See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/315481621

Modelling of Current Transformer Saturation for Power System Protection

Applications

Conference Paper · March 2013

CITATIONS READS

0 1,466

1 author:

Michael Stanbury
Ausgrid
3 PUBLICATIONS   60 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

CT Saturation in Power System Protection View project

All content following this page was uploaded by Michael Stanbury on 06 April 2017.

The user has requested enhancement of the downloaded file.

 Modelling of Current Transformer Saturation for Power
System Protection Applications
Michael Stanbury Ausgrid [email protected]

Keywords: Current transformer, saturation, protection relay, nuisance trip, hysteresis, statistical model

Contents
1 Abstract 1
2 Introduction 1
3 CT Modelling 2
3.1 Ideal Saturation Model 2
3.2 Static Tanh Model 2
3.3 Static Preisach Hybrid Model 3
3.4 Model Comparison 4
3.5 Theoretical Analysis of CT Error 5
4 Relay Tests with Simulated Saturation Waveforms 7
5 Statistical Model of CT Saturation 9
5.1 CT Flux Density During a Fault 9
5.2 Statistical Model 10
5.3 Verification 12
6 Conclusion 12
7 References 13
List of Figures 13
List of Tables 14
8 Biography: Michael Stanbury 14
9 Appendix 14

1 Abstract
This work investigates the effect of current transformer (CT) saturation on protection schemes. Ausgrid and
other utilities would like to reduce the size of their CTs to save space and cost, but this may increase the
likelihood of saturation. Theoretical work is presented which shows that for large fault currents, both simple
and complex CT models give similar results, so simple models can safely be used. It is shown that it's not
necessary to model hysteresis in protection applications where fault levels are thousands of amps.
Secondly, simulated waveforms of CT saturation were injected into a transformer differential protection relay
to assess the relay's ability to detect and handle saturation. Results show that modern relays can handle
significant saturation, however, surprising outcomes were discovered. For example, the saturated
waveforms can cause the second harmonic blocking function to operate. This feature is designed to prevent
nuisance trips for transformer inrush, but was found to also operate during CT saturation. The learning is
that there can be significant non-obvious interactions between relay functions when saturation occurs.
Lastly, a statistical model is presented which predicts the probability of saturation for a given fault level. A
tool was developed that allows protection engineers to interactively adjust burden resistance, cross sectional
area and fault levels to see how the probability of saturation changes. This allows easy quantification of the
risk associated with CT saturation. Until now, this risk has only been discussed in vague, qualitative terms.
This statistical model will allow engineers to attach a probability of saturation to various situations, making it
possible to directly compare different options.

2 Introduction
Ausgrid and Endeavour Energy are currently looking at reducing the size of their protection current
transformers (CTs) to save cost and space. However, this may increase the risk of CT saturation during
faults. If a CT saturates, then the protection relay will not receive an accurate measurement of the fault
current. This may mean that it fails to operate, or it may operate when it should not (i.e. a nuisance trip) in
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

unit protection schemes. A nuisance trip may cause a large number of customers to experience an
unnecessary power outage.
There are many factors which contribute to CT saturation including fault duration, primary system X/R ratio,
the time of the fault on the voltage waveform, and the remanence in the core at the time of the fault.
These issues are complicated by the fact that modern protection relays have the ability to detect CT
saturation and may compensate for it, or restrain the trip threshold. Each relay manufacturer uses a
different algorithm to detect and handle saturation, so there are no general models to predict what will
occur in this situation. This makes it particularly difficult to predict the effect of CT saturation on a power
system.

3 CT Modelling
In order to investigate the effect of CT saturation on protection schemes, three CT models were investigated
and implemented. The first is a simple, linear piecewise model called the Ideal Saturation Model. Secondly,
the Static Tanh Model introduces a gentle roll off into saturation. Lastly, the Static Preisach Hybrid Model
reproduces the B-H curve accurately with hysteresis and minor loop phenomena. These three models were
compared to find which is the best for protection applications.
3.1 Ideal Saturation Model
This model assumes the CT is ideal until the flux density reaches the saturation level, at which point it is
capped at that value. For Grain Oriented Silicon Steel (GOSS) CTs, this is about 1.7 Tesla. At this point, the
secondary voltage and current instantly drop to zero, until the flux density decreases again. The B-H curve
of the Ideal Saturation Model is shown in Figure 1.

