numerical oscillations, although dampened by the resistors oscillations and remove them if so required. It is believed this
and capacitors R1, R2, C1 and C2, as shown in Fig. 4, were still is the reason why damping resistors were not required in the
a feature of the output waveforms. These had the potential to PSCAD model. Chatter and its effects are discussed in Ref
interfere with the correct simulation of tripping conditions. [8,9]. The PSCAD representation of the FCL confirmed the
This aspect was dealt with by averaging the output every two transient voltage and current waveforms predicted by the ATP
time steps, and then operating on the averaged results in the model (as shown in Fig. 5).
MODELS logic to determine if the tripping criteria had been The FCL tripping logic implemented in the PSCAD
satisfied. MODELS was found to be a very powerful tool for Continuous System Modelling Function Library was identical
comparing network values and their time derivatives and to that programmed using the ATP MODELS subprogram.
initiating switch opening sequences based on the results of The electrical network elements were modelled using the
those comparisons. standard components available in the PSCAD library. The
synchronous and induction machines were modelled using the
C. Network Model
PSCAD two axis models to give the correct transient fault
The network was modelled using standard ATP currents.
components. The upstream 132 kV system was modelled as an For a number of test cases the PSCAD model was shown to
infinite bus and three phase RLC source impedance connected behave identically to the ATP model. The asymmetric peak
to the local transmission substation via a number of overhead currents predicted by each model and the trip/no trip decisions
transmission lines. These transmission lines were represented reached by their tripping logic were the same. As both models
as coupled Pi models with the matrices derived using ATP’s were based on the same source data and developed
“line constants” routine. The six star-delta 132/11 kV independently of each other, this was taken to demonstrate the
transformers were represented using the “BCTRAN” model, veracity of the programming of both models. The full system
with the earthing transformers modelled using “XFORMER” studies were then undertaken using the ATP model.
(refer to [5] for details).
The generators were modelled using the Universal Machine VI. RESULTS OF SYSTEM STUDIES
Model UM1, assuming constant excitation and input torque,
Simulations showed that spurious operation of FCL 3 could
which is an appropriate assumption considering that operation
be expected for faults on bus B or D. Different phases of
over only a few cycles needs to be simulated.
FCL 3 were affected depending on the point of wave at which
All other connections in the network were by single core
the fault occurred. The following combinations were found to
11 kV cables which were represented as coupled Pi models
occur i) one phase only, A, B or C, ii) two phases, A and C
with the matrices derived using ATP’s “cable constants”
only, and iii) under certain conditions correct operation (ie no
routine. Motor contribution to the fault current was simulated
operation of FCL 3) was seen to occur.
by lumped motor models using the Universal Machine Model
The cause of the spurious operation was traced to the
UM4 connected at 11 kV or at 3.3 kV and 415 V via lumped
existence of 20 kHz (approximately) current oscillations
equivalent transformer (BCTRAN) models. The one peculiar
superimposed on the fault current waveform, commencing
feature about the plant that is worth mentioning is the fact that
immediately after the initiation of a fault and lasting several
the cables from the transformer terminals T1 through T4 to
milliseconds. An example of these fault initiation transients for
busbars A through D respectively were approximately 200 m
one phase is shown in Fig. 6. It was clearly shown that these
in length, and the tie cable between bus F and bus I was
transients resulted in excessively high values of di/dt which
approximately 500 m in length. All other cable runs in the
were responsible for “confusing” the tripping logic.
network shown in Fig. 1 were relatively short.
