Anda di halaman 1dari 16

A randomised leader progression model for

backashover studies
Ian Cotton
1,
*
,y
and Mohd Zainal Abidin Ab Kadir
2
1
School of Electrical & Electronic Engineering, The University of Manchester, PO Box 88, Manchester M60 1QD, UK
2
Department of Electrical & Electronic Engineering, University Putra Malaysia, 43400 Serdang, Malaysia
SUMMARY
Insulation coordination models are an essential part of power system studies and are used to determine the risk of
damage to plant during abnormal conditions such as those experienced when lightning strikes the power system.
One model used to predict the performance of an overhead line when hit by lightning examines the likelihood of a
backashover; the breakdown of an insulator coordination gap due to a rise in tower potential in comparison to a
phase conductor. A backashover model has many components that must be accurately modelled including the
lightning strike, the tower, the shield wire/phase conductors, the insulator coordination gaps and the tower earthing
system. Of all of these, the only two signicant challenges are the accurate modelling of the insulator coordination
gaps and the tower earthing system (the soil ionisation mechanism in particular). This paper focuses on the
simulation of the insulator coordination gap for which available models range from a simple voltage controlled
switch to the more complex leader progression model (LPM). The paper details the use of experimental data to
produce a modied LPM in which certain model parameters are randomly varied. The development of this model
using laboratory data is detailed, as is the effect that using this model has on the results of two insulation
coordination studies that relate to backashovers. Copyright # 2007 John Wiley & Sons, Ltd.
key words: overhead line; lightning; backashover; insulation coordination
1. INTRODUCTION
Typically, more than 2000 thunderstorms are active throughout the world at a given moment, producing
in the order of 100 lightning ashes per second [1]. For most overhead power transmission lines,
lightning is therefore the primary cause of unscheduled interruptions. Several methods for estimating
lightning outage rates have been developed in the past and many publications have been written on how
to design transmission lines that experience minimum interruption. Even so, there is still a need to
continue the improvement of accurate models that allow the overvoltages and outage rates resulting
fromlightning strikes to transmission lines to be predicted with increasing accuracy. This is particularly
the case considering the increasing interest in compaction of both overhead lines and substations for
which more accurate measures of likely overvoltages at substations is required.
EUROPEAN TRANSACTIONS ON ELECTRICAL POWER
Euro. Trans. Electr. Power 2008; 18:709724
Published online 20 August 2007 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/etep.200
*Correspondence to: Ian Cotton, School of Electrical & Electronic Engineering, The University of Manchester, PO Box 88,
Manchester M60 1QD, UK.
y
E-mail: ian.cotton@manchester.ac.uk
Copyright # 2007 John Wiley & Sons, Ltd.
1.1. The backashover process
For lightning, overhead transmission line outages caused by the backashover mechanismpredominate
over those outages caused by shielding failure [2]. Overhead line shield wires are generally located to
minimise the number of lightning strokes that terminate on the phase conductor [1], the vast majority of
strokes therefore terminate on the shield wire(s).
The backashover process is illustrated in Figure 1. When lightning strikes the top of a transmission
line tower, the combination of the shield wire surge impedance and the injected lightning impulse current
will produce a voltage at the tower top. This surge voltage propagates in three ways, one surge propagates
in each direction of the shield wires while another surge propagates down the tower itself. As the surge
propagating down the tower passes the cross-arms, the voltage across the insulator string is increased.
However, a number of other events can take place to reduce the voltage across the insulator string. The
rst is electromagnetic coupling between the shield wire and the phase conductors that results in a
proportion of the shield wire voltage surge being coupled onto the phase conductor, this is shown as ke
in the gure. Second is the reection of the surge that propagates down the tower once it reaches the base.
Through good earthing and soil ionisation, a signicant negative voltage reection will usually propagate
back up the tower. Dependent on the travel time of the surge and the rise time of the lightning current, this
negative reection may reach the tower cross-arms in time to reduce the voltage seen by the insulators.