Figure 1 The B-H curve for the Ideal Saturation Model. The CT is ideal in the vertical section,
and the secondary voltage is zero in the horizontal sections. Saturation occurs at ±1.7 Tesla.
This model does not include any hysteresis, and it saturates instantly, rather than easing into saturation.
Although this model is very simple, it produces fairly accurate results for the secondary current of protection
CTs during large fault currents. This implementation has the flexibility to incorporate burdens with both
resistive and inductive elements.
3.2 Static Tanh Model
The second model is called the Static Tanh Model. This is more advanced than the last in that it has a
gentle roll-off into saturation. Like the previous model, it does not have hysteresis. This can be thought of
as an anhysteretic model of the real B-H curve. The characteristic is shown in Figure 2.
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

Figure 2 B-H curve of the Static Tanh Model.

In this model, the B-H curve is described by the following formula where is a function of :

where:
is the hyperbolic tangent function
is a parameter describing the slope of the curve

The tanh function was chosen because it is trivial to differentiate. This allows one to conveniently simulate
the CT using well-established techniques. For this work, a system of differential equations was developed
and these were solved numerically using Matlab’s Ordinary Differential Equation (ODE) solvers. The state
variables in the system of differential equations were the secondary current , the magnetic field strength
, and the core flux density .

3.3 Static Preisach Hybrid Model

The last model is the Static Preisach Hybrid (SPH) model. This uses Preisach’s theory of hysteresis to
implement the B-H behaviour of the core. The parameters of this model are calculated from B-H curve
measurements on a core made of the desired material. In this work a core of Grain Oriented Silicon Steel
(GOSS) was used. This model, unlike the previous two, faithfully recreates hysteresis and minor loop
phenomena in the B-H Plot. It is the most complex, and the most accurate of the three models presented.
Preisach’s theory models hysteresis as a large number of ideal relays on the B-H plane. Each of these relays
has different switching values. A weighted sum is performed on all of these elements, which produces the
final B-H curve. The weights for this sum are derived from experimental measurements of a GOSS core. An
example B-H curve is shown in Figure 3.
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

Figure 3 The SPH model in red shows an initially demagnetised core undergoing sinusoidal
excitation with an open circuit secondary. The curve starts at B=0, H=0 and moves up the
initial trajectory. After reaching saturation it then moves into the steady state major loop of
the B-H curve. The blue curve was experimentally determined in the lab at low frequency (<
1Hz). There is close agreement between the SPH model and the experimental results.

Of the three models presented above, the Ideal Saturation is the simplest and least accurate, and the SPH
model is the most complex and most accurate. The Static Tanh Model lies between them. The models were
then compared in order to ascertain the ideal compromise between complexity and accuracy for protection
applications.

3.4 Model Comparison

The three models presented above were compared to each other in a variety of fault simulations. It was
found that each model produces almost exactly the same results for the types of large fault currents seen on
an 11kV power system. For small currents of a few amps, the more complex models are more accurate, but
when the current is thousands of amps, the three models perform almost identically. An example is shown
in Figure 4.
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

Figure 4 Secondary currents versus time for the three models presented above, due to a 5kA
fault. All three curves are on top of each other so the results are practically identical. The
dotted line represents an ideal CT which does not saturate.
Given that the three models all produce identical results for large fault currents, it is recommended that the
simple Ideal Saturation Model be used for fault simulation on the 11kV network as it is the easiest to
implement and use, and it produces the same results as more complex models. In the next section we will
provide theoretical reasons for why all three models produce the same result at large fault currents.

3.5 Theoretical Analysis of CT Error

We now derive an equation which relates the accuracy of a CT model to the size of the fault. We begin by
forming a relationship between the magnetic field strength ( ) and the instantaneous CT error ( ). This will
allow us to explain why the three models give the same results. It will be shown that all three models
presented earlier will be within 1% of each other for the fault levels which are typically seen on 11kV power
systems.
The following derivation will allow us to quantify the accuracy of the Ideal Saturation model. If the Ideal
Saturation model is accurate in a particular situation, then the other two more complex models will be even
more accurate. In this way, by verifying the Ideal Saturation model to a certain level of accuracy, we are
also verifying the other models to be better than that level of accuracy.
We begin by defining the instantaneous CT error at time , which is the deviation from ideal behaviour.
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

Where and are the instantaneous primary and secondary currents at time and are the
number of primary and secondary turns on the CT ( is usually one). A typical value of may be 0.01, to
represent a 1% error. Now we use the standard form of the magnetic field strength , in the core:

