Historia de Las Correcciones de Aire

University of the Witwatersrand, Johannesburg
MASTER’S DISSERTATION
Investigating the Effects of Altitude (Air Density) on the HVDC

Breakdown Voltage of Small Rod-Plane Air Gaps
Author: Tatenda Gora
A dissertation submitted to the Faculty of Engineering and the Built Environment, Uni-
versity of the Witwatersrand, in fulfilment of the requirements for the degree of Master
of Science in Engineering.
Johannesburg, 2016
DECLARATION
I declare that this dissertation is my own unaided work. It is being submitted for a
Degree of Master of Science to the University of the Witwatersrand, Johannesburg. It
has not previously been submitted for any degree or examination to any other university.
Signature:
i
ABSTRACT
The validity of the atmospheric correction method presented in the IEC 60060-1 (2010)
standard is analysed and evaluated by means of theoretical and laboratory work. In
order to understand the problem, the evolution of the atmospheric correction methods,
from as early as 1914, has been presented. A procedure (Calva prediction method) for
predicting the direct current (DC) breakdown voltage for an air gap at any altitude
was discovered and was also analysed along with the IEC 60060-1 (2010). A critique of
some of the atmospheric correction methods commonly used standards was also done.
Experiments were carried out at altitudes of 1 740 m (Wits University), 130 m (UKZN
HVDC centre) and at less than 2 m above sea level (Scottburgh beach, Clansthal). More
tests were conducted using a pressure vessel where high altitude relative air density was
simulated. All tests were conducted on rod-plane air gaps using a 15 mm diameter
flat tip rod. Test results from Scottburgh beach were used as the standard breakdown
voltages of the air gaps tested since the environmental conditions were the closest to the
conventional standard conditions (stp). The test results obtained were compared with
predictions using the Calva method in order to validate the method. The test results
were also corrected according to IEC 60060-1 (2010) and compared to the standard
breakdown voltages obtained at Scottburgh beach. It was shown that the IEC 60060-1
(2010) is quite suitable for atmospheric correction for data obtained at low altitudes
(about 130 m). When applied to high altitude (1 740 m) data, the correction method
is accurate and suitable for very small air gaps less than 0.1 m. As the air gap length
increased, the corrected results began to deviate from the expected standard voltage.
The same trend was shown with the corrected results from the pressure chamber tests.
The prediction method by Calva was accurate when compared to the experimental data
from the high altitude and low altitude test results. When compared to the data from
the pressure chamber, the prediction method had a linear error factor which was different
for each gap length. It was concluded that the IEC 60060-1 (2010) is not only unsuitable
for atmospheric correction for data at relative air densities below 0.8, but also that the
correction method is prone to an increase in error as the air gap length increases when
the relative air density is higher than 0.8. The Calva prediction method was found
to be suitable to use after additional factors are added when applied to high altitude
conditions.
ii
ACKNOWLEDGEMENTS
The author would like to acknowledge with gratitude the mentorship and aid provided by
professor Cuthbert Nyamupangedengu, professor Ian Jandrell and Mr. Nishanth Parus.
The author would also like to sincerely acknowledge the ISH2009-SAIEE Postgraduate
Research Scholarship in High Voltage Engineering for directly funding the research.
Gratitude is also extended to Dr. Andrew Swanson and the UKZN HVDC centre for
providing some equipment and their laboratory for experimental work. Gratitude is
given to Eskom for their support of the High Voltage Engineering Research Group at
Wits University through TESP. The author would also like to express gratitude to the
Department of Trade and Industry (DTI) for THRIP funding and to thank the National
Research Foundation (NRF) for direct funding of the research group.
iii
Contents
DECLARATION i
ABSTRACT ii
ACKNOWLEDGEMENTS iii
Contents iv
List of Figures vi
List of Tables viii
Abbreviations ix
1 Introduction 1
2 Historical evolution of Altitude Correction Methods 4

2.1 Peek Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 IEC 60-1 (1973) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 IEC 60060-1:1989(Pigini Model) . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Ramirez Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.5 Calva Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.6 Some Irregularities In The Commonly Used Standards . . . . . . . . . . . 13
2.6.1 IEC 60060-1 (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6.2 IEC 60071-2 (1996) . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6.3 IEEE Std 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Breakdown Mechanisms in Air 18

3.1 Non-Uniform Field Breakdown . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.1 Ionization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.2 Electron Avalanche Mechanisms . . . . . . . . . . . . . . . . . . . 20
3.1.3 Streamer Development and Propagation . . . . . . . . . . . . . . . 21
4 Experimental Set-Up 24
4.1 The Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.1 Open Air Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.2 Pressure Vessel Tests . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 Laboratory test configuration and Equipment . . . . . . . . . . . . . . . . 25
4.2.1 Test configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
iv
Contents v
4.2.2 Equipment Used in the Tests . . . . . . . . . . . . . . . . . . . . . 26

4.3 Test Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.3.1 Open Air Experiments . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.3.2 Pressure Vessel Experiments . . . . . . . . . . . . . . . . . . . . . 32
5 Experimental Results, Analysis and Discussion 33

5.1 Open Air Gap Breakdown Experiments . . . . . . . . . . . . . . . . . . . 33
5.1.1 Sea Level Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.1.2 Low Altitude Tests (UKZN) (≈ 130 m asl) . . . . . . . . . . . . . 34
5.1.3 High Altitude Tests (Wits University)(≈ 1 740 m asl) . . . . . . . 37
5.1.4 IEC 60060-1 Correction of Low Altitude and High Altitude Results 39
5.2 Pressure Vessel Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.2.1 Why Breakdown Voltage Initially Decreased But Started Increas-
ing With Further Decrease In Pressure . . . . . . . . . . . . . . . . 46
5.2.1.1 Classical Theory of Breakdown Mechanisms . . . . . . . 46
5.2.1.2 Corona Theory Under DC Stress . . . . . . . . . . . . . . 48
5.2.2 Atmospheric Correction Factors on Pressure Vessel Test Results . 50
5.2.2.1 IEC 60060-1 . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2.2.2 Calva’s Method Applied For Comparison With Pressure
Vessel Tests . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2.2.3 Mapping of Calva’s Prediction Method to the Experi-
mental Data . . . . . . . . . . . . . . . . . . . . . . . . . 55
6 Conclusions and Recommendations 62

6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.2 Recommendations For Future Work . . . . . . . . . . . . . . . . . . . . . 63
A Experimental Data And Results 69
B SAUPEC 2015 Discussion Paper 72
C ISH 2015 Publication 79
D Cigré WG D1.50 Contribution 86

List of Figures
2.1 Summary of the development of altitude correction methods in IEC 60060-

1(2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1 Electric field distribution in a rod-to-plane gap . . . . . . . . . . . . . . . 19

3.2 Processes that occur leading to breakdown in a short gap . . . . . . . . . 19
3.3 Streamer propagation in electric field . . . . . . . . . . . . . . . . . . . . . 22
4.1 Open air test set-up in the HV lab at Wits University . . . . . . . . . . . 26

4.2 Pressure vessel test set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3 Pressure vessel test set-up in the HV lab at Wits University . . . . . . . . 27
4.4 Test site for sea level tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.5 Experimental set-up for sea level tests . . . . . . . . . . . . . . . . . . . . 28
4.6 400 kV DC Generator at WITS . . . . . . . . . . . . . . . . . . . . . . . . 28
4.7 500 kV DC Generator at UKZN . . . . . . . . . . . . . . . . . . . . . . . 29
4.8 Picture of pressure vessel used in the experiment . . . . . . . . . . . . . . 30
5.1 Sea level test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

5.2 Low altitude (130 m asl)rod-plane breakdown voltages showing scatter . . 35
5.3 Comparison of low altitude test results and predicted results . . . . . . . 36
5.4 High altitude (1 740 m asl)rod-plane breakdown voltages showing scatter 38
5.5 Comparison of high altitude test results and predicted(Calva) results . . . 39
5.6 Comparison of IEC 60060-1 corrected low altitude and high altitude test
results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.7 Comparison of IEC 60060-1 corrected low altitude and high altitude test
results to sea level results . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.8 Results of breakdown voltage against relative air density in a variable
pressure vessel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.9 Variation of obtained breakdown voltages from the mean for 0.1 m gap . . 44
5.12 Probability of ionisation vs Pressure . . . . . . . . . . . . . . . . . . . . . 47
5.13 IEC 60060-1 correction of 0.1 m air gap results . . . . . . . . . . . . . . . 50
5.16 Comparison of experimental results with Calva’s prediction method for
0.1 m air gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
0.2 m air gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
vi
List of Figures vii

0.3 m air gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.19 Difference function for 0.1 m pressure vessel experiment . . . . . . . . . . 57
5.20 Comparison of prediction including difference method for 0.1 m gap length 58
List of Tables
4.1 Variation of temperature due to pressure in the chamber . . . . . . . . . . 31
5.1 Environmental conditions during sea level experiments . . . . . . . . . . . 34

5.2 Environmental conditions during low altitude tests . . . . . . . . . . . . . 35
5.3 Comparison of test results and Calva prediction . . . . . . . . . . . . . . . 36
5.4 Environmental conditions the high altitude tests . . . . . . . . . . . . . . 37
5.5 Comparison of test results to the Calva prediction at 1 740 m asl . . . . . 38
5.6 Comparison of high altitude and low altitude IEC 60060-1 (2010) cor-
rected results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.7 Table of δ × k values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.8 Approximate altitudes simulated during the tests . . . . . . . . . . . . . . 42
5.9 Pressure Vessel Test Results . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.10 Table showing the average electric fields during breakdown . . . . . . . . 52
5.11 Table showing the calculated g-factors . . . . . . . . . . . . . . . . . . . . 54
5.12 Pressure Vessel Test Results Compared to Calva Prediction . . . . . . . . 55
5.13 Pressure vessel 0.1 m gap test information . . . . . . . . . . . . . . . . . . 57
5.14 Error of mapped data to experimental data for 0.1 m air gap . . . . . . . 58
A.1 Low Altitude Experimental Data . . . . . . . . . . . . . . . . . . . . . . . 69

A.2 High Altitude Experimental Data . . . . . . . . . . . . . . . . . . . . . . . 69
A.3 Sea Level Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
A.4 Pressure Vessel Test Experimental Data . . . . . . . . . . . . . . . . . . . 71
viii
Abbreviations
asl Above Sea Level

Cigré International Council on Large Electric Systems
DC Direct Current
HVDC High Voltage Direct Current
IEC International Electrotechnical Commission
RAD Relative Air Density
WG Working Group
ix
Chapter 1
Introduction
The development of high voltage power system schemes has significantly improved with
higher voltage levels being used and longer distances being covered by the transmis-
sion lines. As these longer distances are achieved, high voltage direct current (HVDC)
becomes more and more favourable and necessary given that load centres are located
further away from the source [1]. As the transmission lines traverse to the load centres,
they pass through different altitudes. This has major implications on the insulation
coordination of the line as is evidenced in countries like South Africa where the altitude
varies from sea level to approximately 3 450 m (peaks along the Drakensburg range) [2].
By so doing, much attention is required on the selection of appropriate insulators and
air gap clearances for the transmission lines.
It is already known that changes in environmental conditions (temperature, humidity

and air pressure) affect the dielectric strength of an air gap, and these conditions change
with altitude. Standards have been put in place to facilitate the correction of voltages
to standard conditions (20◦ C temperature, 11 g/m3 humidity and 1013 mbar pressure).
This helps in the standardisation of equipment and derivation of rating/de-rating factors
to be applied to the equipment and calculation of clearances for high voltage structures
if used at different altitudes.
The standardisation procedure used for voltage correction however does not apply to
data obtained at altitudes above 2 000 m, especially for DC voltages [3]. This problem
has led to the International Council on Large Electric Systems (Cigré) creating a study
1
Chapter 1. Introduction 2
Working Group (WG D1.50) to develop new methods for altitude correction for air gaps
and equipment for installation at altitudes above 6 000 m above sea level (asl).
In combating the same problem, China has set up an HVDC research facility in Tibet, at
an approximate altitude of 4 400 m. It is at this facility that most of the high altitude
data for the Cigré WG D1.50 is obtained. In South Africa, possibilities of increased
HVDC transmission are a reality. Plans have also been put in place for the country to
draw power from the Congo and this will also require the use of HVDC transmission
for it to be feasible [1]. In order to develop the HVDC transmission schemes within
South Africa, a lot of research is being carried out as the initial step towards these
developments. Among some of the research topics proposed in their initiative towards
strategic research into HVDC transmission, Eskom proposed that research be conducted
on the insulation strengths of basic air gaps in the altitude range of 1 200 m to 1 600 m
~[4]. Some work has been done in South Africa under the proposed strategic research
[5].
This dissertation presents the investigation carried out on the effects of altitude (specif-
ically relative air density) on the HVDC breakdown of small rod-plane air gaps. The
research details explanations of the observations noted during the experimental work
along with a critical analysis of the existing correction methods.
The structure of this dissertation is as follows:
• Chapter 2 presents the evolution of the atmospheric correction methods. This

chapter goes through the history of and the research taken to develop atmospheric
correction procedures.
• Chapter 3 is a theoretical section on the mechanisms of the breakdown processes

of an air gap.
• Chapter 4 describes the experimental set up in the laboratory and the procedure
followed.
• Chapter 5 presents the results obtained and a discussion and analysis of the
results.
• Chapter 6 concludes the dissertation and presents recommendations for future

work.
Chapter 1. Introduction 3
• Appendix A contains the tables of all the experimental test data.
• Appendix B is a copy of a discussion paper presented at the SAUPEC 2015 by

the author. The paper constitutes a significant part of Chapter 2.
• Appendix C is a copy of a paper published in the proceedings of the International