Finally, but usually insignicantly due to the signicant travel times involved, a further negative reection
will be generated when the surges propagating along the shield wire reach the next tower.
The combination of all of these processes and the non-linear effect of corona mean that the voltage
waveform that develops across the insulator is certainly not similar to that used in a laboratory
Figure 1. Illustration of backashover process on an overhead transmission line.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
710 I. COTTON AND M. Z. A. AB KADIR
environment to test insulators and coordination gaps before they are used on a high voltage system.
Whatever the shape of this voltage, a backashover will result if it is of a sufcient magnitude. The
backashover will usually take place across the coordination gaps of the insulator that are placed there
to ensure that the insulator is not damaged should the voltage difference between the phase conductor
and the cross-arm exceed a given value.
When a backashover occurs, part of the surge current will be transferred to the phase conductors
through an arc across the insulation strings. This surge current creates a transient voltage on the phase
conductors that will then propagate to the end of line substation and may, if designs are not correct,
cause damage to connected equipment. The backashover will also cause a temporary line-to-ground
fault that will be cleared by a circuit breaker. Therefore, a line outage will result in every case as well as
the possibility of equipment damage.
Much work has gone into producing accurate models of this process and one of the most difcult
items to model accurately is the insulator/coordination gap breakdown. It is clear from the above
discussion that the overvoltage across an insulator resulting from lightning does not have the standard
1.2/50 microseconds waveshape. Non-linear effects such as corona, soil ionisation and reections of
voltages within a tower and between towers themselves tend to distort the surge voltages from the
standard impulse waveshape [4].
The evaluation of the strength of line insulation when exposed to non-standard waves has been
therefore been investigated by a number of researchers and it is clear that there is a volttime (VT)
dependence [5,6]. This time dependence is included in models that estimate the level of backashovers
on a transmission line such as those proposed by IEEE and CIGRE [6,7]. The behaviour of this variety
of insulator coordination gap models when used in transient simulation software is to be examined
within this paper in addition to the formulation of a modied version of the LPM.
1.2. Coordination gap ashover models
The simplest model available of an insulator coordination gap is that of a simple switch (SW) that is
taken to close when the voltage rises above a given value. This, while easily implemented in transient
simulation software, does not model any VT dependence of the ashover process. The state of the
switch is dened as in Equation 1. It is usual to take the critical ashover voltage of an insulator for the
threshold above which breakdown occurs.
0open if U
switch
< U
cfo
1closed if U
switch
U
cfo
(1)
A more complex model to implement in transient simulation software is the VT model [8]. For this
type of model, the ashover voltage of the coordination gap is calculated at each step in time once the
surge has arrived using the equation given as Equation 2. In this equation L is the insulator length, the
VT curve being a function of this length. The equation includes a function of time t and two arbitrary
constants, a and b. In this equation, zero time is taken to be when the surge rst arrives at the
coordination gap as opposed to the start of a simulation in which a lightning voltage may develop some
distance away from and take a nite time to propagate to the insulator in question.
U
flashover
aL
bL
t
0:75
(2)
An alternative empirical model that also includes the effect of time is the disruptive effect (DE)
model [7]. This model integrates the voltage across on a coordination gap once it rises above a certain
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
A MODEL FOR BACKFLASHOVER STUDIES 711
threshold, this integral being given in Equation 3. Once this integral reaches a set value, the
coordination gap is taken to ashover. Some simplied versions of this model set U
o
, the onset voltage,
to zero for ease of computation.
DE
Z
t
0
Ut U
o

k
dt (3)
LPM is based on the three different physical phases in the discharge development process, namely
corona inception, streamer propagation and leader propagation. The model sums the time taken for
these three physical phases of the discharge process to take place. All of the stages have been studied
within a laboratory environment and equations have been developed to represent all of them. In other
words, this is the only model that is the closest approximation of the physical processes taking place
when breakdown occurs. The time to breakdown, t
b
, can be expressed as a sum of three components as
given in Equation 4 [7,9].
t
b
t
i
t
s
t
l
(4)
In this equation, t
i
is the corona inception time, t
s
is the time for streamers to cross the gap or meet the
streamers from the opposite electrode and t
l
is the leader propagation time. It is necessary to describe
each of these model phases in turn to accurately describe the operation of the LPM.