Where is the mean core length. Combining the previous two equations:

From equation (3), it can be seen that is proportional to the instantaneous CT error for a given
instantaneous primary current. is also proportional to the magnetising current. We will now find a
maximum value of for which the error is suitably small. When not in saturation, CT error is largest when
the primary current is large, so we take the worst case where the primary current is at the peak value,
. Taking maximum values of equation (3):

Equation (4) is very useful for planning lab work on CTs as it helps to predict what sort of CT error will be
encountered. More importantly, this equation relates maximum values to maximum CT error. It shows us
that for large fault currents seen in power networks, it doesn’t matter whether we accurately model the
shape of the B-H curve, because various curve shapes result in practically the same secondary current. All
that matters is that the non-saturated portions of the B-H curve fit within the range to . If this
is the case, the CT error will be less than at all times, when not saturated. This means that for network
fault applications, our three models will provide the same results, despite having very different B-H curves.
We now present an example of equation (4). In this example, , , and ,
and the chosen maximum error is . Using equation (4), . This value is very
useful. As long as the non-saturated portion of the B-H curve lies within the limits to , then all
models will give a secondary current which agrees to within 1%. The range to is shown as a
grey box in Figure 5. The B-H curve of GOSS cores fit comfortably within these limits so hysteresis and
saturation roll-off can safely be ignored in our model. In other words, the Ideal Saturation model is good
enough, and the other two complex models add nothing of value for protection applications. In Figure 5,
the range to is shown as a grey box, and we can see that the B-H curve fits within it.

Figure 5 Dark and light grey curves show experimentally determined B-H characteristic of a
GOSS core at 0.5 Hz and 50 Hz respectively. The grey box runs from -186 to +186 A/m. The
experimental curves sit comfortably within the grey box, so all three models presented above
will agree to within 1%.
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

Given that all three models produce the same results for large fault currents, we can safely conclude that it’s
not important to model hysteresis when predicting the behaviour of GOSS protection CTs under large fault
currents. This analysis shows that GOSS has low enough hysteresis to be ignored for protection
applications. For these reasons, it is recommended that the Ideal Saturation model be used as it is far
easier to implement than the other two. Note that neglecting hysteresis may not be valid if a gapped core
CT is used because the B-H curve would be considerably wider and may extend outside of the range
to . In this case, equation (4) is still valid, but will naturally be larger.

4 Relay Tests with Simulated Saturation Waveforms

Once the SPH model was developed, it was used to simulate several faults on Ausgrid’s network. A
transformer differential protection scheme was modelled. This is Ausgrid’s most difficult case because CT
saturation on a through-fault could cause a nuisance trip. The simulated faults were set up to mimic a
through-fault on Transformer 4 at Ausgrid’s City South Zone Substation in Sydney. This is a 132kV to 11kV,
YNd1d1 transformer. Four fault types were generated: three phase, two phase, single phase to ground and
two phase to ground. For each of these fault types, three different CT sizes were used on the 11kV side
(large, medium, small). This produced 12 tests in total, and each of these were injected into an SEL-387E
differential protection relay which is currently used at City South Zone Substation. The results are
summarised in Table 1.

Differential
Fault Type, Core Degree of
Trip? Element 2nd Harmonic Blocking?
Size Saturation
Picked Up?
LLL, Large None No No Negligible
LL, Large None No No Negligible
LG, Large None No No No
LLG, Large Slight No No Yes
LLL, Medium Moderate No Briefly Yes
LL, Medium Moderate No Yes Yes
LG, Medium High No Yes Yes
LLG, Medium Moderate No Yes Yes
LLL, Small High Almost Yes Yes
LL, Small High Yes Yes Yes, but not at all times
LG, Small Extreme Yes Yes Yes, but not at all times
LLG, Small High Almost Yes Yes, but not at all times

Table 1 Summarised results of 12 though-fault simulation tests in a transformer differential

protection scheme. The large and medium cores were not susceptible to nuisance trips but the
small core had two nuisance trips, and came close to tripping for the other two.

One of the 12 test results is shown in Figure 6. It can be seen that there is considerable saturation on the
low side CTs (see centre plot), yet the relay did not initiate a trip. The relay picked up a large mismatch
between high side and low side currents so the 87R logic bits were enabled. The relay did not trip,
however, because a large second harmonic was detected and this initiated the second harmonic blocking
feature. Saturated CT waveforms have a large second harmonic component. The second harmonic blocking
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

feature is designed to prevent trips on transformer inrush currents, but we can see that it is also activated
by CT saturation, as an unintended side-effect.