Symposium on High Voltage Engineering (ISH) in 2015. Part of the findings
highlighted in this paper are portions of Chapter 5.
• Appendix D is a report submitted as a contribution to the Cigré WG D1.50 task

force on short air gaps.
Chapter 2
Historical evolution of Altitude

Correction Methods
Opening Summary
Work to create a way of correcting voltages for insulators began as far back as 1914 when
Peek investigated altitude effects on flashover voltages of insulators [6]. Ever since, a
lot of work has been done to develop a standard procedure of atmospheric correction.
This chapter looks at the evolution of knowledge in the creation of the commonly used
IEC 60060-1(2010) and also critically analyses how atmospheric correction is done in
the commonly used international standards. The knowledge gaps are identified and the
research questions of the present work are developed. A summary of the development
of the IEC 60060-1 standard is given in Figure 2.1. Further work has been done after
the development of the IEC 60060-1 standard and the work is also highlighted in this
chapter.
2.1 Peek Model
The effect of altitude on the breakdown voltage of insulators was noted as early as 1914
by Peek [6]. A significant amount of research has been done since then with regard to
air density effects on the breakdown voltage of air gaps and external insulation, and
4
Chapter 2. Historical evolution of Altitude Correction Methods 5
Peek (1914) investigates the variation of flashover voltage with

altitude. Concludes that the flashover voltage of an insulator at a
higher altitude is directly proportional to the breakdown voltage of
the same insulator under standard conditions with the factor of
proportionality being the air density. Also concludes that air density
changes due to pressure and those due to temperature have the same
effect on the breakdown voltage
IEC 60-1 (1960) is amended to IEC 60-1 (1973). Factors m and n are
introduced as indices that represent the variations of pressure and
temperature respectively. The expression for m shows that m
decreases proportionally with the gap length and is not related to the
air density or the type of voltage used. Shows input of Peeks
research that air density change due to pressure or temperature have
the same effect, therefore m and n can be equal.
Research by Geldenhuys and Feser and Schmidt [15] suggest that the
change in air density due to pressure change and that due to
temperature change have different effects therefore temperature and
air density should be corrected separately. Geldenhuys derives an
expression for the electric field required for streamer propagation
and concludes that under standard conditions, the electric field
required for streamer propagation is 500 kV/m
Pigini et al (1985) introduce the G-factor that determines the value of

n, a factor that represents the variation of the relative air density. The
G-factor is dependent on the length of the discharge, the air density
and the electric field required for streamer propagation. The obtained
expressions give much more accurate results compared to the
previous models such that it is incorporated into the 1989 version of
IEC 60-1, IEC 60060-1 (1989). The same model is later slightly
modified by giving equations in the place of graphs in the latest
standard; IEC 60060-1(2010)
Figure 2.1: Summary of the development of altitude correction methods in IEC

60060-1(2010)
more recently air gaps under DC voltage stress [5, 7–12]. Peek [6] investigated the
effect of altitude on the spark-over voltages of bushings and insulators by placing them
in a wooden cask and conducting breakdown tests under variable pressure. Pressure
was varied to obtain relative air densities between 0.5 and 1.0 while the temperature
variations due to the pressure changes were recorded. In his research, Peek [6] was able
to produce tables of correction factors to be used on specific insulators by reading off the
relative air density. From the obtained results, the correction factors for insulators were
different from those of bushings and leads under the same relative air density [6]. Peek
attributed these to the uniformity of the field suggesting that the flashover voltage for a
uniform field decreases directly with relative air density (RAD), whereas for non-uniform
fields, the flashover voltage decreases at a smaller rate than the RAD [6]. Peek therefore
concluded that, given that the flashover voltage of an insulator at sea level (δ=1) is v1 ,
then the breakdown voltage v at δ=x was given by
v = xv1 (2.1)
where δ is the relative air density at a specific altitude and read from the tables derived
from empirical test results by Peek [6].
Peek’s method for atmospheric correction did not incorporate vital factors to be consid-
ered when it comes to the breakdown of air gaps. The method did not include factors
like the length of the discharge path and the variation of the electric field under different
environmental conditions and different voltage types. Therefore, this method was not
robust enough to be used as the standard atmospheric correction procedure.
2.2 IEC 60-1 (1973)
In 1973, the existing altitude correction standard then, IEC 60-1, was amended such that
the breakdown voltage of an air gap or insulator at relative air density δ, uδ is related
to the breakdown voltage of the same air gap or insulator under standard conditions, u0
by [13]
kd
uδ = u0 (2.2)
kh
where kd is the air density correction factor defined as
P m T0 n
kd = ( ) ( ) (2.3)
P0 T
and kh is the humidity correction factor defined as
kh = k w (2.4)
In equation 2.3, P and P0 are the pressure and standard pressure, in mbar, respectively
while T and T0 are the temperature and standard temperature, in ◦ C, respectively.
In equation 2.4, k is the humidity correction factor given in a graph as a function of
absolute humidity and w is an exponent that defines the humidity factor depending on
the test type voltage and electrode arrangement [14].
The exponents m and n are the variations of pressure and temperature respectively that
are supposed to be determined before kd is evaluated. Since the air density is dependent
on pressure, the determination of the exponent m became the area of concern regarding
atmospheric corrections. The exponent m was defined in IEC 60-1 (1973) as



 1.0 for d ≤ 1,

m= −0.12d + 1.12 for 1<d<6, (2.5)



 0.4 for d ≥ 6.
where d is the air gap length. The factor m in this standard implied that pressure
variation was linearly dependent on the length of the air gap. Compared to the model
proposed by Peek, the model in the IEC 60-1 does not give a linear relationship between
the flashover voltage and the relative air density. However, the standard suggests that
the effect of pressure varies at a quicker rate with the increase in gap length. For short
gaps (less than 1 m long) and large gaps (above 6 m), m is considered constant and
the gap sizes between 1 m and 6 m are where linearity was observed for m. For all
voltage types of either polarity, the standard specifies that m and n can be considered
equal despite the geometry of the electrodes. Some of the electrode geometries for some
voltage types have the exponents m and n set to a value of 1.0 while other geometries
refer to equation 2.5. Some geometries did not have enough data when the standard
was put together, therefore m and n were set as 0. Three electrode geometries (sphere-
sphere, rod-rod and rod-plane) were used in the standard. If any electrode geometry
besides the three geometries mentioned earlier was used, m and n were set to 1.0 and
there was no humidity correction to be applied.
The evaluation of m, n, w and the restriction of the correction method to specific

electrode arrangements made the standard undergo revision and new correction methods
were to be introduced. This led to the development of the IEC 60060-1(1989) discussed
in the next section.
2.3 IEC 60060-1:1989(Pigini Model)
During his research, Geldenhuys [15], was able to derive the approximate expression of
the average electric field required for positive streamer propagation as given in equation
(2.6).
Es+ = 425δ 1.5 + (4 + 5δ)H (2.6)
where H is the absolute humidity and δ is the relative air density. By varying the
humidity, the average electric field required for streamer propagation under standard
conditions was determined to be 500 kV/m. The findings achieved by Geldenhuys [15]
gave way for the determination of the breakdown voltage of an air gap in terms of the
electric field of the leader zone of the discharge (El ), the length of the leader (ll ), the
electric field of the streamer zone (Es ) and the length of the streamer (ls ) as given by
equation (2.7) [16].
U = El ll + Es ls (2.7)
If breakdown occurs purely due to the streamer mechanisms (without any leader for-
mation), the leader terms El and ll are set to zero leaving the breakdown voltage to be
dependent only on the streamer field Es and streamer length ls . With this information,
Pigini et al [16] were able to derive a factor G to be used to determine the index of the
relative air density correction factor by incorporating the discharge mechanism used and
the relative air density. To correct the voltage, a similar expression to equation (2.2)
was used but modified as follows,
U50 = U50(std) (δ.K)n (2.8)
where K is the humidity correction factor graphically determined, n is the air density
correction factor and U50(std) and U50 are the 50% probability breakdown voltage under
standard and non-standard conditions respectively. To determine n in equation (2.8),
the G factor, equation (2.10), was used and n took the following values [16] depending
on the range of values of G

G(G−0.2)

0.8 for 0.3 ≤ G ≤ 1,
n= (2.9)
3−G

2G for 1 ≤ G ≤ 2,
where G is the G factor under standard conditions (δ and k =1). The factor G was defined
as in equation 2.10 and was determined on the assumption that the major process during
the breakdown of the gap was streamers [16].
U50
G= (2.10)
500δkd
where d is the length of the discharge gap in metres and k is a factor dependent on the
type of voltage used and δ is defined as:

P 273 + t0
δ= (2.11)
P0 273 + t
where t0 and t are the standard temperature and actual temperature in ◦ C respectively
and P0 and P are the standard pressure and actual pressure in mbar respectively. The
results by [16] were incorporated into the IEC 60060-1 (1989), including the graphs of
equation (2.9) as the air density correction factors [17]. The final correction procedure
in IEC 60060-1(1989) was as follows:
U
U0 = (2.12)
kt
where U is the non-standard breakdown voltage in kV, U0 is the standard breakdown

voltage in kV and kt is the atmospheric correction factor defined as
kt = k1 k2 (2.13)
where k1 is the air density correction factor and k2 is the humidity correction factor.
The air density correction factor is defined as
k1 = δ m (2.14)
where δ is defined according to equation 2.11. The humidity correction factor is defined
as
k2 = k w (2.15)
where k is a parameter that takes into account the type of test voltage used and is
graphically defined for each test voltage as a function of the ratio of the absolute test
humidity to the relative air density. It is at this point that the type of voltage is taken
into consideration during the correction procedure. The G factor is then calculated as
U
g= (2.16)
500Lδk
where U is as earlier defined, L is the length of the air gap in metres, δ is defined in
equation 2.11 and k is the parameter described above.
The exponents m and w in equations 2.14 and 2.15 respectively were determined using
the value of g. Once g was determined, the values of m and w were then read of a graph
that relates g to m and w. Thereafter, k1 and k2 can be determined and the standard
voltage U0 can be calculated.
The procedure in the IEC 60060 (1989) standard still plays a major role in the recent
IEC 60060-1(2010) which, instead of using graphs for determining m and w as in IEC
60060-1(1989), uses equations that are similar to equations (2.9) [3]. It is important to
note that the model derived by Pigini et al takes into account the voltage type through
the factor k in equation (2.10) whereas IEC 60-1 (1973) did not include any factor that
took into account the type of voltage used. Since the standard was produced from the
work conducted by Pigini et al, the standard is limited to an altitude of 1 800 m. Given
that the only difference in the altitude correction procedures highlighted in the current
version of IEC 60060-1 (2010) with those of IEC 60060-1 (1989) is the use of equations
to determine the index n instead of graphs, it therefore holds that the current version
of IEC 60060-1 is also limited to an altitude of 1 800 m.
Due to the altitude limit of the work done by Pigini et al, extra work by Ramirez [18]
was done in an attempt to modify the procedure to cover altitudes up to 3 000 m asl.
This results of the work are described in the following section.
2.4 Ramirez Model
A year after IEC 60060-1 (1989) was published, Ramirez and his counterparts [18] pro-
posed a model claimed to be more accurate than the one incorporated in IEC 60060-
1(1989) and takes account of the influence of air density on rod-plane gaps. The model
is expressed in equation (2.17)
U 0.8[1 + T (1 − δ)](δ − 0.2G00 )

= + 0.2 (2.17)
U0 (1 − 0.2G00 )
where δ is the relative air density calculated using equation (2.11) and
1 − 0.8G00
T = 1.4 (2.18)
1 − 0.2G00
with G0 0 defined as
U0
G00 = 1+(F0 −1)
(2.19)
500[ 3 D]
where F0 is the gap factor and D is the gap length. This proposed model satisfies
corrections for altitude up to 3000 m for positive impulse voltages [18].
The method proposed by Ramirez et al was not a full correction in that the effects of
humidity were not yet incorporated in the method. The derivation of the method was
done using the fundamentals used by Pigini et al and using positive impulse voltages.
Therefore, the correction procedure is questionable as there are a number of missing
factors which are critical in the breakdown process. The search for improved correction
factors continued with one of the notable efforts being contributed by Calva et al [9] as
discussed in the next section.
2.5 Calva Model
Calva et al [9] proposed a model to use for altitude correction for gaps under DC voltage
stress. This model determines the breakdown voltage Vb as
Vb = Eso d[k1 + k2 ]S (2.20)
where k1 is the air density correction factor
k1 = δ m (2.21)
and m=1.4 for positive polarity and 0.44 for negative polarity and δ is as in equation
2.11. k2 is the humidity correction factor
h − 11

−0.83
k2 = 1.3δ (2.22)
100
and S is the gap factor. Eso is the electric field required for streamer propagation and
considered to be 500 kV/m for positive polarity and for negative polarity is given by
equation (2.23)
Eso = 1476.4 × 1121.91d (2.23)
The models presented by Ramirez et al [18] and Calva et al [9] were not considered
as inputs to the latest version of the IEC 60060-1(2010) and it would therefore be of
interest to compare these models. It is now generally agreed that there are shortfalls
in the IEC 60060-1(2010) standard with regard to altitude correction factors. In that
regard, a Cigre work group (WG D1.50) “Atmospheric and altitude correction factors
for air gaps and clean insulators” has been established to coordinate further studies and
investigations on altitude correction. It has also been proven by [5, 10] that the IEC
60060-1 is still inconsistent in its calculated results compared to the experimental results
obtained as altitude increases for air gaps under DC voltage stress and that the model
proposed by Calva et al yields better accuracy in results than the IEC 60060-1(2010).
It has been observed that when the models by Ramirez, Pigini and Calva were derived,
there were inconsistencies in the type of voltage used. It is suspected that the incon-
sistency in the voltage type used to derive the models could be among the causes of
the variations in the models. IEC 60060-1 was derived from empirical results conducted
on switching and lightning impulse voltages only [15, 16, 19]. The model presented by
Ramirez was also derived using switching and lightning impulse voltages but the data
was limited to only the rod-plane electrode configuration and could not be extended
into a general method as the IEC 60060-1. Finally, the model presented by Calva et al
[9] was developed using DC voltage conditions. This procedure is currently unique in
literature and further work regarding this procedure is still under way.
2.6 Some Irregularities In The Commonly Used Standards
There is a fair number of standards used internationally that entail atmospheric correc-
tion voltages. The three notable standards widely used are the IEC 60060-1(2010) [3],
IEC 60071-2 (1996) [20] and IEEE Std 4 (2013) on high voltage testing techniques [14].
In as much as these standards are widely used, the standards have notable irregularities
regarding atmospheric corrections that are highlighted in the following subsections.
2.6.1 IEC 60060-1 (2010)
The most commonly used standard is the IEC 60060-1 (2010), largely based on Pigini
et al [16] and discussed previously in Section 2.3. The standard defines two methods of
atmospheric correction:
(i) The standard procedure; This procedure is used to correct voltages from non-
standard conditions to standard conditions using the relationship,
U
U0 = (2.24)
Kt
where U0 is the disruptive-discharge voltage under standard conditions, U is the