With regards the corona inception time, most practical coordination gaps have a non-uniform eld.
At fairly low voltages, the corona inception voltage is therefore usually far below the breakdown
voltage. When this eld non-uniformity is coupled with the fast rate of rise of voltages during studies
connected with lightning, the corona inception time can usually be regarded as irrelevant when it is
compared to the streamer and leader propagation times.
The time for streamer propagation, t
s
, is normally computed using Equation 5 where E
max
is the
maximum average electric eld strength in the gap and E
50
is an experimentally derived value as
detailed in data from Pigini et al. [10] and shown in Table I.
1
t
s

1:25E
max
E
50
0:95 (5)
The values in Table I were produced for a rodplane gap when only one streamer is expected. This
approach can be modied to account for the formation of two streamers in a conguration such as a
rodrod gap if necessary.
The time for leader propagation, t
l
, is calculated by determining the leader velocity at a given time
step of a model and thus calculating the time it takes for the leader to propagate across the gap. As a
Table I. Constants for use with leader progression model equations.
Conguration Polarity K (m
2
v
2
s
1
) E
50
(kVm
1
)
Air gaps, post and long rod insulators 0.8 10
6
600
1 10
6
670
Cap and pin insulators 1.2 10
6
520
1.3 10
6
600
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
712 I. COTTON AND M. Z. A. AB KADIR
practical simplication, the formula below was introduced to calculate the change in the leader length
in a particular time step.
dl
dt
K Ut
Ut
g 1
E
50

(6)
In Equation 6, K is a constant as detailed in Table I, g is the gap length, U(t) is the voltage across the
gap at a particular time step and l is the leader length at the present time.
1.3. Modied leader progression model
If all of the ashover models described were run for a particular waveshape and a particular voltage
they would always yield the same answer. This is not correct in the practical case as if an insulator
coordination gap is tested with a set shape of waveform at its 50% ashover voltage, it will breakdown
50%of the time. The time to breakdown will also vary for each occasion when a breakdown does occur.
In insulation coordination modelling, this random nature of the operation of a coordination gap could
have a signicant inuence on overvoltage distributions given the fast rise time of lightning currents.
For this reason, in the studies to be detailed, an alteration will be made to the LPM model by
introducing a statistical distribution to the constants E and K in Equation 6. This is done by expressing
E and Kusing a Weibull distribution. However, to formulate the appropriate parameters to express these
values in the Weibull distribution, experimental testing is necessary. AWeibull distribution was chosen
for this work owing to the ability to have signicant control (relative to other distributions) over its
shape by modication of the shape parameter.
2. LABORATORY TESTING TO VERIFY INSULATOR COORDINATION GAP MODELS
Figure 2 shows a simplied version of the experimental arrangement that was used to carry out testing
to help dene suitable parameters for use with all of the coordination gap models described in the
previous section. In this diagram the multi-stage impulse generator actually used in the laboratory has
been represented as a single stage. The test object was a shortened 400 kVinsulator string tted with a
coordination gap with a length of 1.1 m. Astandard impulse generator circuit connects to the test object
and a chopping gap is connected in parallel to it via a resistance. The chopping gap is represented as a
capacitance, C
G
, in parallel with a switch. The chopping gap allows the production of non-standard
waveforms across the test object. In particular, it allows the sudden reduction in the voltage across the
test object in the same way that an insulator would see a reduction due to a reection from the tower
base.
Three different voltage waveforms were produced using this circuit. The rst was a 1/47microseconds
waveform, this being within the allowable tolerances of a standard 1.2/50microseconds waveform. The
second was a 2.4/51microseconds waveform and therefore had a signicantly longer front time and a
marginally longer tail time than the rst waveform. The nal waveform was the chopped version and had
a 2.4/18microseconds waveform of the type shown in Figure 3. All waveforms were negative in polarity.