Figure 6 Simulated LLG through-fault on a 132 to 11kV transformer differential scheme. The
upper plot shows the high side currents, the centre plot shows the low side currents. The lower
plot shows various logic states within the differential protection relay. Saturation can be seen
on the low side about 30ms after the fault occurs.
Given that these 12 simulated tests are through-faults, we do not want the protection to trip. In one sense,
the second harmonic blocking is useful as it prevented a nuisance trip in 6 out of the 8 tests where the
differential element (87R) operated. However, this behaviour does not seem intended by the relay
manufacturer and came as a surprise to Ausgrid protection engineers.
All 12 tests showed the same second harmonic blocking behaviour, however on 2 of the tests with the
smallest core, the saturation was so severe that the second harmonic blocking was only initiated for part of
the fault duration so the relay sent a trip signal in one of the gaps. This tells us that the smallest core
tested is certainly too small to be used on the network, but the medium or large cores could possibly work
with appropriate protection settings.
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

After these tests were performed, they were modified to produce 12 new tests. The low side currents were
inverted to simulate 12 in-zone faults on the transformer. All 12 of these tests produced a trip, as desired.
These tests have raised some potential issues with CT saturation handling of relays. If CT sizes are reduced
to the point where saturation becomes more common, then protection engineers will need to be aware of
potentially unintended consequences related to second harmonic blocking. Ausgrid plans to run these same
tests on other relay models and other protection schemes.

5 Statistical Model of CT Saturation

Within electrical utilities, the risk of CT saturation has typically been discussed in vague terms. For Ausgrid
to reduce the size of its CTs, a more rigorous understanding of saturation risk is required. This will ensure
that a new, smaller CT specification is created with an informed understanding of how the risk of saturation
may change. In this section we derive a statistical model of CT saturation which estimates the probability
that a CT will saturate for a given fault. This will enable network utilities to quantify the risk associated with
saturation.
5.1 CT Flux Density During a Fault
To discover whether or not a CT saturates during a fault, we must find out whether the flux density ( ) rises
to the saturation level . We assume the CT follows the Ideal Saturation model presented in section 3.1,
and first described in reference [2]. This simple model does not include hysteresis, but this is acceptable
because hysteresis does not have a large effect on the secondary current as shown in section 3.5.
A CT with a resistive burden undergoing a network fault will have a flux density waveform described by
equation (5). A full derivation of (5) is presented in the Appendix.

Where and are the number of primary and secondary turns, is the RMS fault current at steady state,
is the secondary resistance, is the core’s cross sectional area, is the power system frequency, is
time, is the X on R ratio of the primary system fault, , is the angle of the sinusoidal voltage
waveform when the fault occurs (which dictates DC offset), and is the core remanence prior to the
fault. and are major contributors to core saturation of power system CTs [4]. An example of
equation (5) is shown in the lower plot of Figure 7.
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

Figure 7 A typical fault current waveform is shown at top with full DC offset. Below is the
corresponding flux density in the CT core, calculated with (5). The remnant flux ( ) is
assumed to be zero. The saturation flux density is shown as a dotted line at 1.7 T. In this
example, the core does not saturate because the flux does not reach 1.7 T.
We will now separate (5) into an exponential term and a sinusoidal term. We will call the exponential term
because it follows the midpoint of the sinusoidal oscillations in (5). It is found by setting the first
cosine term in (5) to zero:

The amplitude of the sinusoid is the constant value:

will quickly rise to its final value, which is found when . The final value is:

Using (8) we can find out whether or not a CT saturates by checking if is outside these limits:

where is the saturation flux density of the core (~1.7 T for GOSS). If from (8) is found to be
within the limits of (9), there will be no saturation. Otherwise the CT will saturate.
5.2 Statistical Model
We now develop a statistical model of CT saturation. Equation (8) contains two independent, random
variables from a utility’s perspective ( and ). Each of these two random variables has some associated
Probability Density Function (PDF). We will derive a PDF for , based on the PDFs of and ,
using the Method of Transformations [5].
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

Typical values of are such that the DC offset will generally be less than 70% [6], however we will
conservatively assume a uniform distribution for which can take any value between and . We also
assume a uniform distribution for , which can take any value in this range:

Where is the “remanence factor” which can be as high as 0.8 in electrical steels. We now transform the
PDFs of and to find the derived PDF of , according to [7]. This produces the PDF of the final flux
density after a fault:

where:

is the unit step (or Heaviside) function.