measured disruptive-discharge voltage under non-standard conditions and Kt is the
atmospheric correction factor [3]. The derivation of this procedure was covered in
Section 2.3.
(ii) The converse procedure; This is an iterative method used to predict breakdown
voltages at non-standard conditions using equation (2.24) applied on known break-
down voltage at standard atmospheric conditions.
The determination of Kt in [3] involves the calculation of the g-factor to determine the
exponents m (exponent n in Section 2.3) and w (exponent used for humidity correction).
The function that relates m to the g-factor is derived from a curve that averages the
results obtained during the work done by [16] of different gap geometries and positive
polarities of different impulse shapes [21]. This gives rise to the possibility of over-
correction or under-correction of the voltage [21]. With regards to negative polarity
impulses, the value m=1 was assumed by [16] and incorporated into the standard. This
implies that the value of m for negative impulses is independent of the air density. Since
the variation of air density has an effect on the breakdown voltage, the assumption of a
constant value of m is an approximation that cannot be considered accurate.
The relationship between m, w and g in [3] is only valid when δ × k ≈ 1 [22]. This
condition is valid for high voltage laboratories located at altitudes near sea level. For
laboratories at much higher altitudes however, δ × k < 1 [22] implying that the ap-
plication of the correction method in [3] becomes invalid for results obtained in such
laboratories.
For one to be able to use the g-factor, the U50 of the apparatus/gap is assumed to be
known. However, for high voltage apparatus, the type test withstand voltage (the spec-
ified prospective value that characterises an insulator [3]) is the one known and cannot
be assumed to be the U50 of the apparatus [22]. Therefore the g-factor calculated using
the type test withstand voltage can be invalid. Also, as mentioned earlier, the g-factor
is defined with the assumption that streamer propagation is the dominant discharge
mechanism and that the discharge is directly proportional to the relative air density.
Therefore for streamer dominant discharges, the g-factor remains accurate but cannot
be trusted to be accurate for other types of discharges especially in larger air gaps.
These irregularities make the IEC 60060-1 standard be deemed unreliable for correction
of voltages at altitudes greater than 1 800 m and for discharges that are not streamer
dominated.
2.6.2 IEC 60071-2 (1996)
The IEV 60071-2 (1996) standard was based on the same data used to develop IEC 60060-
1 and therefore by default, makes the standard to be valid only up to 1 800 m, although
it is used for higher altitudes [20]. The relationship between the required withstand
voltage (Urw ) and the co-ordination withstand voltage (Ucw ) is given by equation (2.25)
Urw = Ucw × Ka (2.25)
where K a is the atmospheric correction factor given by equation (2.26)
H
Ka = em( 8150 ) (2.26)
where H is the altitude in m. In this standard, the value of m is defined as 1 for lightning
impulses and short duration AC voltage tests. For switching impulses, the value of m
is taken from a plot of curves that relate m to (Ucw ). However, (Ucw ) is defined as
the 10% probability breakdown voltage therefore, when designing insulation for a lower
breakdown voltage probability, the value of (Ucw ) cannot be used [22]. The curves
relating m and (Ucw ) in the standard are for phase-to-earth insulation, longitudinal
insulation, phase-to-phase insulation and rod-plane gap. These cannot fully describe
the dielectric strength and and discharge characteristics of different gap configurations,
rod-to-rod for instance. Also, a value for m to be used for AC voltages is recommended
but no recommendation is given for DC voltage.
Even though this standard is assumed to be able to correct for altitudes greater than
1 800 m, there are irregularities in the determination of m and no recommendation for
DC voltages is given.
2.6.3 IEEE Std 4
The IEEE Std 4 standard has two methods for atmospheric correction [14]:
1. Method 1- This method uses the factor K and similar as the one used in IEC
60060-1
2. Method 2- This method uses the factors kd (air density correction factor) and kh
(humidity correction factor) for correction and is applicable to gaps less than 1 m.
Since Method 1 is similar to the one presented in IEC 60060-1, it will not be discussed
again in this section. Method 2 states the following relationship,
kd
U= U0 (2.27)
kh
where U0 and U are the standard voltage and non-standard voltage respectively.
The values of m and n used for evaluating kd are assumed in the standard to be the
same irrespective of the gap geometry and RAD . For other electrode configurations and
voltage types, the values of m, n and w are determined from a graph that relates the
three variables to the gap length [14]. The values of m, n and w will also be the same
(that is, m=n=w ) for the same gap length, configuration. This does not agree with
the variation of the air density correction factor with the type of discharge that governs
breakdown in the air gap.
Closing Summary
The atmospheric correction method stated in IEC 60060-1 (2010) is derived from data
based on empirical work conducted up to an altitude of 1 800 m. This makes the
method unsuitable for use when applied to altitudes higher than 1 800 m. Literature
has shown discrepancies in the factors m and w thereby suggesting changes in breakdown
mechanisms since the factors are dependent on the dominant breakdown mechanism. A
prediction method proposed by Calva et al for DC voltages has been discovered and will
be investigated. With the evolution history discussed in the previous pages, the problem
statement and the research question were put together as follows:
The Problem
As altitude increases, air density decreases and air breakdown mechanisms change. The
IEC 60060-1 (2010) standard has been found to cover only a limited altitude range.
Some models that can possibly extend the altitude correction factors to cover a higher
altitude range have been suggested in the literature but are not yet incorporated into
the standards. Very little knowledge is available in literature regarding the variation in
breakdown mechanisms under HVDC with altitude, leading to limited knowledge on the
correction methods. Also, the currently used standard, IEC 60060-1(2010) is built on
the empirical results of impulse tests over a limited altitude range. No empirical data
on AC and DC voltages was used towards the establishment of the standard. Further
verification of the proposed models is required, especially for DC voltages, and the
reliability of IEC 60060-1 (2010) needs to be tested.
Research Question
It is known that the breakdown voltage of an air gap is dependent on pressure. However,
the altitude correction factors given by the current version of IEC 60060-1 (2010) (based
on the model proposed by Pigini et al) are not reliable for higher altitudes. A procedure
proposed by Cavla et al, described in the Chapter 2, has not been incorporated into the
current standard even though they have been suggested to be more accurate at higher
altitudes. The following questions therefore arise:
“What are the similarities and differences between the different altitude correction models
(IEC 60060-1 (2010) and Calva procedure) when tested on the same experimental data
set?”
and
“Out of the two procedures, which one can be recommended for current use and is there
any possible modification that will give more accurate correction factors for wider altitude
ranges?”
By answering these questions, it is hoped that insight into a more informed approach
to altitude correction will be established. The first step towards understanding altitude
correction is a study into the electrical breakdown mechanisms in air. This is discussed
in the next chapter.
Chapter 3
Breakdown Mechanisms in Air
Opening Summary
The electrical breakdown of air is governed by processes that include corona, streamer
development and propagation, leader development and propagation and then the break-
down arc. The detailed explanation of each of these processes are laid out in the following
sections. The process is explained by considering the rod-plane geometry as it gives a
typical non-uniform field profile compared to other geometries. By studying the break-
down mechanisms, one can understand altitude correction factors from first principle
since environmental conditions (which strongly influence the breakdown mechanisms),
especially pressure, vary with altitude.
3.1 Non-Uniform Field Breakdown
Rod-plane electrode configurations are commonly used as they give electric field profiles
common in high voltage electric equipment. The voltage breakdown mechanism for
a rod-plane short gap under positive DC stress is dominated by the streamer process
[23, 24]. In a rod-plane gap with voltage applied, a divergent field is created between the
rod and the plane. This field is stronger around the tip of the rod and weaker towards
the plane as illustrated in Figure 3.1. When breakdown of a rod-plane air gap occurs,
the processes in Figure 3.2 occur. In larger gaps, streamer-leader transition can occur
18
Chapter 3. Breakdown Mechanisms in Air 19
Anode
Critical field line
Cathode
Figure 3.1: Electric field distribution in a rod-to-plane gap
Streamer Arc
Ionization Corona development and formation
propagation (breakdown)
Figure 3.2: Processes that occur leading to breakdown in a short gap
before arc formation and streamer propagation [25].
3.1.1 Ionization
The process in which a neutral molecule loses an electron thereby gaining a positive
net charge is called ionization [26]. The process of ionization is the first step towards
breakdown of air. Ionization can be achieved by the process of collision, photo-ionization
and secondary ionization [26].
Ionization by collision occurs when a free electron or metastable collides with a neutral
molecule causing the neutral molecule to release an electron. The process of dislodging
an electron from a neutral molecule by collision requires that the energy being transferred
during the collision be greater than the ionization energy of the molecule [26]. In the
case where the energy transferred during collision is greater than the ionization energy,
the process can be represented by equation (3.1).
e− + M → 2e− + M + (3.1)
where e− is an electron, M is the molecule and M+ is the produced positive ion. In

a given electric field, the ability of an electron to gain enough energy to dislodge an
electron from the molecule is dependent on the spacing between the molecules (mean
free path, τ ). Generally, further apart the molecules are, the more energy gained by the
electron during acceleration (because of the surrounding electric field), whereas if the
molecules are close together, there will not be enough space between the molecules for
the electrons to gain enough energy to dislodge an electron. Therefore, the ionization
of molecules is highly dependent on the “mean free path” between molecules. This
means that ionization by collision can be easily achieved in a gas under low pressure
compared to that under higher pressure. Ionization by collision is the most common
form of ionization processes in gases, therefore making the air breakdown mechanism
highly dependent on the air pressure.
Ionization of gases can also occur through the process of photo-ionization. This process
is achieved by the molecule or atom being able to absorb radiation energy that exceeds
the ionization energy of the molecule or atom [26]. Other secondary processes like
electron emission due to positive ion impact with the cathode, emission of electrons
due to photons and electron emission due to metastable and neutral atoms also occur
during the ionization process of gases [26]. For this research, ionization due to molecular
collisions is of most relevance as this process is directly linked to atmospheric pressure.
Electrical breakdown of a gas is through electron avalanches as reviewed in the next
section.
3.1.2 Electron Avalanche Mechanisms
If a free electron appears between the cathode and anode with a voltage applied, the
electron accelerates towards the anode due to the electric field. While the electron
accelerates towards the anode, it collides with molecules along the way and therefore
initiating ionization in accordance with equation (3.1). One collision adds a free electron
to the system, therefore as the initial electron propagates through the gap, the number
of electrons in the system increases exponentially causing an avalanche of electrons. The
number of electrons (nd ) reaching the anode a distance d from cathode can be calculated
by
R x=d
αdx
nd = n0 e x=0 (3.2)
where n0 is the initial number of electrons emitted from the cathode and α (Townsend’s
first ionization coefficient) is the average number of ionizing collisions made per electron
per unit distance travelled in the direction of the field [26]. The ions produced from
the process propagate towards the cathode. This leads to a charge separated system
with the electrons travelling towards the anode and ions towards the cathode. Since
electrons are lighter than the ions, the electrons get to travel a greater momentum
compared to the ions therefore reaching the anode before the ions reach the cathode.
The initial avalanche propagation towards the anode is mainly driven by the potential
difference between the electrodes. In the presence of secondary ionisation mechanisms,
the overall number of secondary electrons produced per incident positive ion, photon,
excited particle or metastable particle is a function of the secondary ionisation coefficient
(γ) [26]. As the voltage is increased, the avalanche will transform into a streamer. The
development and propagation of a streamer is discussed in the next section.
3.1.3 Streamer Development and Propagation
In a rod-plane electrode configuration, a non-uniform electrical field is produced in the

air gap. The maximum electric field (Emax ) for a gap length of d and rod tip radius r
can be calculated by equation (3.3) [27].
2V
Emax = 4d
(3.3)
rln(1 + r )
The non-uniformity of the field leads to the first ionization coefficient varying across the
gap [26, 28] resulting in the breakdown criterion as given in equation (3.5). In order
to initiate a streamer, the charge in the avalanche head should have reached a critical
value of
n0 eαxc ≈ 108 (3.4)
where xc is the length of the path taken by the avalanche in the direction of the electric
field. In order for breakdown to occur, the criterion in equation 3.5 should be met.
Z xc
αdx = lnNcr ≈ 18 − 20 (3.5)
0
where Ncr is the critical charge in the avalanche. For the criterion in equation (3.5) to be
met, it was empirically discovered that at standard atmospheric pressure, the electrical
field should be ≈2.5 kV/mm for a very small (few mm) streamer length within the range
ln(Ncr )≈9-21 [25]. Since the field will be non-uniform, streamer inception occurs in the
areas of the highest electric field. When the electric field strength reaches approximately
2.5 kV/mm the breakdown criterion is met and the air around the rod begins to break-
down. The breakdown of the air occurs around the rod tip and is usually accompanied
by a cracking noise and is known as corona [26, 28]. Since the field is non-uniform, the
electric field strength varies from a maximum at the rod tip to a minimum at the plane
electrode.
As the streamer propagates across the gap, the streamer head distorts the background
electric field thus producing a resultant electric field modified by the space charge.
The resultant electric field, partially, drives the process of self-sustained ionization and
growth of avalanches [25] as depicted in Figure 3.3. Having the resultant field at the
-
d
Induced electric field between anode
and streamer head
+ +
-
-
E Distorted resultant electrical field