Impulses of all three waveforms were repeatedly applied to the test object at various voltage levels.
For all of the tests, the prospective peak voltage and whether breakdown occurred was recorded. If
breakdown did occur, the voltage and time at which breakdown took place were also noted.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
A MODEL FOR BACKFLASHOVER STUDIES 713
The data were initially used to select suitable parameters for the constants used in Equations 2 and 3
for the VTand DE models. No tting of data was done to the SW model owing to its simplicity. Based
on the experimental data, suitable parameters such as a and b in Equation 2 and the value of the
integral in Equation 3 could be chosen to provide a best t between the experimental data for the 1/
47 microseconds waveform and models of each gap run within the PSCADenvironment. The models in
the PSCAD environment consisted of the circuit shown as Figure 2 plus control circuitry that measured
the voltage across the test object and compared it with the various outputs of the coordination gap
equations to establish whether a ashover should occur.
The rst graph given in Figure 4 shows the VT curve obtained during tests using the 1/
47 microseconds waveform along with a set of data points taken from the best t to the relationship
previously described in Equation 2. The experimental data plotted used an average from three shots at
each voltage level. The experimental data were referred to standard atmospheric conditions, pressure
and temperature having been measured during the tested. A good match between the experimental data
and the simulations was achieved in this particular case.
Figure 2. Experimental arrangement used to test insulator coordination gaps.
Figure 3. Chopped waveform used to test insulator coordination gaps.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
714 I. COTTON AND M. Z. A. AB KADIR
To signicantly test the models, they were then applied without alteration to the parameters
previously selected to the 2.4/18 microseconds waveform. This would allow the applicability of a
single model to more than one waveform type to be evaluated. In the case shown in Figure 4, the match
between the results from the model and the actual test data is shown to be good with the exception of
those obtained for a very long time to breakdown. However, the range of experimental data available for
comparison with the computations was limited. The inclusion of waveforms with stepped fronts and/or
oscillatory components would have been benecial to clarify this point further.
Figure 4. Results from t of VT model to 1/47 microseconds data (1st graph) and application to 2.4/18 micro-
seconds waveform (2nd graph).
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
A MODEL FOR BACKFLASHOVER STUDIES 715
Two data t procedures were applied for the LPM model. First, the same procedure as that described
above was applied to a LPMmodel where the values of E &Kused in the dening equations were xed.
A second data t procedure was then used in the case where the parameters E & K were taken to
statistically vary, i.e. the modied LPM model. From the experiments, data were available that
provided information about the probability of breakdown as a function of the prospective peak test
voltage. The selection of parameters to describe the Weibull distributions representing the statistically
varying E and K was done using a brute force approach where many combinations of alternative
distributions of E and K were evaluated until the best t was made between the experimental data and
the computational model.
In this case, the LPMwas run many times to produce a curve giving probability of breakdown against
prospective peak voltage. This curve was then tted against the experimental data for the 1/
47 microseconds waveform before the same model was tested against the other waveforms. A good
match between the experimental results and the model was found for the other waveforms. A selection
of the results for this case is shown in Figure 5.
To compare all models against each other, the RMS errors of the model ts in comparison to the
experimental data were calculated using results from VT graphs of the type shown in Figure 4. The
results are shown in Table II. The switch model was taken to be as described in Equation 1 with the
voltage limit being taken to be the critical ashover voltage of the 1.1 m coordination gap.
The results show that the lowest errors for all models except the DE are for the 1/47 microseconds
waveform. This should be the case as the data from the experiments using this waveform were used to
select the model coefcients. However, when different waveforms are applied, the LPM model
generally outperforms the VT and the DE models. In all cases it is clear that the simulation of a
coordination gap by a simple switch is inappropriate. The lowest average error when considering all of
the waveforms is seen for the standard LPM, the modied one not performing particularly well for the
2.4/18 microseconds waveform. This may suggest that the randomised parameters are not the most
appropriate.