The function has in radians and is defined to be zero where is outside of the
range -1 to 1.

Figure 8 is a graph of (11). If we integrate (11) over the range in (9) we find the probability that the CT will
not saturate. This integral was performed by the Mathematica software, but a wide variety of methods can
be used.

Figure 8 PDF of the ideal CT flux density after a fault, calculated by (11). The non-saturated
region is shown in the middle (lighter area), based on the range shown in (9). The two outer
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

(darker) regions represent flux densities that the CT will try to reach, but will fail due to
saturation. The integral of the outer areas represents the probability that any saturation will
occur (Probability = 49.2% for this example).
Equation (11) is very useful for protection engineers who are concerned about the possibility of saturation in
a CT specification. If it is desired to make a CT smaller, (11) provides a way to understand how the risk of
saturation may change. For example, this statistical model tells us that we can compensate for a lower
cross sectional area with more turns on the CT. This would specify a smaller CT with the same saturation
risk. By comparing diagrams like Figure 8, engineers can easily visualise how various parameters affect
saturation risk including the saturation flux density, number of turns, remanence factor, fault level, X/R ratio,
burden resistance, cross sectional area and power system frequency. This new understanding will help
protection engineers to specify CTs with a more rigorous understanding of the risk that smaller CTs could
bring to the network.
5.3 Verification
The derivation of the final PDF in (11) was verified using the Monte Carlo method. A frequency histogram
was produced of the final flux densities of 100,000 simulated faults, and this distribution was seen to
accurately follow (11). The 100,000 faults had randomly chosen values of and , according to the
uniform distributions described above. An example of one of these verification tests is shown in Figure 9.

Figure 9 Monte Carlo verification of the Statistical Model of Saturation. On the left is the
frequency histogram of the final flux density of 100,000 randomly chosen fault scenarios
simulated with equation (5). On the right is the PDF plotted using equation (11). Close
agreement between the two methods is seen.

6 Conclusion
This work has contributed to a thorough understanding of CT saturation which can be used in protection
applications. A rigorous understanding is important for utilities wishing to reduce the size of their CTs.
Three models were introduced to model saturation and these were seen to all produce identical results for
large faults, so we can safely ignore hysteresis for these applications. We can also ignore the gentle roll-off
into saturation, and therefore the Ideal Saturation Model is seen to perform as well as the more complex
models.
Results were also presented from injection tests of saturated waveforms into a transformer differential
protection relay. It was seen that the saturated waveforms initiated the second harmonic blocking feature
of the relay, preventing a nuisance trip, but also greatly reducing the sensitivity of the relay during that time.
If CT sizes are to be reduced, this is an important factor to consider.
Lastly, a statistical model of saturation was presented which allows engineers to estimate the probability that
a CT will saturate for a given fault level. This is important for understanding the risk to the network caused
by a change of CT specification.
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

Further work will involve high current testing to further verify the models presented above. Ausgrid will also
perform more current injection tests into different relays and for different protection schemes to see how
they respond to saturated waveforms.

7 References

[1] H. Yu, J. Yuan, and J. Zou, “Design of novel structure current transformer with shielding coils
for overcoming the saturation of core,” IEEE Trans. on Magn., vol.42, no. 4, pp. 1431-1434,
2006.
[2] A. Wright, Current Transformers: Their Transient and Steady State Performance , London:
Chapman and Hall Ltd, 1968.
[3] IEEE Power System Relaying Committee, “Mathematical models for current, voltage and
coupling capacitor voltage transformers,” IEEE Trans. on Power Delivery, vol. 15, no. 1, pp. 62-
72, 2000.
[4] L. A. Kojovic, “Comparison of different current transformer modeling techniques for protection
system studies,” Power Engineering Society Summer Meeting, vol. 3, pp. 1084-1089, 2002.
[5] R. Scheaffer and L. Young, Introduction to Probability and Its Applications, Third Edition,
Boston: Brooks/Cole, 2010, p. 363.
[6] G. Ziegler, “Co-ordination of relays and conventional current transformers,” Cigre Working
Group B5. 02, 2003.
[7] F. King, “Probability,” 2004. [Online]. Available:
http://www.cl.cam.ac.uk/teaching/2003/Probability/prob11.pdf. [Accessed 8 October 2012].
[8] J. Glover, M. Sarma and T. Overbye, Power System Analysis and Design, Fifth Edition, Cengage
Learning, 2010, p. 383.
[9] M. Giridharan, Electrical Systems Design, New Delhi: International Publishing House, 2010, p.
153.