(kV/m) due to the propagation of the
streamers
d (m) Background electrical field

due to voltage across the
electrodes
Figure 3.3: Streamer propagation in electric field
head of the streamer of approximately 2.5 kV/mm or more allows for the streamer to
be sustained. Due to the photo-ionization process that drives the streamer propagation,
the path which the streamer takes is probabilistic. Therefore a bundle of streamers can
be formed due to the creation of two differently directed avalanches as the streamer
propagates. When the main stem of the streamer has enough current to achieve a tem-
perature above 5000 K (4726.85 0 C), then the streamer transforms to a leader (if the
gap is more than 2 m long) [25]. Once the leader bridges the gap, it creates a conductive
path and the gap breaks down.
It is important to note that there are four possible outcomes when voltage is applied
across a rod-plane air gap. These outcomes are highly dependent on the applied voltage
and gap length.
1. The voltage is applied and the avalanches propagate towards the anode. However,
along the way the avalanches are unable to initiate the self propagation process
and the streamer breakdown criterion is not met. This scenario occurs if the gap
is weakly stressed therefore there will not be enough energy for the streamers to
propagate. In such a case, the field is so weak such that the streamers produced
do not cross the whole gap.
2. The streamers propagate across the gap and are able to lead to breakdown without
any leader formation. This usually happens to air gaps between 5 cm and 2 m
[25].
3. For gaps bigger than 2 m, streamers develop into leaders and breakdown occurs
through the leader.
Closing Summary
In order to understand and develop altitude correction methods, the study of breakdown
mechanisms is essential. it has been noted that the atmospheric pressure has great
influence on the ionisation process, which is the backbone of the electrical breakdown
process. In the next chapter, the experimental set-up and procedure used in the research
on the effect of altitude (pressure) on air gap breakdown voltage are introduced.
Chapter 4
Experimental Set-Up
Opening Summary
This chapter highlights the experimental set-up and procedures for the experiments
conducted to investigate the effect of altitude on the DC breakdown voltage of rod-
plane air gaps. Four sets of experiments were conducted and are categorised as open
air tests and pressure vessel tests. The open air tests were conducted at three different
altitude locations while the pressure vessel was used to simulate higher altitudes that
are not physically easily accessible. The following sections give detailed descriptions of
the set-ups and procedures followed.
4.1 The Tests
The work done in the laboratory included two types of tests; open air tests and the
pressure vessel tests.
4.1.1 Open Air Tests
The open air tests carried out in this project were conducted in three different locations.
These locations were selected according to a criteria whereby a high voltage DC supply
kit is available and the altitude of the locations is at either end of the altitude extremities
24
Chapter 4. Experimental Set-Up And Procedure 25
that are easily accessible in South Africa. The result of the selection criteria left two
possible locations, Wits University(≈ 1 800 m asl) and the University of KwaZulu-Natal
(UKZN) Westville Campus DC Centre (≈ 130 m asl). The third test was conducted
less than 2 m asl at Scottburgh beach, Clansthal. These tests were conducted in order
to obtain results closest to the standard breakdown voltage. These results are used for
comparison with the results from the high altitude and low altitude tests.
By conducting the tests at the above mentioned locations, data was captured at both
low altitude and high latitude. The data was then used in producing standard condition
breakdown voltages by being corrected according to IEC 60060-1 and was also compared
to the prediction method proposed by Calva et al [9]. For higher altitudes greater than
that at Wits University, the tests were conducted in a variable pressure vessel.
4.1.2 Pressure Vessel Tests
The pressure vessel tests were conducted at Wits University using a cylindrical pressure
vessel borrowed from the UKZN. Due to the locking mechanism of the top cover of the
vessel, air was only drawn out of the vessel to simulate high altitudes (low pressure) since
pumping air in (high pressure) would cause structural damage to the vessel. Further
physical details of the pressure vessel are presented in the next section. A maximum of
300 mbar was able to be drawn from the ambient pressure within the chamber. This
resulted in a relative air density (RAD) of 0.53, corresponding to an altitude of ≈ 5200
m. The results obtained from this test were compared to the prediction method proposed
by Calva et al [9] and were also used to evaluate the applicability of the IEC 60060-1
standard on altitudes greater than 1 800 m asl (RADs lower than 0.8).
4.2 Laboratory test configuration and Equipment
4.2.1 Test configurations
The same set-up in Figure 4.1 was used in the tests in UKZN. For the sake of consistency,
the exact same rod was also used in the tests in UKZN and at sea level.
HVDC
Generator
Rod
Plane
Figure 4.1: Open air test set-up in the HV lab at Wits University
Figure 4.2: Pressure vessel test set-up
4.2.2 Equipment Used in the Tests
The DC Generators
The two generators used for the tests are shown in Figures 4.6 and 4.7.
HVDC
Generator
Rod
Input
Transformer
Pressure
Vessel
Figure 4.3: Pressure vessel test set-up in the HV lab at Wits University
Test Sea
Site
Figure 4.4: Test site for sea level tests
The generator at Wits University (Figure 4.6) is a 400 kV rated 3 stage Cockroft-
Walton DC unit which can reach a maximum of 325 kV due to high altitude effects.
The generator at UKZN (Figure 4.7) is a 2 stage Cockroft-Walton DC unit rated at 500
kV. The generator can only operate at a maximum voltage of approximately 250 kV
due to clearance limitations in the lab, as can be seen in Figure 4.7. The generators
HVDC
Structural
Generator
support
Water
Resistor
Rod
Plane
Figure 4.5: Experimental set-up for sea level tests
Figure 4.6: 400 kV DC Generator at WITS
are equipped with a stack of resistors at the output end. These resistors form a voltage
divider and a panel on the control board displays the output voltage read from the
voltage divider. The output of the divider was compared to an already resistive potential
divider calibrated at NETFA and the readings fell within a 2% error. The voltage divider
Figure 4.7: 500 kV DC Generator at UKZN
on the generator was therefore deemed accurately calibrated and usable for the project.
Rod and Plane Electrode Set-Up
A rod-plane electrode configuration was used for the tests conducted because this ge-
ometry is widely used in literature and the results can be comparable to other similar
work done in South Africa [5]. For the open air tests, a copper rod 1.5 m long and with
a 15 mm diameter was used as the rod electrode. The rod dimensions and electrode set
up is commonly used in other research centres around the world that are taking part in
the round robin tests for the Cigré D1.50 working group. The end of the rod was fitted
with an end cap that was smoothly soldered for contact purposes and to avoid electric
field distortions around the end cap. The rod was suspended from the roof by a nylon
rope pulley system in order to maintain clearances. A 1.8 m by 2 m steel plate was used
as the grounded plane.
The rod used in the pressure vessel experiment was a 0.7 m aluminium rod 15 mm in
diameter. The rod was machined at the Wits Genmin laboratory to have the same tip
profile as the copper rod used in the open air experiments. A circular plate 112 mm in
diameter was machined in the Genmin laboratory and was used as the ground plane for
the experiment.
The Pressure Vessel
The pressure vessel used in the experiment was a 0.55 m long cylindrical tube with a
0.12 m inner radius borate glass wall. A picture of the pressure vessel is shown in Figure
4.8. The pressure vessel had two valves that were used to control the inlet and outlet of
Aluminium
Rod
Ground
Plate Pressure
Vessel
Pressure
Gauge
Figure 4.8: Picture of pressure vessel used in the experiment
the air. A new top cap for the chamber was machined in the Genmin laboratory. The
top cap was designed such that the rod electrode can go through the cap. A clamping
mechanism to hold the rod in place after setting the gap length was designed on the top
cap. The design of the top cap came with the constraint that the pressure vessel could
only be used under ambient and low pressure conditions. If pressure was increased, the
top cap would have been blown off.
Furthermore, due to the dimensions of the chamber and the design of the top cap,
only gap sizes up to 0.3 m could be tested. These two limitations meant that only
high altitude RADs could be simulated. In order to verify whether the gas laws could
affect the experiment due to drawing pressure, a pilot test was carried out. By drawing
pressure in the vessel with a digital thermometer inside, the temperatures in Table 4.1
were recorded. The external temperature at the beginning of the test was the same as the
Table 4.1: Variation of temperature due to pressure in the chamber
Pressure Temperature (0 C)
Ambient 22.2
Ambient-100 mbar 22.0
initial temperature inside the vessel. At the end of the test, the external temperature was
22.4 0 C. Therefore, the temperature variations inside the chamber were not significant
during the pressure vessel tests as the temperature was approximately the same as
outside the chamber. Therefore all temperature readings during the tests were taken
outside the chamber.
4.3 Test Procedure
4.3.1 Open Air Experiments
The following procedure was used in conducting the open air tests:
1. Set gap length to 0.1 m
2. Record temperature, pressure and humidity
3. Slowly increase the voltage at approximately 3 kV/sec until breakdown occurs

and record the breakdown voltage. Breakdown would have occurred when an arc
completely bridges the air gap.
4. Observe 5 minute waiting period to allow dissipation of any accumulated space

charge
5. Repeat 3 and 4 until 5 breakdowns have been recorded. 5 breakdowns were used
because it was discovered that the breakdown of an air gap under HVDC stress is
repeatable and there is not much scatter in the results.
6. Increase gap length by 0.05 m
7. Repeat 2 to 6 until readings for 0.4 m air gap have been recorded.
4.3.2 Pressure Vessel Experiments
The pressure vessel tests were similar to those of the open air experiments. Due to
limitations in the size of the pressure vessel, gap sizes that were tested were 0.1 m, 0.2 m
and 0.3 m. Starting at ambient pressure, a set of 5 breakdowns were recorded for each
gap length per pressure level. The pressure was decreased in the chamber by 100 mbar
using an EdwardsTM pump once 5 breakdowns were recorded for the current pressure
level. Tests were done until tests on a pressure level 300 mbar below the ambient were
conducted. Due to the confined space in the chamber, a 10 minute waiting period
was observed between breakdowns to allow adequate dispersion of accumulated space
charge. Environmental conditions were recorded for each gap length and each pressure
level tested.
In the next chapter, the test results, analysis and discussion are presented.
Chapter 5
Experimental Results, Analysis

and Discussion
Opening Summary
This chapter lays out the results obtained from the tests conducted at high altitude, low
altitude, sea level and in the variable pressure vessel. All experiments were conducted
using a rod-plane air gap with a 15 mm flat tip diameter rod. All the results, except
the sea level results, were corrected according to IEC 60060-1 (2010) and were also
compared to predictive calculations using the Calva prediction method. Corrected IEC
60060-1(2010) results were compared to the sea level results as they should ideally be
similar. Trends in the results were analysed and functions were derived to modify the
Calva model for improved accuracy.
5.1 Open Air Gap Breakdown Experiments
5.1.1 Sea Level Tests
The results of the sea level tests are shown in Figure 5.1 (average of 5 breakdowns per
air gap length) and Table 5.1 shows the environmental conditions during the tests.
33
Chapter 5. Experimental Results and Discussion 34
Table 5.1: Environmental conditions during sea level experiments
Gap length Temp. Hum. Press. Breakdown Std

(m) (◦ C) (%) (mbar) Voltage(kV) Deviation(%)
0.10 23.6 89 1012.4 63.6 0.8
0.15 23.5 89 1012.4 87.4 0.6
0.20 23.4 90 1012.5 112.0 0.6
0.25 23.4 90 1012.6 139.8 0.3
0.30 23.3 92 1012.8 164.2 0.2
0.40 23.3 92 1012.7 218.0 0.6
220
Mean
200 Maximum
Minimum
180
Breakdown Voltage (kV)
160
140
120
100
80
0.1 0.15 0.2 0.25 0.3 0.35 0.4

Gap length (m)
Figure 5.1: Sea level test results
Figure 5.1 shows the deviation of the results from the mean. The results had a maximum
deviation of 1% from the mean. The small variation shows that the results obtained
were precise and instils confidence that they are accurate. The environmental conditions
during the tests were the closest to the defined standard conditions in IEC 60060-1,
therefore the results will be used as the reference standard breakdown voltages.
5.1.2 Low Altitude Tests (UKZN) (≈ 130 m asl)
The results obtained during the tests at UKZN are in Figure 5.2 and the recorded
environmental conditions during the test are in Table 5.2.
Figure 5.2 is a plot of the average voltage of 5 breakdowns per each gap length and
shows how the breakdown voltages per each gap length varied in range and how the
maximum and minimum breakdown voltages vary from the mean.
Table 5.2: Environmental conditions during low altitude tests

(m) (0 C) (%) (mbar) Voltage (kV) Deviation (%)
0.10 27.6 49 987.8 56.9 1.2
0.15 26.2 56 986.3 82.2 0.6
0.20 27.8 52 986.1 107.6 0.2
0.25 26.5 55 985.4 131.5 0.1
0.30 27.2 55 985.2 157.1 0.3
0.35 21.9 56 985.8 174.9 0.1
0.40 25.5 46 985.5 196.8 0.1
Mean
180 Maximum
Minimum
160
140
120
100
80
60
0.1 0.15 0.2 0.25 0.3 0.35 0.4
Gap length (m)
Figure 5.2: Low altitude (130 m asl)rod-plane breakdown voltages showing scatter
A maximum deviation of 2.4% from the mean breakdown voltage was observed for the
entire data set. This occurred at the 0.1 m air gap length results. Therefore, the rest of
the 5 individual breakdown voltages obtained per each gap length falls within 2.4% of
the mean breakdown voltage of each gap length. This is an acceptable error value and
therefore the data set gives high confidence in its accuracy.
The low altitude results are plotted against the predicted results using the Calva method
in equation 5.1 where d is the gap length, k1 and k2 are defined in equations
Vb = 500d(k1 + k2 )S (5.1)
5.3 and 5.3 respectively and S is the gap factor and is dependent on the rod electrode
geometry and polarity [9]. S was defined as 1.01d−0.10 and in equation 5.3, m was
defined as 1.4 for positive polarity [9].
k1 = δ m (5.2)
h − 11

−0.83
k2 = (1.3δ ) (5.3)
100
Table 5.3: Comparison of test results and Calva prediction
Gap size (m) Breakdown Calva (%) Difference

Voltage (kV) Prediction (kV)
0.10 56.9 59.0 3.7
0.15 82.2 86.3 5.0
0.20 107.6 111.1 3.2
0.25 131.5 136.3 3.6
0.30 157.1 161.0 2.5
0.35 174.9 184.2 5.3
0.40 196.8 202.1 2.7
Figure 5.3 shows the plot of the comparison of the obtained low altitude experimental
results and the predicted results while Table 5.3 shows the tabulated data. The plot in
220
Experimental
200 Predicted
180
160
140
120
100
80
60
40
0.1 0.15 0.2 0.25 0.3 0.35 0.4
Gap Length (m)
Figure 5.3: Comparison of low altitude test results and predicted results
Figure 5.3 shows that the prediction method proposed by Calva provides fairly accurate
results since the average error for the low altitude test results is 3.7%. The maximum
error when considering individual air gaps is 5.3% observed for the 0.35 m air gap. By
observing the trend of the predicted and experimental results, it can be noted that the
Calva prediction method over estimates the breakdown voltage by an average of 3.7%
(obtained error). This error is advantageous when considering clearances during design
as it will naturally add in an extra safety factor. It is deduced that the Calva prediction
method is therefore reasonably accurate for gaps in the range 0.1 m - 0.4 m at an altitude
of ≈ 200 m asl.
5.1.3 High Altitude Tests (Wits University)(≈ 1 740 m asl)
High altitude tests were conducted at Wits University at an altitude of ≈1 740 m asl.
The corresponding relative air density at this altitude is 0.82 and is close to the IEC
60060-1 limit of δ=0.8. The plot of the results is shown in Figure 5.4. Table 5.4 presents
the conditions under which the results were obtained.
Table 5.4: Environmental conditions the high altitude tests