3. APPLICATION OF THE INSULATOR COORDINATION GAP MODELS TO
BACKFLASHOVER STUDIES
Two forms of modelling were carried out to investigate the application of the various insulator
coordination gap models to backashover studies. One model examined the likely system
backashover rate due to a mid-span lightning strike to the overhead line, i.e. the substation was
considered to be a signicant distance away from the lightning strike location. The second model
examined the impact of a lightning strike close to a substation in which a transformer was connected to
the end of line. This model was used to examine the impact of the various insulator coordination gap
models on the predicted level of damage to the transformer. Both models were produced in the PSCAD
software [8].
3.1. Backashover current estimation
The conguration of the rst model is shown in Figure 6 and is based on a version of the British 400 kV
system[11]. Aline is modelled complete with three towers that are separated by 250 m. Line sections of
50 m are used and these have a corona model applied to them. At both ends of the line, a matching
impedance is used to model an innite line. The lightning strike is placed in one of four locations along
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
716 I. COTTON AND M. Z. A. AB KADIR
Figure 5. Modied leader progression model comparison of experimental and tted data for the 1/47 micro-
seconds waveform (1
st
graph) and the comparison of the resulting data for the 2.4/18 microseconds waveform.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
A MODEL FOR BACKFLASHOVER STUDIES 717
the overhead line from the tower itself to mid-span. Insulator coordination gap models are included on
the tower, T2. All the types of models described earlier in this paper were converted into a formsuitable
for PSCADand implemented in the studies. For the LPM models, no adjustment was necessary. For the
switch, VTand DE models, adjustments had to be made to account for the longer length of coordination
gap. Further specic details relating to the model are described in Table III.
Table IV describes the results obtained from the PSCAD software. To achieve these results, negative
lightning currents from 5 kA upwards were injected into the model at each location in 1 kA steps. The
current that caused backashover was then recorded. This was repeated for each form of insulator
coordination gap model. The value shown for the modied LPMmodel is an average value produced by
multiple iterations of this process. As the model has statistical variation, the current that is required to
cause backashover will also vary.
Two obvious issues from the results are the failure to obtain a backashover for any position other
than the tower being struck for the VT model and the very low currents suggested by the SW model.
Moving away fromthese issues, the LPMmodels then showsignicantly higher levels of backashover
currents than the DE models. It is therefore useful to compare the values predicted by the various
models to literature. The LPM values seem valid when compared with a backashover current reported
by CIGRE [7]. For a double circuit 230 kV line with a height of 35 m, the backashover current is
116 kA. For a single circuit 500 kV line with a height of 25 m, the backashover current is 164 kA.
Another example is the data reported for lightning performance of transmission lines of 132 kV [13].
Table II. RMS error of ts to all waveforms using alternative insulator coordination gap ashover models.
Waveform (microseconds) Root mean square error (RMSE) in %
Std LPM Mod LPM VT DE SW
1/47 2.6 1.5 3.3 6.0 20.3
2.4/18 2.6 9.6 13.0 32.0 29.2
2.4/51 3.9 2.3 7.0 5.2 19.1
Average 3.0 4.5 7.8 14.1 22.9
Figure 6. Schematic of PSCAD model used to examine backashover rate.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
718 I. COTTON AND M. Z. A. AB KADIR
For a 30 mtower height, the current is 93 kAwhilst for a 60 mheight, the current is 126 kA. Considering
both of these cases refer to lower voltage systems than that described in this paper, it would suggest that
the values predicted using the DE model are particularly low.
In general, the modied LPM model predicts slightly lower values of backashover current in
comparison to the standard invariant LPM model. The small difference shows that the predicted
backashover rate will be virtually identical in both cases. The use of the modied LPM model in
accurately predicting backashover rates is not, therefore, necessary. However, the values shown are
the average value for the modied LPM. When carrying out simulations, backashovers will occur for
currents either side of this value according to the probabilistic value of E and K. The possible impact
that these varying backashover currents may have on the probability of damage to a transformer is an
important concern and this is addressed in the next section.