List of Figures
FIGURE 1 THE B-H CURVE FOR THE IDEAL SATURATION MODEL. THE CT IS IDEAL IN THE
VERTICAL SECTION, AND THE SECONDARY VOLTAGE IS ZERO IN THE HORIZONTAL
SECTIONS. SATURATION OCCURS AT ±1.7 TESLA. 2
FIGURE 2 B-H CURVE OF THE STATIC TANH MODEL. 3
FIGURE 3 THE SPH MODEL IN RED SHOWS AN INITIALLY DEMAGNETISED CORE UNDERGOING
SINUSOIDAL EXCITATION WITH AN OPEN CIRCUIT SECONDARY. THE CURVE STARTS AT
B=0, H=0 AND MOVES UP THE INITIAL TRAJECTORY. AFTER REACHING SATURATION IT
THEN MOVES INTO THE STEADY STATE MAJOR LOOP OF THE B-H CURVE. THE BLUE CURVE
WAS EXPERIMENTALLY DETERMINED IN THE LAB AT LOW FREQUENCY (< 1HZ). THERE IS
CLOSE AGREEMENT BETWEEN THE SPH MODEL AND THE EXPERIMENTAL RESULTS. 4
FIGURE 4 SECONDARY CURRENTS VERSUS TIME FOR THE THREE MODELS PRESENTED ABOVE,
DUE TO A 5KA FAULT. ALL THREE CURVES ARE ON TOP OF EACH OTHER SO THE RESULTS
ARE PRACTICALLY IDENTICAL. THE DOTTED LINE REPRESENTS AN IDEAL CT WHICH DOES
NOT SATURATE. 5
FIGURE 5 DARK AND LIGHT GREY CURVES SHOW EXPERIMENTALLY DETERMINED B-H
CHARACTERISTIC OF A GOSS CORE AT 0.5 HZ AND 50 HZ RESPECTIVELY. THE GREY BOX
RUNS FROM -186 TO +186 A/M. THE EXPERIMENTAL CURVES SIT COMFORTABLY WITHIN
THE GREY BOX, SO ALL THREE MODELS PRESENTED ABOVE WILL AGREE TO WITHIN 1%.
6
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

FIGURE 6 SIMULATED LLG THROUGH-FAULT ON A 132 TO 11KV TRANSFORMER DIFFERENTIAL

SCHEME. THE UPPER PLOT SHOWS THE HIGH SIDE CURRENTS, THE CENTRE PLOT SHOWS
THE LOW SIDE CURRENTS. THE LOWER PLOT SHOWS VARIOUS LOGIC STATES WITHIN
THE DIFFERENTIAL PROTECTION RELAY. SATURATION CAN BE SEEN ON THE LOW SIDE
ABOUT 30MS AFTER THE FAULT OCCURS. 8
FIGURE 7 A TYPICAL FAULT CURRENT WAVEFORM IS SHOWN AT TOP WITH FULL DC OFFSET.
BELOW IS THE CORRESPONDING FLUX DENSITY IN THE CT CORE, CALCULATED WITH (5).
THE REMNANT FLUX ( ) IS ASSUMED TO BE ZERO. THE SATURATION FLUX DENSITY IS
SHOWN AS A DOTTED LINE AT 1.7 T. IN THIS EXAMPLE, THE CORE DOES NOT SATURATE
BECAUSE THE FLUX DOES NOT REACH 1.7 T. 10
FIGURE 8 PDF OF THE IDEAL CT FLUX DENSITY AFTER A FAULT, CALCULATED BY (11). THE
NON-SATURATED REGION IS SHOWN IN THE MIDDLE (LIGHTER AREA), BASED ON THE
RANGE SHOWN IN (9). THE TWO OUTER (DARKER) REGIONS REPRESENT FLUX DENSITIES
THAT THE CT WILL TRY TO REACH, BUT WILL FAIL DUE TO SATURATION. THE INTEGRAL
OF THE OUTER AREAS REPRESENTS THE PROBABILITY THAT ANY SATURATION WILL
OCCUR (PROBABILITY = 49.2% FOR THIS EXAMPLE). 11
FIGURE 9 MONTE CARLO VERIFICATION OF THE STATISTICAL MODEL OF SATURATION. ON
THE LEFT IS THE FREQUENCY HISTOGRAM OF THE FINAL FLUX DENSITY OF 100,000
RANDOMLY CHOSEN FAULT SCENARIOS SIMULATED WITH EQUATION (5). ON THE RIGHT
IS THE PDF PLOTTED USING EQUATION (11). CLOSE AGREEMENT BETWEEN THE TWO
METHODS IS SEEN. 12