(m) (0 C) (%) (mbar) Voltage (kV) Deviation (%)
0.10 15.3 41 844.7 52.5 1.9
0.15 14.4 45 841.7 67.8 1.1
0.20 15.4 34 840.2 81.2 0.3
0.25 14.9 42 841.1 100.2 0.4
0.30 15.1 46 842.3 118.2 0.7
0.35 15.8 33 839.7 129.9 0.3
0.40 14.8 47 842.3 156.8 0.2
The results shown in Figure 5.4 are an average of 5 breakdowns per each gap length.
Figure 5.4 also shows the breakdown range observed for each air gap and how they vary
from the mean breakdown voltage. Out of this set of data, a maximum variation of
3.2% was observed between the maximum breakdown voltage and the mean breakdown
voltage during the 0.1 m test. With this, the results pose high confidence in their
accuracy.
As done for the low altitude test results, the plot for the results of the high altitude test
compared to the predicted results is shown in Figure 5.5. Table 5.5 shows the tabulated
data for Figure 5.5.
Unlike the low altitude results, it is observed that the Calva prediction method under-
predicts the breakdown voltage of the 0.1 m and 0.15 m air gaps, then over-predicts (as
expected) for the bigger air gaps. This trend is approximately the same trend observed
Mean
150
Maximum
140 Minimum
Breakdown Voltage (kV) 130
120
110
100
90
80
70
60
0.1 0.15 0.2 0.25 0.3 0.35 0.4

Gap length (m)
Figure 5.4: High altitude (1 740 m asl)rod-plane breakdown voltages showing scatter
Table 5.5: Comparison of test results to the Calva prediction at 1 740 m asl
Gap size (m) Breakdown Calva (%) Difference

Voltage (kV) Prediction (kV)
0.10 52.5 46.3 11.8
0.15 67.75 67.0 1.1
0.20 81.18 83.7 3.1
0.25 100.15 105.0 4.8
0.30 118.2 126.8 7.3
0.35 129.9 134.8 3.8
0.40 156.8 162.2 3.4
by [5] at an altitude of 1 880 m asl. However, the crossing point of their results and the
Calva prediction method occurs between 0.3 m and 0.4 m, whereas in Figure 5.5 the
crossing point occurs at an approximate 0.19 m gap length. This trend seems to suggest
that as the altitude increases, the Calva prediction method under-predicts for smaller
gaps and as the altitude increases, the method over-predicts.
However, the comparison of the results show that the Calva prediction is reasonably
accurate as the average error between the two sets of data is 5.0%. This error falls
within the generally allowed 5% error, therefore the Calva prediction is proven to be
reasonably accurate at an altitude of 1 740 m asl for air gaps ranging in size between
0.1 m to 0.4 m.
180
Predicted
Experimental
160

140
120
100
80
60
40
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Gap Length (m)
Figure 5.5: Comparison of high altitude test results and predicted(Calva) results
5.1.4 IEC 60060-1 Correction of Low Altitude and High Altitude Re-
sults
Since IEC 60060-1 standard correction method serves to correct voltages to standard
conditions, both sets of data were corrected according to the standardised method. The
IEC 60060-1 standard has been described in Chapter 2. Table 5.6 shows the tabulated
comparison of the high altitude and low altitude corrected results.
Table 5.6: Comparison of high altitude and low altitude IEC 60060-1 (2010) corrected
results
Gap Length High Altitude Low Altitude %

(m) Corrected Voltage (kV) Corrected Voltage (kV) Difference
0.10 64.4 59.6 8.1
0.15 82.8 84.9 2.5
0.20 102.8 111.6 7.9
0.25 122.9 136.1 9.7
0.30 140.3 162.0 13.4
0.35 163.0 181.9 10.4
0.40 187.1 210.2 11.0
The plot of the corrected data is shown in Figure 5.6. IEC 60060-1 states that the
correction methods are applicable and accurate when applied to results obtained between
0.8<δ<1. Therefore, if the same electrodes are used to conduct experiments for the same
gap lengths in different altitude locations within the δ range stated in the standard,
220
High Altitude
200 Low Altitude
Breakdown Voltage 180
160
140
120
100
80
60
40
0.1 0.15 0.2 0.25 0.3 0.35 0.4
Gap Length (m)
Figure 5.6: Comparison of IEC 60060-1 corrected low altitude and high altitude test
results
then the corrections should converge to the same values. That is, in the figure showing
the comparison of the two tests corrected according to IEC 60060-1 (Figure 5.6), it is
expected that the two trend lines overlap but this is not the case. Below the gap length
of 0.155 m, the corrected values are an average of 4.1% above the low altitude corrected
values whereas above 0.155 m they are an average 9.2% below the low altitude corrected
values.
Work also done in South Africa by Parus et al [5] shows that the IEC 60060-1 correction
does not correct the breakdown voltages obtained at different altitudes to the same
value. In their work, Parus et al compared IEC 60060-1 corrected results obtained at
altitudes of 822 m, 1 500 m and 1 880 m asl. It was also observed that the corrected
results at these three different altitudes do not converge to the same value as they should
if the corrections methods were correct. In order to verify the above mentioned claims
regarding the standard, tests were carried out at sea level. The comparison of the low
altitude and high altitude corrected voltages to that of the sea level tests are shown in
Figure 5.7. The environmental conditions during the sea level tests were not exactly
as the conventional standard conditions mentioned in IEC 60060-1 (2010) (20◦ C, 1 013
mbar and 11 g/m3 ). The temperature, pressure and humidity during testing averaged
approximately 23◦ C, 1 012.8 mbar and 18 g/m3 m. The humidity was higher than the
standard whereas the other parameters were comparable to the standard conditions.
220
High Altitude
200 Low Altitude
Sea Level
180
160
140
120
100
80
60
40
0.1 0.15 0.2 0.25 0.3 0.35 0.4
Gap Length (m)
Figure 5.7: Comparison of IEC 60060-1 corrected low altitude and high altitude test
results to sea level results
However, the conditions at sea level were the closest to the actual standard conditions,
therefore the results are taken as the standard voltages. The low altitude data has an
average error of 5.6% compared to the standard results, whereas the high altitude data
has an average error of 7.1%. The high altitude data shows that as the gap length
increases, the corrected voltage begins to deviate from the standard breakdown voltage.
Both high altitude and low altitude corrections deviate from the sea level data by more
than the acceptable 5% measurement error. However, the data shows that the corrections
of low altitude data are more accurate than that of high altitude data.
Rickmann et al [22] state that in order to use the IEC 60060-1 correction, the relationship
between m, w and g should be valid and this is when (δ×k )≈1. This condition can be
confirmed by the Cigré brochure [11] used to determine the IEC 60060-1 correction. The
brochure states that the standardized procedure was between 0.9 <(δ × k ) <1.1. From
the available data, this criteria is only met by the low altitude data as shown in Table
5.7.
This confirms the corrected data closest to the actual standard breakdown voltages is
that of the low altitude tests as the results show.
Table 5.7: Table of δ × k values
Gap Length (m) δ k δ ×k

0.10 0.95 1.03 0.98
0.15 0.95 1.05 1.00
0.20 0.95 1.05 0.99
Low Altitude
0.25 0.95 1.04 0.99
0.30 0.95 1.05 1.00
0.35 0.97 1.00 0.97
0.40 0.95 1.01 0.96
0.10 0.85 0.93 0.79

0.15 0.85 0.93 0.79
0.20 0.84 0.91 0.77
High Altitude 0.25 0.85 0.93 0.79
0.30 0.85 0.94 0.80
0.35 0.84 0.91 0.77
0.40 0.85 0.94 0.80
5.2 Pressure Vessel Results
Since pressure was drawn from ambient in the pressure vessel, it was difficult to do a
test at precisely the same pressure level for different gap lengths. Therefore, in order
to have comparable data sets, the results were plotted against the relative air density
due to the drawn vacuum for each test. By varying the relative air density, different
altitudes were simulated. Table 5.8 shows the different relative air densities and the
corresponding altitudes calculated as per equation 5.4 that were simulated during the
experiments.

−H
p = 1013.e 8150 (5.4)
Where H is the height above sea level in metres. Due to the limitation in gap length
Table 5.8: Approximate altitudes simulated during the tests
δ ≈ Altitude (m)
0.82 1 600
0.77 2 100
0.72 2 700
0.67 3 200
0.62 3 800
0.57 4 500
0.53 5 200
because of the size of the chamber, only gaps up to 0.3 m were tested and the results
are shown in Figure 5.8.
Table 5.9: Pressure Vessel Test Results
Gap Length (m) RAD Average Breakdown (kV) Std Deviation (%)
0.82 49.7 1.7
0.77 49.8 1.7
0.1 0.72 47.5 5.9
0.67 48.4 3.7
0.57 55.6 1.0
0.82 99.6 4.3
0.72 94.6 0.8
0.2 0.63 86.9 1.4
0.58 78.0 2.5
0.53 85.9 2.2
0.82 155.9 7.0
0.72 148.2 1.1
0.3
0.62 129.8 2.8
0.53 115.8 2.4
160
0.1 m
0.2 m
140 0.3 m
120
100
80
60
40
0.85 0.8 0.75 0.7 0.65 0.6 0.55 0.5
Relative Air Density
Figure 5.8: Results of breakdown voltage against relative air density in a variable
pressure vessel
The results of the pressure vessel tests are shown in Table 5.9 and Figure 5.8 is a plot
of the mean of 5 breakdowns per air gap per relative air density. The results obtained
followed the expected trend of the breakdown voltage decreasing with a decrease in
relative air density. However, it was observed that the breakdown voltage would start
increasing with further decrease in relative air density after a certain threshold relative
air density. This trend was observed for the 0.1 m air gap and, more clearly, for the 0.2
m air gap.
Plots showing the statistical deviations about the mean of the recorded voltage data (5
breakdowns per air gap) are shown in Figure 5.9 to Figure 5.11.
56 Mean
Maximum
Minimum
54
52
50
48
46
44
0.6 0.65 0.7 0.75 0.8
Figure 5.9: Variation of obtained breakdown voltages from the mean for 0.1 m gap
Mean
105 Maximum
Minimum
100
95
90
85
80
0.8 0.75 0.7 0.65 0.6 0.55

The maximum variation from the mean for the 0.1 m set of results is 7.9% at δ=0.72,
8.6% at δ=0.82 for the 0.2 m air gap and 9.8% at δ=0.82 for the 0.3 m air gap length.
These errors are particularly large and are mainly caused by an anomalous datum point
170 Mean
Maximum
160 Minimum
150
140
130
120
0.8 0.75 0.7 0.65 0.6 0.55

from the 5 points collected for that gap length at that voltage. These irregularities can
be attributed to the presence of space charge in the chamber that can distort the electric
field in an unexpected detrimental way. The average errors over the set of results are
3.9% for the 0.1 m air gap, 3.3% for the 0.2 m and 4.7% for the 0.3 m air gap. Therefore
the results are deemed reasonably accurate as they fall within the acceptable error of
5% and therefore worthy to be used for the following analysis.
For the 0.1 m air gap, breakdown in air densities greater than 0.72 was not precursed
by any visible corona on the electrode. However, lower than 0.72 relative air density, a
bluish film glow was observed to ignite around the tip of the rod electrode. The blue glow
would extinguish and replenish as the voltage was increased. It was at this stage that
the breakdown voltage of the air gap began to deviate from the usual trend of getting
smaller with a decrease in relative air density. The frequency of the extinguishing and
ignition of the blue glow increased as the the air gap approached breakdown then erratic
breakdown arcs would occur.
When conducting the same experiment on the 0.2 m gap, corona was observed along
with a blue film glow identical to that observed in the 0.1 m air gap. The blue film
glow became more prevalent as pressure was decreased until a point whereby a pressure
threshold was met and instead of the breakdown voltage decreasing as per norm, the
breakdown voltage began to increase. The same behaviour was noticed during the exper-
iment with the 0.3 m air gap, however, the pressure threshold whereby the voltage trend
starts increasing with decrease in air density was not met due to the aforementioned
chamber dimension constraints. From the results of the 0.1 m and 0.2 m air gaps, the
threshold point occurred at a lower relative air density as the gap length is increased.
Therefore, given a stronger vacuum pump and a bigger chamber, it would be expected
that the 0.3 m air gap would behave similar the 0.1 m and 0.2 m air gaps.
5.2.1 Why Breakdown Voltage Initially Decreased But Started In-

creasing With Further Decrease In Pressure
In general, it is expected that as pressure decreases, the voltage required to break down
an air gap also decreases. However, the results of the pressure vessel experiment showed
an interesting trend in that the expected trend was true only up to a certain threshold
beyond which the breakdown voltage would begin increasing with further reduction of
pressure. This threshold pressure seems to be related to the size of the air gap under
test. That is, as the gap length increased, then a lower pressure threshold would be
required for the breakdown voltage to start increasing.
A search in literature provided two possible explanations for the increase in voltage as
pressure decreased. One explanation is from the classical theory of breakdown mecha-
nisms and the other explanation is based on the observed corona activity before break-
down occurred.
5.2.1.1 Classical Theory of Breakdown Mechanisms
In the classical theory of breakdown mechanisms in a gas, the mean free path between
molecules in a gas plays a significant role in the breakdown process. It has been already
established that the mean free path is a function of pressure according to equation 5.5.
1
λα (5.5)
p
where λ is the mean free path between molecules and p is the pressure the molecules
are subject to. The relationship in equation (5.5) implies that the higher the pressure,
the smaller the mean free path between the molecules. A small mean free path limits
the ability of molecules to gain enough energy to be able to cause ionisation and initiate
the breakdown process. Therefore, in order to breakdown the air under high pressure,
a higher voltage is required for electrons to initiate the ionisation, thus implying that
the probability of ionisation is low for gases under high pressure. But, as the pressure
decreases, the mean free path increases and less voltage can be applied to initiate break-
down. Therefore the expected trend of the breakdown voltage of an air gap to decrease
with and increase in altitude.
However, there is a point whereby the pressure is very low and the mean free path so
large that the probability of ionisation becomes very low. That is, considering a closed
system, the molecules available for ionisation will be fewer and far apart. All this follows
Figure 5.12(Adapted from [29]) derived from equation (5.6) [28, 29]
−Bp
α = Ape E (5.6)
where α is Townsend’s first ionisation coefficient, A and B are constants that depend
on the gas type and temperature, p is the gas pressure and E is the electric field. The
Ionisation Coefficient ( α )
← Pressure too low; Optimal Pressure too high; →