3.2. Predicting the probability of damage to a substation transformer
A new model was produced using the basic template given in Figure 6. For this model the matching
impedance section on the right hand side of the model was replaced with a simple model of a
Table III. Key parameters used in modelling backashover rate to 400 kV system.
Model Sub-component Details/reference
Lightning strike Double exponential current source with a varying front time
according to the peak current [7]
Overhead line Shield wire Single shield wire, 1.43 cm radius (Zebra), 537 V surge impedance,
50 m height. Modelled with frequency dependent travelling wave model
Phase conductor Single phase conductor, 1.43 cm radius (Zebra), 300 V surge impedance,
lowest phase conductor at 20 m. Modelled with frequency dependent
travelling wave model
Corona model According to Carneiro and Marti model [12] implemented every 50 m
along line
Tower Main Structure Surge impedance of 155 V and travelling wave velocity of speed of light
modelled with a Bergeron model
Tower footing DC resistance of 10 V in a soil resistivity of 100 Vm. Soil ionisation
modelled [8]
Coordination gap Rodrod gap with distance of 2.54 m
Table IV. Current required to cause a backashover for the different coordination gap models and the different
lightning strike positions.
CG model Current required to cause backashover (kA)
Position of lightning attachment
1 2 3 4
Std LPM 110 155 175 171
Mod LPM 100 144 168 170
VT 164 NO

NO

NO

DE 42 65 75 77
SW 17 34 47 53

No backashover occurred in these cases for current values below 200 kA.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
A MODEL FOR BACKFLASHOVER STUDIES 719
transformer that included the distributed inductance and capacitance of the windings. No mutual
coupling between phases was modelled. The transformer was itself protected by coordination gaps set
at a distance of 1.68 m. It is noted that it is preferable to also protect transformers with surge arresters
owing to the high rates of dU/dt that can be produced by ashover of a coordination gap. However,
many transformers on the UK system are still protected in this manner. The overhead line shield wire
was not connected into the substation. For this set of simulations only a lightning strike to the tower was
modelled. For all coordination gap models the peak voltage observed across the transformer was
measured for values of lightning current ranging from 5 to 200 kA in 5 kA steps. For the modied LPM
model, simulations were run multiple times to allow the statistical nature of the LPM model to have an
inuence on the results.
Figure 7 gives the results of this analysis (although only one run of the modied LPMmodel is shown
for simplicity). For all models, the initial, near linear, phase of the graph shows when no backashovers
occur. The maximumvoltage seen at the transformer is therefore purely determined by electromagnetic
coupling between the shield wire and the phase conductors. The non-linearity of the graph at this point
is explained by effect of the corona models and the changing rate of rise of current as a function of the
prospective current peak.
Above certain current levels (e.g. 50 kA for the DE model and around 120 kA for the LPM models),
backashovers take place and the voltages seen at the transformer suddenly increase. In the case of the
switch coordination gap model, the ideal nature of the model allows the voltage to be clamped to a
maximum of around 1500 kV no matter what the current. The other models all show a varying
relationship between the injected current and the peak voltage. As would be expected from the results
in the previous section the VT model does not produce backashovers unless the current is extremely
high, i.e. above around 175 kA.
As the probability of lightning current exceeding a specic value is easily calculated from formulae
[14], it is straightforward to convert the results given in Figure 7 into curves giving the probability of the
Figure 7. Peak voltage seen at transformer versus lightning current magnitude for all coordination gap models.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
720 I. COTTON AND M. Z. A. AB KADIR
voltage at the transformer exceeding a specic value, the basic impulse level usually being the one of
interest as voltages above this value can be taken as causing damage to the transformer. Figure 8
displays the relevant curves from this analysis. All runs of the modied LPM model have been taken
into account to produce this gure.
Using the gure allows the production of Table Vthat gives the probability of transformer damage by
a single lightning strike to the tower closest to the substation. Small percentages are clearly expected in
this form of analysis as it would otherwise suggest that the coordination gaps used on a system of the
design described are not appropriate for providing protection against transients.