List of Tables
TABLE 1 SUMMARISED RESULTS OF 12 THOUGH-FAULT SIMULATION TESTS IN A
TRANSFORMER DIFFERENTIAL PROTECTION SCHEME. THE LARGE AND MEDIUM CORES
WERE NOT SUSCEPTIBLE TO NUISANCE TRIPS BUT THE SMALL CORE HAD TWO NUISANCE
TRIPS, AND CAME CLOSE TO TRIPPING FOR THE OTHER TWO. 7

8 Biography: Michael Stanbury

Michael is a graduate engineer with Ausgrid. He was an Ausgrid cadet during his electrical engineering
degree and his final year thesis was on CT saturation. In 2011 he developed a grading tool now used by
Ausgrid to specify protection settings on 11kV feeders with complex branching topologies. He now works
within the Control & Protection section, specifying sub-transmission protection settings.
Prior to his engineering degree, Michael worked in visual effects on several feature films including the 2006
Oscar-winning Happy Feet.

9 Appendix
Here we will derive the flux density waveform of a CT undergoing a typical fault with a DC offset and
remanence. The CT is assumed to follow the Ideal Saturation Model presented above. The CT will have a
resistive burden and the fault will be a sinusoid with an exponentially decaying DC offset.
We begin by deriving the CT’s flux density ( ) waveform for a given fault. All symbols represent
instantaneous values (not RMS). Faraday’s Law of Induction says the CT’s secondary EMF ( ) will be:

where is the number of turns of secondary winding, and is the flux in the core. The core’s flux can be
calculated by:
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

where is the core cross sectional area, and is the magnetic flux density. We model the CT secondary
circuit as a pure resistance ( ) due to the CT winding and the leads from the CT to the relay. Applying
Ohm’s law to the CT secondary circuit, we have

Combining and rearranging the above 3 equations gives a differential equation which can be used to find
over time.

We now assume that the CT does not saturate, i.e. the CT operates in the vertical portion of the B-H curve
shown in Figure 1. If the final flux density is outside of the range to , then this assumption does
not hold, and the CT will be saturated. Given our assumption of no saturation, the secondary current is
equal to the primary current times the turn ratio (as in an ideal transformer):

We substitute this secondary current into (A.4):

We now integrate to find the flux density over time in terms of the primary current .

To perform this integral we must know , which in this case is the fault current waveform. Typical electrical
fault waveforms have a sinusoidal component and an exponentially decaying “DC offset” component (see
the upper plot of Figure 7). They take the following form [8] [9]:

where:
 is the time instant
 is the power system frequency in rad/s

 is the X on R ratio of the primary system up to the point of the fault. This can easily be

obtained with a fault study.

 (Typically about +89˚, as X is much larger than R in a power system)

 is the angle of the voltage sine wave when the fault occurs ( occurs at a voltage zero
crossing, resulting in a large DC offset)
 is the RMS, steady-state fault current. This can easily be obtained by a fault study.

Now that we have the fault waveform , we can evaluate the integral in (A.7).
 CIGRE Australia APB5
SEAPAC 2013
Brisbane 12-13 March
Modelling of Current Transformer Saturation for Power System Protection Applications

We now calculate the constant of integration . We choose this value to be such that at (when the
fault occurs), the flux density in the core ( ) equals the remnant flux density :

By substituting , and into (A.9), we can solve for . Having found it, we can now substitute it
back into (A.9) to obtain the final solution of the flux density waveform:

Equation (A.10) is used in the main part of the paper, referred to as equation (5). Note that this equation
assumes there is no saturation, and that the CT is ideal. This assumption is checked distinguishing between
final flux densities within the range of equation (9), and those outside of this range. If the final flux density
is outside of this range, then the assumption of no saturation is not upheld, and therefore the CT has
saturated.

View publication stats