Higher voltage for breakdown Pressure P→ Higher Voltage for breakdown
Figure 5.12: Probability of ionisation vs Pressure
variation of the pressure threshold as the air gaps get bigger can be attributed to the
electric field the air was subjected to. Therefore, as the gap length increased, the electric
field within the chamber also increased moving the threshold pressure lower.
5.2.1.2 Corona Theory Under DC Stress
Another explanation for the observed increase in voltage as pressure decreases is the
involvement of ion space charge during corona and breakdown. As observed in all three
gaps, the presence of a blue film glow at lower relative air densities corresponded to an
increase in breakdown voltage instead of the expected trend of decrease in breakdown
voltage. This is similar to the positive glow mode of corona discharge [27]. The theory on
positive corona discharge states that there are three modes of positive corona discharge
under DC voltage stress. These modes listed according to increasing field intensity for
their onset are [27]:
1. Burst corona
2. Onset streamer
3. Positive glow
Brief explanations for each of the modes are as follows.
Burst Corona
Burst corona is a result of ionisation occurring on the anode surface due to the electron
avalanches from the cathode losing its energy as they neutralise at the anode [27]. As a
result, positive ions are built up around the area where neutralisation occurred therefore
creating a local positive space charge, which eventually kills the discharge at that point.
Available free electrons are then neutralised at a different point on the anode while the
positive space charge earlier developed fades away[27].
Onset Streamer
As the voltage (and electric field) is increased, the local space charge around the anode
leads to field enhancement and attraction of the subsequent electron avalanches [27].
Due to the attraction of the subsequent electron avalanches to the positive space charge
field, streamers are developed and extend farther into the areas of low electric field
within the air gap thereby creating more positive ions (positive space charge) within
the low electric field region in the air gap. The overall effect of the subsequent electron
avalanches, the free electrons being absorbed at the anode and low positive ion mobility
leads to a net positive space charge being developed in front of the anode [27]. The
electric field gradient in front of the anode therefore drops below the critical field (due
to the positive space charge) required to sustain ionisation and the developed streamer
is suppressed. The presence of an applied field, due to the voltage source, removes the
created positive ion space charge and restores the initial field conditions and the process
restarts. This ultimately leads to pulsating discharges occurring around the anode [27].
The attributes of the onset streamer were observed prior to breakdown during tests on
the 0.1 m and 0.2 m air gaps at RADs of and below 0.77 and 0.57 (the knee points)
respectively. Breakdown of the 0.1 m air gap at RAD 0.77 occurred soon after the
presence of the onset streamer and no other corona mode was observed. The same
occurred for the 0.2 m air gap between RADs of 0.72 and 0.62.
Positive Glow
A continual increase in electric field increases the discharge and ionisation activities
over the anode surface. Within the region of the intense ionisation activity, a luminous
layer is created [27]. The positive glow is developed due to the applied electric field
being strong enough to quickly repel the positive ion space charge from around the
anode, therefore promoting ionisation activity on the anode surface. However, the field
intensity is not strong enough to develop radial discharges that lead to the formation
of streamers [27]. This leaves the surface of the anode with high intensity ionisation
activity and the luminous glow is developed. The development and presence of the
positive glow causes a significant increase in breakdown voltage [27].
During the experiments carried out in the lab, a faint blue luminous glow was observed
at low RAD values. As the RAD values became lower, the blue luminous glow became
brighter and that is when an increase in breakdown voltage was observed for the 0.1 m
and 0.2 m air gaps. The blue luminous glow was also observed during the experiment
with the 0.3 m air gap but breakdown occurred when the glow was not as bright as
observed in the 0.1 m and 0.2 m experiments. The visual observations and the behaviour
of the breakdown voltage suggests that the blue luminous glow observed was indeed a
positive glow and signalled a change in the breakdown mechanisms. This observation
suggests that when improved correction factors are determined, they should include a
factor that corresponds to the change in breakdown mechanisms for the air gap under
test at the appropriate altitude.
5.2.2 Atmospheric Correction Factors on Pressure Vessel Test Results
5.2.2.1 IEC 60060-1
IEC 60060-1 is the most common standard being used for atmospheric correction of air
gaps and some high voltage equipment, like insulator strings. Despite the fact that the
standard application is limited to a minimum relative air density of 0.8 [3] ( ≈2 000
m altitude), the correction factors were applied to the pressure vessel tests to find out
whether indeed the IEC standard was not applicable to altitudes above 2 000 m.
80
Sea Level
75 Uncorrected
Corrected
70
65
60
55
50
45
40
35
30
0.82 0.81 0.8 0.79 0.78 0.77 0.76 0.75 0.74 0.73 0.72
Figure 5.13: IEC 60060-1 correction of 0.1 m air gap results
The data of the three air gap sizes were corrected according to IEC 60060-1(2010) and
the plots of the obtained data are shown in Figures 5.13 to 5.15. Data points beyond
the knee point were removed from the correction as the data belongs to another regime
of breakdown mechanisms. These data points are for the 0.1 m air gap where RAD is
below 0.72 and for the 0.2 m air gap where RAD is below 0.58.
Figure 5.13 shows that the IEC 60060-1 standard corrects reasonably well for very small
gaps, in this case, 0.1 m air gap. The correction applied to the data according to IEC
150
Sea Level
140 Uncorrected
Corrected
120
110
100
90
80
70
60
0.8 0.75 0.7 0.65 0.6
240 Sea Level

Uncorrected
220 Corrected
200
180
160
140
120
100
80
60
0.8 0.75 0.7 0.65 0.6 0.55
60060-1 had an average 3% deviation from the expected sea level breakdown voltage.
As was also observed in Figure 5.7, the IEC correction of the smallest gap size (0.1 m)
is the most accurate despite the altitude. Figures 5.14 and 5.15 also confirm the trend
shown by the low altitude and high altitude data in that as the gap length increases, the
correction becomes more inaccurate. The average deviation from the sea level breakdown
voltages for the 0.2 m and 0.3 m air gaps was 15.5% and 20.75% respectively. These
deviations are large compared to the low altitude and high altitude data from the open
air experiments (5.6% and 7.1% respectively), thereby confirming that the IEC standard
is indeed unsuitable for data obtained from RAD less than 0.8. The observed data also
suggests that for very small gaps (in the order of 0.1 m and less), IEC 60060-1 is suitable
when RAD is as small as 0.72.
The trend of the corrected voltages seem to suggest that the IEC 60060-1 has a mathe-
matical flaw that does not allow it to correct appropriately for small relative air densities,
or that there is change in breakdown mechanisms as RAD reduces. According to [30],
the average breakdown electric field at time of breakdown is determined by equation 5.7
and can be used to determine the mechanism that caused the breakdown.
Vbreakdown
EBreakdown = (5.7)
dbreakdown
An electric field of approximately 500 kV/m implies a streamer dominated breakdown

mechanism. This is because 500 kV/m has been adopted as the mean electric field for
streamer propagation under standard atmospheric conditions [11]. By calculating the
average electric fields of the obtained results using equation 5.7, the calculated electric
fields imply streamer breakdown in all three air gaps. The calculated electric fields are
shown in Table 5.10.
Table 5.10: Table showing the average electric fields during breakdown
Gap Length (m) RAD Ebreakdown (KV/m)

0.82 497.00
0.77 497.60
0.1 0.72 474.60
0.67 484.20
0.57 556.00
0.82 498.20
0.72 472.90
0.2 0.63 434.40
0.58 390.00
0.53 429.30
0.82 519.67
0.72 493.87
0.3
0.62 432.80
0.53 386.00
However, it is expected that electric field required for positive streamer propagation
should be proportional to air density [11]. This is to imply that if the breakdown of
the air gaps was dominated by positive streamers, it would have been expected that
the electric field also decrease with a decrease in relative air density for all three gap
lengths, but this is not so for the 0.1 m air gap. The air gap seems to be displaying
varying electric fields as the relative air density decreases.
The Cigré brochure [11] states that the parameter g (calculated from equation 5.8 where
L is the discharge length and k is a parameter for the voltage type used) is used to give
an estimate of the similarity of the discharge to that of a pure positive streamer.
U50
g= (5.8)
500δLk
The brochure states that [11]:
• g≈1 implies a positive streamer discharge process. This is when the dielectric
strength is expected to be proportional to the relative air density
• g<1 implies a streamer and leader dominated discharge with the dielectric strength
is not as influenced by the relative air density
• 1<g<2 implies a positive and negative streamer dominated discharge process and
the situation is deemed ”very complicated”
• g>2 implies a uniform air gap and the breakdown is corona dominated.
To be able to determine the breakdown mechanisms of the air gaps, the parameter g
was calculated for each air gap and shown in Table 5.11. The calculated g parameters
were all found to be between 1 and 2, implying both negative and positive streamer
dominated discharge thus presenting a “complicated”scenario.
From the principles used to derive the IEC 60060-1 standard, the relationship in equation
5.9 was established
Ub
δm = (5.9)
500d
where m is a parameter that covered the influence of the environmental conditions,

the waveform applied and the parameters of the insulation under test [30]. When the
breakdown process of an air gap is streamer dominated, then m=1 [30]. According
to IEC 60060-1, m=1 when 1<g<2, which is when the positive and negative streamers
dominate the discharge process. For the collected data, the calculated m values according
Table 5.11: Table showing the calculated g-factors
Gap Length (m) δ g

0.1 0.82 1.22
0.77 1.30
0.72 1.33
0.67 1.45
0.57 1.96
0.2 0.82 1.23
0.72 1.32
0.63 1.39
0.58 1.37
0.53 1.63
0.3 0.82 1.25
0.72 1.35
0.62 1.37
0.53 1.42
the equation 5.9 were mostly negative. This scenario shows a problem with the IEC
60060-1 whereby the same m value is used to correct voltages governed by different
environmental, waveform and insulation material parameters. This concern was also
shared by [22]. Therefore, there is a need for a more refined method to determine the
m parameter in order to obtain accurate corrections.
5.2.2.2 Calva’s Method Applied For Comparison With Pressure Vessel Tests
The results of the experiment were compared to the procedure proposed by Calva et al
[9] according to equations 5.1 to 5.3. The comparison of the test results to the Calva
prediction are found in Table 5.12 and in Figure 5.16 to Figure 5.18. The plots show
that the proposed method by Calva does not cater for the increase in voltage with the
further decrease in pressure, but it does provide some form of trend that agrees with the
experimental data.
The trend in the results shows that the Calva prediction method can be modified to fit
the experimental data.
Table 5.12: Pressure Vessel Test Results Compared to Calva Prediction
Gap RAD Average Calva Error

Length (m) Breakdown (kV) Prediction (kV) (%)
0.82 49.7 45.9 8
0.1 0.77 49.8 41.7 16
0.72 47.5 37.6 21
0.82 99.6 85.3 14
0.72 94.6 69.9 26
0.2
0.63 86.9 56.7 35
0.58 78.0 47.8 39
0.82 155.9 126.1 19
0.72 148.2 105.1 29
0.3
0.62 129.8 85.4 34
0.53 115.8 68.4 41
50
48
46
44
42
40
38
Experimental
Predicted
36
0.82 0.81 0.8 0.79 0.78 0.77 0.76 0.75 0.74 0.73 0.72
Figure 5.16: Comparison of experimental results with Calva’s prediction method for
0.1 m air gap
5.2.2.3 Mapping of Calva’s Prediction Method to the Experimental Data
In order to map the Calva predicted results to the experimental data, a difference func-
tion was derived and added to the Calva prediction function. By so doing, the Calva
prediction function was therefore scaled in gradient and mapped upwards to fit the ex-
perimental data. This method was done considering that a linear trend was observed in
the results and therefore a linear function was deemed suitable for the mapping. The
data points where the breakdown voltage began to increase due to the corona were not
considered.
100
Experimental
Predicted
90
80
70
60
50
40
0.8 0.75 0.7 0.65 0.6 0.55
0.2 m air gap
160
Experimental
150 Predicted
140
130
120
110
100
90
80
70
60
0.85 0.8 0.75 0.7 0.65 0.6 0.55 0.5
0.3 m air gap
Consider the pressure vessel experiment for the 0.1 m gap size with information in Table
5.13.
The difference between the experimental breakdown voltage (Ub ) and the Calva pre-
dicted breakdown voltage (Ub(predicted) ) is calculated and plotted against the relative air
density (δ) as shown in Figure 5.19. A line of best fit was then determined to create
Table 5.13: Pressure vessel 0.1 m gap test information
δ Ub (kV) Ub(predicted) (kV) Ub -Ub(predicted) Calva+Difference function

0.82 49.7 45.92 3.78 49.90
0.77 49.76 41.74 8.02 48.77
0.72 47.46 37.59 9.87 47.76
the difference equation (equation (5.10)) which would be added to the Calva prediction
function.
Udif f = −61δ + 54 (5.10)
The resulting function, shown in equation 5.12, where Udif f is defined in equation 5.10
and the other terms are as previously defined. Equation 5.12 shows the adjusted pre-
diction method for the 0.1 m air gap tests and it is plotted to compare how accurate it
is in relation to the experimental data in Figure 5.20.
Vb = [500d(k1 + k2 )S] + Udif f (5.11)
Vb = [500d(k1 + k2 )S] − 61δ + 54 (5.12)
11
Difference
10 Difference Trendline
9
Voltage Difference(kV)
7
y = − 61*δ + 54
6
3
0.82 0.81 0.8 0.79 0.78 0.77 0.76 0.75 0.74 0.73 0.72
Figure 5.19: Difference function for 0.1 m pressure vessel experiment
The application of the difference method to the 0.1 m gap pressure vessel experiments
produced a result with absolute errors shown in Table 5.14 of each reading taken.
50
48
44
42
40
38 Experimental
Predicted
Experimental + Difference function
36
0.82 0.81 0.8 0.79 0.78 0.77 0.76 0.75 0.74 0.73 0.72
Figure 5.20: Comparison of prediction including difference method for 0.1 m gap
length
Table 5.14: Error of mapped data to experimental data for 0.1 m air gap
δ Ub (kV) Ub(mapped) (kV) Absolute Error (%)