The presence of these current oscillations at this high
D. Model Verification frequency gives rise to very high values of di/dt. This defeated
The FCL model was verified by the development of an the “same sign” criteria of the tripping logic as the time
alternative model that was implemented using the PSCAD- derivative of the current through FCL 3 and the time derivative
EMTDC software. The basic FCL representation of a switch of the sum of the currents through all three FCLs oscillated
in parallel with a variable resistance was implemented within across both sides of the zero axis, and at one stage were of the
same sign while all other criteria were satisfied, thus resulting Two basic tripping configurations are required. One is
in an unnecessary operation. shown in Fig. 8 below and applies to bus pairs A, D and B, C,
both of which can be connected to the generator buses J, K and
12000
L. At the faulted bus shown in Fig. 8 the current through
FCL 1 (I2) is positive. Hence the summation current IS = I1 + I2
10000
is greater than the summation current at the adjacent bus,
where I5 is negative and equal to I4 + I6. This factor was used
8000
to provide an additional tripping criteria similar to a bus zone
Fault Current (Amps)
to have less reliance on di/dt criteria and more reliance on bus E FCL5 FCL6 bus G
instantaneous current criteria to avoid the problems caused by FCL1 FCL2 FCL3 FCL4
numerous fused isolators in the network. A fault on the load bus J bus K bus L
Generator
side of these fused isolators, which were equipped with
G1 G2
standard current limiting fuses, would result in the
commencement of fuse melting before the detection time for
FCL operation had commenced, and it was not possible to Fig. 7 Reconfigured System
guarantee that the let through current of the FCL would not
combine with the fault current from the source on the faulted If the above are satisfied, then operation of the FCL is
side to exceed the 50 kA limitation of the downstream fuses, required to occur. Without the last criterion, as I2 and I5 in
unless the prospective fault current was kept below 50 kA Fig. 8 are the same, transient oscillations can result in the first
RMS. Simplification of the tripping criteria required each FCL four criteria being satisfied for both FCLs. The addition of the
to operate and protect one bus only, and the 50 kA limit on fifth criteria prevents this due to the fact that I5 = -(I4 + I6), and
prospective fault current required the network to be split into hence the sum of I4 and I5 will not reach the necessary
three pairs of transformers. The four aged transformers vary in instantaneous value to enable a trip to occur. The second
size from 28 to 40 MVA, and are all to be replaced with new tripping configuration applies to buses E, F and G and is
47 MVA units to enable N-1 capacity to be maintained. shown below in Fig. 9. Fault current limiters FCL 5 and FCL 6
An additional six-unit switchboard is to be installed and have similar tripping criteria to that discussed above for faults
some reconfigurations are planned to enable different pairs of on bus E or bus G. For faults on bus F, an additional current
transformers to be connected together for increased security summation tripping criteria IS = – I2 – I5 + I8, together with
during transformer outages. The resultant system is shown in similar di/dt and instantaneous current tripping criteria, is
Fig. 7. implemented to trip FCL 6.
IX. REFERENCES
[1] A.M. Dán, Zs. Czira, L. Prikler; “Comparison of Traditional and
Thyristor-Controlled Fault Current Limiters for Medium Voltage
Application”, Proc. of IPST 1999, Budapest, June 20-24, 1999, pp. 602-
607.
I1 I4 [2] J. C. Das, “Limitations of Fault-Current Limiters for Expansion of
Electrical Distribution Systems”, IEEE Transactions on Industry
Applications, Vol. 33, No. 4, July/August 1997, pp 1073 – 1082.
IF [3] A. Petit, G. St-Jean, and G. Fecteau, “Empirical Model of a Current-
I3 I6
Fault Limiting Fuse using EMTP”, IEEE Transactions on Power Delivery,
M M
Vol. 4, No. 1, January 1989, pp 335 – 341.
I5 [4] Lj. A. Kojovic, S. P. Hassler, K. L. Leix, C. W. Williams, E. E. Baker,
I2
FCL1 FCL2 “Comparative Analysis of Expulsion and Current-Limiting Fuse
Operation in Distribution Systems for Improved Power Quality and
Protection.” IEEE Transactions on Power Delivery, Vol. 13 No. 3, July
1998, pp 863 – 869.
[5] Canadian-American EMTP Users Group, Alternative Transients
Program Rule Book, 1998.
[6] Fuse Breaking Test Versuchsbericht N r 2667_74 N V KEMA Test
Laboratories
Legend [7] Hartung, Kientoff, Grafe, “Application of Is-Limiter – Description and
I1 Transformer Incomer contribution Calculation”, ABB Calor Emag., Ratingen, Germany, Tech. Rep.
I2 Current through FCL DECMS Order No. 02/7017740, June 2002.
I3 Motor Contribution on faulted bus [8] P. Kuffel, K. Kent, G. Irwin, “The Implementation and Effectiveness of
I4 Adjacent Transformer Incomer Contribution Linear Interpolation Within Digital Simulation”, Proceedings,
I5 Current through adjacent FCL International Conference on Power Systems Transients (IPST '95), pp.
I6 Motor Contribution on adjacent bus 499-504, Lisbon, September 3-7, 1995.
[9] A. M. Gole, I. T. Fernando, G. D. Irwin, O. B. Nayak, “Modeling of
Positive direction of current is towards protected bus Power Electronic Apparatus: Additional Interpolation Issues”,
International Conference on Power Systems Transients (IPST '97), pp.
Fig. 8 Tripping Configuration for Bus Pairs A, D and B, C 23-28, Seattle, June 22-26, 1997.
X. BIOGRAPHIES