The VT model predicts no damage at all while the SW model predicts one incident of damage for
every 10 lightning strike events. The DE model predicts damage for just under 5 out of every 100
lightning strikes while the LPM models both suggest damage will occur for just 2 out of 10 000
lightning strikes. The analysis shows the wide variation in results that can be obtained from the use of
different coordination gap models. It would suggest that the DE and SW models are inappropriate as
this level of damage would not be expected in a real transmission system.
4. DISCUSSION AND CONCLUSIONS
Five different types of insulator coordination gap have been modelled within the PSCAD software
environment. The operation of these models has been compared against laboratory data produced by
testing a real coordination gap with lightning impulses. The models have then been applied to two
systems within the PSCAD environment, one to estimate the backashover rate of a 400 kV system and
one to estimate the probability of damage to a 400 kV transformer located at a substation next to a
struck tower.
Figure 8. Cumulative probability distribution of maximumvoltage observed at substation transformer being lower
than the stated value.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
A MODEL FOR BACKFLASHOVER STUDIES 721
The laboratory testing was carried out using a 1.1 m coordination gap. This is shorter than the 2.5 m
length expected on a 400 kV system but the maximum voltage available from the test apparatus limited
the expansion of this gap. The basic conclusions expected from this work are, however, expected to be
valid for gap sizes in the same order of magnitude as that tested, i.e. for gaps used on typical
transmission systems.
In respect of the experimental testing, all the model parameters that could be adjusted were set
to values that gave the best match when experimental data were compared with model output for the
1/47 microseconds negative lightning impulse waveform. Without adjusting these model parameters, the
models were then run in software in which two alternative lightning impulse waveforms were applied, the
2.4/18 and the 2.4/51microseconds. The results from these simulations were then compared with
experimental data gained from testing the physical coordination gap with these waveforms.
Both forms of LPM were shown to give good matches to all of the experimental results giving the
lowest RMS error when a direct comparison was performed. This in itself should give condence to the
ability of the LPM models to operate accurately for a wide range of waveshapes. This ability is,
presumably, based on the need to physically model the growth of the leader within the model. The good
match with experimental results should also infer that if these models are used in insulation
coordination models, they should give a high accuracy in comparison to other models.
Two comprehensive models of a 400 kV system were then produced. Both models had accurate line
models, tower models and included non-linear processes such as corona and soil ionisation. Initially,
backashover rates of the system were calculated for all types of coordination gap model. In
comparison to literature, the LPM model appears to give good results. When examining the probability
of damage to a transformer in the second model, signicant differences were observed in the results.
Probabilities of damage to the transformer ranging from 0 to 15% were predicted by the models. It is
more difcult to say which model is correct in these circumstances but the LPM model must have
condence associated with it given the performance in the comparison with experimental testing.
Overall it is clear that the choice of coordination gap model in insulation coordination studies is an
important one. Signicant variations in results will occur between models that use different
coordination gap models, and care must therefore be taken when trying to produce comparative
analyses of research work/insulation coordination studies.
The modication of the LPM model to account for statistical variation in the E and K parameters
and the natural variation observed in practice was successful. The modied model when implemented
in PSCAD could reproduce experimental curves that showed the probability of breakdown for a given
value of voltage. However, the differences between the results produced by the standard LPM model
versus the modied model during the actual insulation coordination studies were not signicant. The
modication of the LPM model may therefore prove on occasions to be a useful research tool but is not
essential for insulation coordination studies.
Table V. Probability of transformer damage due to a lightning strike at the tower closest to the substation.