0.82 49.70 49.90 0.40
0.77 49.76 48.77 2.03
0.72 47.46 47.67 0.44
The error of the mapping for the 0.1 m gap size is within the acceptable statistical
variation 5%. The same procedure was repeated for the 0.2 m and 0.3 m air gap sizes.
The difference functions to be added to the Calva prediction method, as done in equation
5.12, for the 0.2 m and 0.3 m air gaps are shown in equations 5.13 and 5.14 respectively.
Udif f = −68δ + 72 (5.13)
Udif f = −57δ + 79 (5.14)
Similar plots to Figure 5.20 were done for the 0.2 m and 0.3 m gaps and are shown in
Figure 5.21 and Figure 5.22 respectively.
Tables 5.15 and 5.16 show the errors of the mapping for each gap length. The tables
show that the mapping applied reduced the error in experimental and predicted results
to within 5%.
110
100

90
80
70
60
50 Experimental
Predicted
Experimental + Difference Function
40
0.8 0.75 0.7 0.65 0.6 0.55
length
160
150
140
130
120
110
100
90
80
Experimental
70 Predicted
Predicted + Difference Function
60
0.85 0.8 0.75 0.7 0.65 0.6 0.55 0.5
length

0.82 99.64 101.49 1.86
0.72 94.58 92.96 1.71
0.63 86.88 85.84 1.20
0.58 78.00 80.36 3.03

0.82 155.90 158.99 1.98
0.72 148.16 143.59 3.09
0.62 129.84 129.50 0.26
0.53 115.80 117.60 1.55
Tables 5.14 to 5.16 show that the prediction method proposed by Calva is viable for
predicting the breakdown voltage of small air gaps at low relative air densities when
modified by adding equations 5.10 to 5.14 to the Calva prediction method for air gaps
of sizes 0.1 m - 0.3 m respectively in order to attain accuracy within 5%. Due to
limitations in supply voltage, pressure vessel size and pump drawing ability, more data
could not be collected for a wider range of relative air density and gap size. Therefore,
the adjustments made to the Calva prediction method for the respective gap sizes are
specifically for the mentioned gap sizes and relative air density range.
Due to the lack of a large data set and the number of variables involved in the breakdown
process of air (relative air density, humidity, rod geometry etc), the difference equations
could not be generalized into a single function to be added to the Calva prediction
method so that it applies to any gap length and this could be an interesting extension
of this work in the future. Also, by collecting more data, the difference equation may
change in order to fit all the possible RAD ranges and also increase the accuracy of the
function.
Closing Summary In Answering the Research Question Of This Dissertation
Tests were conducted and it has been shown that the atmospheric correction procedure in
the IEC 60060-1 (2010) is more suitable for use when applied to low altitude data. When
applied to high altitude data (1 740 m), the procedure has been found to introduce errors
as the gap length increases. The Calva prediction method has proved to be accurate
when applied to data below 1 740 m in altitude. Despite the IEC 60060-1 (2010) stating
that the atmospheric procedure is inapplicable when RAD is less than 0.8, corrected
data from the pressure vessel suggests that the procedure is accurate for very small
air gaps (0.1 m and below) up to RAD 0.72. The results from the other air gap sizes
tested (0.2 m and 0.3 m) agree with the claim by IEC 60060-1 (2010) in the standards
inability to obtain accurate atmospheric correction factors when RAD is lower than
0.8. Calva’s prediction method was inaccurate when applied to the pressure chamber
conditions, but the prediction methods results followed the trend of the experimental
results. Factors were derived to add to the Calva prediction method for the method to
fit the experimental data.
The experiment conducted in the pressure vessel showed that as pressure decreases,
the breakdown voltage of the same gap length decreases until a threshold whereby the
voltage begins to increase. This behaviour was explained by means of the classical
breakdown theory and the corona mechanisms under HVDC stress.
Therefore, the altitude correction method stated in IEC 60060-1 (2010) and the Calva
prediction method are both applicable and accurate when applied to low altitude data.
Calva’s model has an advantage in that it can predict the breakdown voltage of an air
gap by considering the environmental conditions and the air gap length only. Whereas,
the IEC 60060-1 (2010) standard requires more parameters. Both IEC 60060-1 (2010)
standard and the Calva prediction method are not applicable for a wide altitude range.
However, the Calva prediction method can be modified so as to increase the accuracy.
The method is therefore recommended for further modification.
Chapter 6
Conclusions and
Recommendations
6.1 Conclusions
The following are the important findings and conclusions of the research:
Open Air Tests (0 m - 1 740m asl)
1. IEC 60060-1 altitude correction factors’ accuracy improves with reduction in alti-
tude.
2. IEC altitude correction factors confirmed to be applicable for altitudes below 1

800 m asl. However, the increase in error is apparent as the altitude and air gap
length increases.
3. The Calva prediction method accurately predicts the DC breakdown voltage for
air gaps within the 5% acceptable error at altitudes up to 1 740 m asl.
62
Chapter 6. Recommendations and Conclusion 63
Pressure Vessel Tests Simulating High Altitudes (1 740 m - 5 200 m

asl)
1. At simulated altitudes higher than 1 740 m, IEC 60060-1 correction factors are
inaccurate (resulting in errors greater than 5%). This confirms that the standard
does not apply for altitudes above 1 740 m asl.
2. Each gap size breakdown voltage as a function of altitude has a knee point thresh-
old where the trend changes from decrease of breakdown voltage with decrease in
pressure to an increase in voltage with further decrease in pressure.
3. For each gap length, the Calva prediction method can be modified to accurately
predict the breakdown voltage at high altitude.
6.2 Recommendations For Future Work
The following recommendations have been suggested to further the current research:
• It is imperative that tests be done at sea level to obtain baseline breakdown volt-
ages. These tests should be done at the standard temperature of 200 C and can
be taken as the standard breakdown voltage of the tested gap sizes. All proposed
atmospheric correction methods applied should produce results that compare to
the standard voltages obtained.
• The work conducted in this work was done using positive polarity HVDC. It is
recommended that the work be redone using negative polarity so that it can be
compared whether the gaps perform in a similar mechanism to that of the positive
polarity. That is, to test whether a significant increase in breakdown voltage
is observed at lower pressures and if the Calva prediction method does produce
similar results to those experimentally obtained.
• Tests results for the same gap sizes are required for altitudes with RADs of 0.77,
0.72, 0.62 and 0.53. These results can be used to compare with those obtained in
the pressure vessel.
Chapter 6. Recommendations and Conclusion 64
• The work can repeated for increased gap sizes. This would imply a bigger pressure
vessel and higher voltages and would pose the problem of a bushing. Therefore a
new pressure vessel that could meet the given requirements should be designed.
• It is recommended that more results be obtained for various gap sizes at various
RADs. This is to increase the data available to accurately derive the appropriate
function to add to the Calva prediction method. Overall, it would be desired that
a general function be derived for adding to the Calva prediction method instead
of different functions specific to an air gap as done in this work.
• In order to obtain photographic evidence of the breakdown mechanism that occur

during the breakdown process, it is recommended that a high speed camera be
acquired.
References
[1] P. Tuson. HVDC Strengthening in Southern Africa. In Power Engineering

Society Conference and Exposition in Africa, 2007. PowerAfrica ’07. IEEE,
pages 1–6, July 2007. doi: 10.1109/PESAFR.2007.4498096.
[2] Marco Glastra. Drakensberg-mafadi peak. URL http://www.footprint.co.

za/drakensberg_mafadi.htm. Last accessed: 25 February 2016.
[3] SANS 60060-1. High-voltage test techniques. part 1: General definitions and
test requirements, 2011.
[4] T. Britten, L. Pillay, P. Naidoo, F F. Bologna, D. Muftic, and N. Ijumba.

Eskom’s proposed strategic research into hvdc transmission. IEEE Power
and Energy Magazine, pages 28–32, July 2005.
[5] N. Parus, I R. Jandrell, N. Mahatho, T. Govender, and H A. Roets. Live

work under hvdc voltage: An investigation into altitude effects of flashover
voltage for rod-plane air gaps. In Proceedings of the Eighteenth International
Symposium on High Voltage Engineering, Seoul, South Korea, August 2013.
[6] F.W. Peek. Effect of altitude on the spark-over voltages of bushings, leads
and insulators. Transactions of the American Institute of Electrical Engineers,
XXXIII(2):1721–1730, June 1914. ISSN 0096-3860. doi: 10.1109/T-AIEE.
1914.4765197.
[7] T. Udo and Yasuo Watanabe. Dc high-voltage sparkover characteristics of

gaps and insulator strings. IEEE Transactions on Power Apparatus and Sys-
tems,, PAS-87(1):266–270, Jan 1968. ISSN 0018-9510. doi: 10.1109/TPAS.
1968.291998.
65
Reference 66
[8] N.L. Allen, M. Boutlendj, and H.A. Lightfoot. Dielectric breakdown in nonuni-
form field air gaps. IEEE Transactions on Electrical Insulation,, 28(2):183–
191, Apr 1993. ISSN 0018-9367. doi: 10.1109/14.212243.
[9] P.A. Calva, V. del Moral, M.G. Marquez, and G.P. Cabrera. In Conference
on Electrical Insulation and Dielectric Phenomena, 2003. Annual Report.
[10] N.L. Allen. Correction procedures for high-voltage measurements under a

variable air density. IEEE Proceedings on Science, Measurement and Technol-
ogy, 153(5):194–200, September 2006. ISSN 1350-2344. doi: 10.1049/ip-smt:
20050040.
[11] WG 33-07. Guidelines for the evaluation of the dielectric strength of external
insulation, 1992.
[12] A. Robledo-Martinez and P A. Calva. Dc breakdown characteristics of rod-rod

and rod-plane gaps in reduced-density air. In Proceedings of the Sixth Inter-
national Symposium on High Voltage Engineering, volume 3, New Orleans,
USA, 1989.
[13] M. Boutlendj and N L. Allen. Assessment of air-density correction for prac-

tical electrode systems. European Transactions on Electrical Power, 6(4):
267–274, July 1996.
[14] IEEE standard for high-voltage testing techniques. IEEE Std 4-2013 (Revi-
sion of IEEE Std 4-1995), pages 1–213, May 2013. doi: 10.1109/IEEESTD.
2013.6515981.
[15] H J. Geldenhuys. The breakdown voltage of air in a 50 cm rod-plane gap over

a practical range of air density and humidity. 5th International Symposium
on High Voltage Engineering, 1987.
[16] A. Pigini, G. Sartorio, M. Moreno, M. Ramirez, R. Cortina, E. Garbagnati,

A.C. Britten, and K.J. Sadurski. Influence of air density on the impulse
strength of external insulation. IEEE Transactions on Power Apparatus and
Systems, PAS-104(10):2888–2900, Oct 1985. ISSN 0018-9510. doi: 10.1109/
TPAS.1985.319134.
Reference 67
[17] SANS 60060-1. High-voltage test techniques. part 1: General definitions and
test requirements, 1989.
[18] M. Ramirez, M. Moreno, A. Pigini, G. Rizzi, and E. Garbagnati. Air density

influence on the strength of external insulation under positive impulses: Ex-
perimental investigation up to an altitude of 3000m a.s.l. IEEE Transactions
on Power Delivery, 5(2):730–737, April 1990.
[19] K. Feser and J. Schmid. Temperature effect on the streamer breakdown of

air. In Proceedings of the Sixth International Symposium on High Voltage
Engineering, volume 3, New Orleans, USA, August 1989.
[20] IEC 60071-2. Insulation co-ordination- part 2: Application guide, 1996.
[21] I. Gutman, A. Pigini, J. Rickmann, J. Fan, D. Wu, and E. Gockenbach.

Atmospheric and altitude correction of air gaps, clean and polluted insulators:
state-of-the-art within Cigré and IEC. Number D1-213, France, August 2014.
[22] J. Rickmann et al. Analysis and comparison of atmospheric and altitude

correction methods for air gaps and clean insulators. In Proceedings of the
Eighteenth International Symposium on High Voltage Engineering, number
OC2-03, pages 542–547, Seoul, South Korea, August 2013.
[23] H. Isa, Y. Sonoi, and M. Hayashi. Breakdown process of a rod-to-plane gap

in atmospheric air under dc voltage stress. IEEE Transaction on Electrical
Insulation, 26(2), April 1991.
[24] T. Suzuki, K. Miyake, and T. Hara. Breakdown process in rod-to-plane gaps

with dc voltages. IEEE Transactions on Industry Applications, IA-21(1), Jan-
uary/February 1985.
[25] T. Christen, H.Bohme, A. Pedersen, and A. Blaszczyk. Streamer line model-

ing. In Scientific Computing in Electrical Engineering, volume 16 of Mathe-
matics in Industry. Springer-Verlag, New York, 2012.
[26] M.S. Naidu and V. Kamaraju. High Voltage Engineering. Tata McGraw Hill
Education Private Limited, 2009. ISBN 9780070669284.
Reference 68
[27] F.A.M. Rizk and G.N. Trinh. High Voltage Engineering. Taylor & Francis,
2014. ISBN 9781466513761.
[28] E. Kuffel, W.S. Zaengl, and J. Kuffel. High Voltage Engineering: Funda-
mentals. Applied Electricity and Electronics. Butterworth-Heinemann, 2000.
ISBN 9780750636346.
[29] F.H. Kreuger. Industrial High Voltage 1, 2, 3. Delft University Press, Delft,
Netherlands, 1991.
[30] D. Wu and J. Rickmann. Challenges in developing a unified method for at-

mospheric correction. Cigré SCD1 Colloquium, Cigré, Rio de Janeiro, Brazil,
2015.
Appendix A
Experimental Data And Results
Table A.1: Low Altitude Experimental Data
Gap Length(m) 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Temperature(0 C) 27.6 26.2 27.8 26.5 27.2 21.9 25.5

Pressure (mbar) 987.8 986.3 986.1 985.4 985.2 985.8 985.5
Rel. Humidity (%) 48.6 56.0 51.8 54.8 54.9 56.0 46.0
Test 1(kV) 57.4 81.8 107.6 131.7 157.3 174.8 197.0

Test 2(kV) 55.6 81.6 107.8 131.6 156.5 174.9 196.9
Test 3(kV) 57.4 83.0 107.3 131.6 156.8 174.7 196.3
Test 4(kV) 56.9 82.6 107.6 131.3 157.6 174.8 197.0
Test 5(kV) 57.3 82.2 107.8 131.2 157.3 175.4 196.6
Ave. Breakdown Voltage (kV) 56.92 82.24 107.62 131.48 157.1 174.92 196.76
Table A.2: High Altitude Experimental Data
Gap Length(m) 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Temperature(0C) 15.3 14.4 15.4 14.9 15.1 15.8 14.8