Coordination gap model Probability of damage to
transformer (%)
VT 0.00
DE 4.57
Mod LPM 0.02
Std LPM 0.02
SW 10.26
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
722 I. COTTON AND M. Z. A. AB KADIR
5. LIST OF ABBREVIATIONS AND SYMBOLS
5.1. Abbreviations
DE disruptive effect model
LPM leader progression model
Mod modied
RMS root mean square
Std standard
SW simple switch model
VT voltagetime model
5.2. Symbols
a, b constants
C
G
chopping gap capacitance
E
max
maximum electric eld in gap
E
50
experimentally derived constant
g gap length
K constant
K
e
coupling coefcient
l leader length
t
b
time to breakdown
t
i
corona inception time
t
s
streamer propagation time
t
l
leader propagation time
U
cfo
critical ashover voltage
U
switch
voltage across switch
U
ashover
ashover voltage
U
o
onset voltage
U(t) voltage across gap at given time step
REFERENCES
1. Hileman AR. Insulation Coordination For Power Systems. Marcel Dekker: New York, USA, 1999.
2. Gilman DW, Whitehead ER. The mechanismof lightning ashover on high voltage and extra high voltage transmission lines.
Electra 1973; 27:6596.
3. IEC. Insulation Coordination Part 1: Denitions, Principles and Rules. IEC 60071-1, 1993.
4. Chowdhuri P. Electromagnetic Transients in Power Systems. Research Studies Press Ltd.: Somerset, UK, 1996.
5. IEEE Power Engineering Society. IEEE guide for improving the lightning performance of transmission lines. IEEE Std.
1243-1997, 1997.
6. CIGRE. Guide to procedures for estimating the lightning performance of transmission lines. CIGRE Brochure 63, 1991.
7. Woodford D. PSCAD/EMTDC: Getting Started. Manitoba HVDC Research Centre, Inc.: Canada, 1998.
8. IEEE Modelling and Analysis of System Transients Working Group. Modelling guidelines for fast front transients. IEEE
Transactions On Power Delivery 1996; 11:493506.
9. Pigini A, Rizzi G, Garbagnati E, Porrina A, Baldo G, Pesavento G. Performance of large air gaps under lightning
overvoltages: experimental study and analysis of accuracy of predetermination methods. IEEE Transactions On Power
Delivery 1989; 4:13791392.
10. British Electricity International (BEI). Modern Power Station Practice: EHV Transmission, Vol. K. Pergamon Press: Oxford,
UK, 1991.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
A MODEL FOR BACKFLASHOVER STUDIES 723
11. Carneiro S, Marti JR. Evaluation of corona and line models in electromagnetic transient simulation. IEEE Transaction On
Power Delivery 1991; 6:334342.
12. Irwin T, Ryan HM. Transmission and distribution: part 2, Chapter 2. Ryan HM (ed.). High Voltage Engineering and Testing:
IEE Power Series 17, 2nd edn. IEE: Stevenage, UK, 1998.
13. Popolansky F. Frequency distribution of amplitudes of lightning currents. Electra 1972; 22:139147.
AUTHORS BIOGRAPHIES
Ian Cotton received a Class I B.Eng. (Hons) degree in Electrical Engineering from the
University of Shefeld in 1995 and a Ph.D. in Electrical Engineering from UMISTin 1998. He
is currently a Senior Lecturer in the Electrical Energy and Power Systems Group of the School
Of Electrical and Electronic Engineering at the University of Manchester. His current research
interests include power systems transients, the use of higher voltage systems in aerospace
applications and power system induced corrosion. He is a Senior Member of the Institute of
Electrical and Electronic Engineers, a Member of the Institution of Electrical Engineers and a
Chartered Engineer.
Mohd Zainal Abidin Ab Kadir graduated with a B.Eng. degree in Electrical and Electronics
from the Universiti Putra Malaysia in 2000 and obtained his Ph.D. from the University of
Manchester, UK in 2006 in High Voltage Engineering. Currently, he is a Lecturer in the
Department of Electrical and Electronics Engineering, Faculty of Engineering, University
Putra Malaysia. His research interests include high voltage engineering, insulation coordi-
nation, lightning protection and power system transients.
Copyright # 2007 John Wiley & Sons, Ltd. Euro. Trans. Electr. Power 2008; 18:709724
DOI: 10.1002/etep
724 I. COTTON AND M. Z. A. AB KADIR

Anda mungkin juga menyukai