Pressure (mbar) 844.7 841.7 840.2 841.1 842.3 839.7 842.3
Rel. Humidity (%) 41 45 34 42 46 33 47
Test 1(kV) 54.2 68.2 81.1 100.8 118.8 129.4 156.8

Test 2(kV) 52.5 66.3 80.9 99.7 119.0 130.0 157.2
Test 3(kV) 52.4 68.3 81.1 100.2 118.4 129.6 156.2
Test 4(kV) 51.2 68.1 81.4 100.0 116.6 130.0 157.2
Test 5(kV) 52.2 67.9 81.5 100.1 118.2 130.6 156.6
Ave. Breakdown Voltage (kV) 52.50 67.75 81.18 100.15 118.20 129.92 156.80
69
Appendix A. Experimental Data And Results 70
Table A.3: Sea Level Results
Gap length (m) 0.10 0.15 0.20 0.25 0.30 0.40
Temperature 23.6 23.5 23.4 23.4 23.3 23.3

Pressure (mbar) 1012.4 1012.4 1012.5 1012.6 1012.8 1012.7
Rel. Humidity (%) 89 89 90 90 92 92
Test 1 (kV) 64.0 87.0 112.0 140.0 164.0 218.0

Test 2 (kV) 64.0 87.0 112.0 140.0 164.0 218.0
Test 3 (kV) 63.0 88.0 113.0 140.0 164.0 216.0
Test 4 (kV) 63.0 88.0 111.0 140.0 165.0 220.0
Test 5 (kV) 64.0 87.0 112.0 139.0 164.0 218.0
Ave. Breakdown Voltage (kV) 63.6 87.4 112.0 139.8 164.2 218.0
Table A.4: Pressure Vessel Test Experimental Data
Gap Length (m) 0.1 0.2 0.3
Ambient Pressure(mbar) 835.9 835.6 834.7 835.1 831.5 831.5 834.4 837.2 837.4 837.1 838.8 838.4 837.9 837.5
Test Pressure(mbar) 835.9 785.6 734.7 685.1 581.5 831.5 734.4 637.2 587.4 537.1 838.8 738.4 637.9 537.5
Temperature(0 C) 23.2 23.2 23.1 23.2 23.2 20.4 20.2 20.3 20.3 20.2 22.1 22.3 22.5 23.1
Rel. Humidity(%) 42 42 42 43 44 48 49 49 45 49 49 50 50 49
δ 0.82 0.77 0.72 0.67 0.57 0.82 0.72 0.63 0.58 0.53 0.82 0.72 0.62 0.53
Test 1 (kV) 48.6 50.9 44.5 47.3 56.7 98.1 94.9 87.2 76.7 83.2 171.1 145.7 124.5 113.7
Test 2 (kV) 50.3 49.6 49.2 51.4 55.6 98.2 95.3 88.0 80.1 86.5 146.3 149.9 126.7 114.0
Test 3 (kV) 49.0 50.0 50.7 49.3 55.2 97.0 93.8 86.2 76.4 86.2 150.3 147.9 132.2 117.8
Test 4 (kV) 49.7 48.3 43.7 48.0 55.3 96.7 93.5 88.1 76.6 88.4 - 147.2 134.6 120.3
Test 5 (kV) 50.9 50.0 49.2 46.1 55.2 108.2 95.4 84.9 80.2 85.0 - 150.1 131.2 113.2
Appendix A. Experimental Data And Results
Ave. Breakdown Voltage(kV) 49.7 49.8 47.5 48.4 55.6 99.6 94.6 86.9 78.0 85.9 155.9 148.2 129.8 115.8
71
Appendix B
SAUPEC 2015 Discussion Paper
This appendix contains a discussion paper handed in to the South African Universities
Power Engineering Conference in 2015.
72
A CRITICAL ANALYSIS OF THE CURRENT HIGH VOLTAGE
ALTITUDE CORRECTION METHODS
T. Gora and Dr. C. Nyamupangedengu∗
∗ Faculty of Engineering and the Built Environment, Private Bag 3, Wits 2050, South Africa E-mail:
tatenda.gora@students.wits.ac.za
cuthbert.nyamupangedengu@wits.ac.za
Abstract: Altitude correction factors are a critical component in the design of high voltage systems at
high altitudes. Unfortunately, the common standards currently used for altitude correction are mostly
valid for altitudes of 1 800 m and have been proven to be unreliable for altitudes higher than this. This
paper critically analyses and discusses the discrepancies in three popular standards used for altitude
correction. The research work on altitude correction being done at Wits University is highlighted in the
context of the worldwide efforts in this knowledge area.
Key words: Altitude, DC, Correction Factor
1. INTRODUCTION incorporated into the standards. Further verification of the

proposed models is required, especially for DC voltages.
The desire to meet electricity demands, especially in
emerging economies, has led power utilities to embark EVOLUTION HISTORY OF THE ALTITUDE CORREC-
on projects that involve the transmission of power from TION FACTORS
sources that are far away from the load centres. The
number of High Voltage Direct Current (HVDC) schemes 2.1 Peek Model
around Europe and Asia and the lengths of the transmission
lines show the extent at which HVDC is being used to The effect of altitude on the breakdown voltage of
connect the load centres to remote generation sources [1]. insulators was noted as early as 1914 by Peek [6], and
In China, the Xiangjiaba-Shanghai HVDC link connects a lot of work has been done since then with regard
the load centre from the generation source which is to air density effects on the breakdown voltage of air
approximately 4 000 km away. The Xiangjiaba-Shanghai gaps and external insulation, and more recently air gaps
link passes through the Tibet region which is quite under DC voltage stress [3, 4, 7, 8, 9, 10, 11, 12]. Peek
mountainous, resulting in the line passing through altitudes [6] investigated the effect of altitude on the spark-over
above 4 000 m. Research is still being done on insulator voltages of bushings, leads and insulators by placing them
performance at such high altitudes [2]. In Africa, Eskom in a cask and varying the pressure. Pressure was varied
is considering possibilities of importing power from the to obtain relative air densities between 0.5 and 1.0 while
Congo. This project will involve the transmission of the temperature variations due to the pressure changes
power over distances of approximately 4 000 km. For this were recorded. In his research, Peek [6] was able to
project to be economically viable, the power will have to produce tables of correction factors to be used on specific
be transmitted using high voltage direct current (HVDC) insulators by reading off the relative air density. Therefore,
technology. However, the available standards for altitude the correction factors for insulators were different from
correction have discrepancies that can hinder the design of those of bushings and leads under the same relative air
a reliable system at high altitudes. density [6]. Peek attributed these to the uniformity of the
This paper highlights the evolution of the altitude field suggesting that the flashover voltage for a uniform
correction factors being applied and discusses the field decreases directly with δ, whereas for non-uniform
discrepancies in the common standards used in the world. fields, the flashover voltage decreases at a lesser rate than
A brief discussion on current work being done to solve δ [6]. This means that given that the flashover voltage of
the problem of unreliable altitude correction factors is also an insulator at sea level (δ=1) is v1 , then the breakdown
given. voltage v at δ=x was given by
v = xv1 (1)
2. BACKGROUND
where δ is the relative air density at a specific altitude
As altitude increases, air density decreases and air and read from the tables derived by Peek [6]. This
breakdown mechanisms change. The IEC 60060-1 (2010) therefore implied that the breakdown voltage was directly
standard has been found to cover only a limited altitude proportional to the air density. Peek also concluded that
range [3, 4, 5]. Some models that can possibly extend the effects of the change in δ due to pressure changes and
the altitude correction factors to cover a higher altitude constant temperature were the same as the change in δ due
range have been suggested in literature but are not yet to temperature changes and constant pressure [6].
2.2 IEC 60-1 If breakdown occurs purely due to the streamer
mechanisms (without any leader formation), the leader
In 1973, the IEC 60-1 was amended. In the 1973 version terms E and l are set to zero leaving the breakdown voltage
of IEC 60-1, it was stated that the breakdown voltage of to be dependent on only the streamer field Es and streamer
an air gap or insulator at relative air density δ, uδ is related length ls . With this information, Pigini et al [15] were able
to the breakdown voltage of the same air gap or insulator to derive a factor G to be used to determine the index of
under standard conditions, u0 by [13] the relative air density correction factor by incorporating
the discharge mechanism used and the relative air density.
uδ = kd u0 (2) To correct the voltage, a similar expression to equation (2)
was used but modified as follows,
where kd is the air density correction factor defined as
U50 = U50(std) (δ.K)n (7)
P T0
kd = ( )m ( )n (3) where K is the humidity correction factor, n is the
P0 T
air density correction factor and U50 is the breakdown
where P and P0 are the pressure and standard pressure voltage under non-standard conditions. The factor G
respectively while T and T0 are the temperature and was determined on the assumption that the major process
standard temperature respectively. The exponents m during the breakdown of the gap was streamers [15]. G
and n are the variations of pressure and temperature was defined as
respectively that are supposed to be determined before U50
G= (8)
kd is evaluated. Since the air density is dependent on 500δkd
pressure, the determination of the exponent m became the where d is the length of the discharge gap and k is a factor
area of concern regarding atmospheric corrections and was dependent on the type of voltage used and δ is defined as:
defined in IEC 60-1 (1973) as
P 273 + t0
m = −0.12d + 1.12 (4) δ= (9)
P0 273 + t
where d is the air gap length between 1 m and 6 m. For To determine n in equation (7), the G factor, equation (8),
d≤1 m, m=1.0 and m=0.4 for d≥6 m. The factor m in this was used and n took the following values [15] depending
standard was therefore independent of the type of voltage on the range of values of G
used and the relative air density being corrected for. This
therefore implied that the variation of pressure was only G0 (G0 − 0.2)
n= (10)
dependent on the length of the air gap. Compared to the 0.8
model proposed by Peek, the model in the IEC 60-1 does 3 − G0
not give a linear relationship between the flashover voltage n= (11)
2G0
and the relative air density. For DC and impulse voltages
of either polarity, the standard specifies that m and n can where G0 is the G factor under standard conditions (δ and
be considered equal. k=1). Equation (10) was valid for 0.3≤G≤1 and equation
(11) was valid for 1≤G≤2. These results by [15] were
incorporated into the IEC 60060-1 (1989), including the
2.3 Pigini Model(IEC 60060-1) graphs of equation (10) and equation (11) as the air density
correction factors [16]. The results still play a major role
During his research, Geldenhuys [14], was able to derive in the recent IEC 60060-1(2010) which, instead of using
the approximate expression of the average electric field graphs for determining G as in IEC 60060-1(1989), uses
required for positive streamer propagation as given in equations that are quite similar to equations (10) and (11)
equation (5). [17]. It is important to note that the model derived by
Pigini et al takes into account the voltage type through
Es+ = 425δ1.5 + (4 + 5δ)H (5) the factor k in equation (8) whereas IEC 60-1 (1973) did
not include any factor that took into account the type of
where H is the humidity and δ is the relative air density. voltage used. Since the standard was produced from the
By varying the humidity, the average electric field required work conducted by Pigini et al, the standard is limited to
for streamer propagation under standard conditions was an altitude of 1 800 m. Given that the only difference in the
determined to be 500 kV/m. The breakthrough achieved altitude correction procedures highlighted in the current
by Geldenhuys [14] gave way for the determination of the version of IEC 60060-1 (2010) with those of IEC 60060-1
breakdown voltage, by taking reference of the final jump (1989) is the use of equations to determine the index n
that leads to breakdown, of an air gap in terms of the instead of graphs, it therefore holds that the current version
electric field of the leader zone of the discharge (El ), the of IEC 60060-1 is also limited to an altitude of 1 800 m.
length of the leader (ll ), the electric field of the streamer
zone (Es ) and the length of the streamer (ls ) to be given by
equation (6) [15].
U = El ll + Es ls (6)
2.4 Ramirez Model model proposed by Calva et al yields more accurate results
than the IEC 60060-1(2010).
A year after IEC 60060-1 (1989) was published, Ramirez It has been observed that when the models by Ramirez,
and his counterparts [18] proposed a model claimed Pigini and Calva were derived, there were inconsistencies
to be more accurate than the one incorporated in IEC in the type of voltage used. It is suspected that the
60060-1(1989) and takes account of the influence of air inconsistency in the voltage type used to derive the models
density on rod-plane gaps. The model is expressed in could be among the causes of the variations in the models.
equation (12) The variation in the types of voltages used for the models
is as follows:
U 0.8[1 + T (1 − δ)](δ − 0.2G00 )
= + 0.2 (12)
U0 (1 − 0.2G00 ) • IEC 60060-1(2010) has input from empirical results
conducted with switching impulses [14, 19] and
where δ is the relative air density calculated by equation
lightning impulses [14, 19, 15]
(9) and
1 − 0.8G00 • The model presented by Ramirez was derived under
T = 1.4 (13)
1 − 0.2G00 impulse voltage conditions [18]
with G0 defined as • The model by Cavlva et al is purely under DC voltage
U0 conditions [4, 9]
G00 = 1+(F0 −1)
(14)
500[ 3 D]
Therefore work to determine the validity of these models
where F0 is the gap factor. This proposed model satisfies using the same voltage type is required.
corrections for altitude up to 3000 m [18].
3. IRREGULARITIES IN THE COMMONLY USED
2.5 Calva Model STANDARDS
Calva et al [9] proposed a model to use for altitude There is a fair number of standards used internationally
correction for gaps under DC voltage stress. This model that involve atmospheric correction voltages. The three
determines the breakdown voltage Vb as notable standards used are the IEC 60060-1(2010) [17],
IEC 60071-2 (1996) [20] and IEEE Std 4 (2013)
Vb = Eso d[k1 + k2 ]S (15) on high voltage testing techniques [21]. However,
these standards have notable irregularities regarding
where k1 is the air density correction factor atmospheric correction that can bring to question which
correction method is reliable. The atmospheric correction
k1 = δm (16) discrepancies in the three above mentioned standards are
highlighted in this section.
and m=1.4 for positive polarity and 0.44 for negative
polarity. k2 is the humidity correction factor 3.1 IEC 60060-1 (2010)
h − 11
k2 = 1.3δ−0.83 (17) This is the standard derived from the work done by Pigini
100
et at [15] and is the most commonly used standard. This
and S is the gap factor. Eso is the electric field required standard defines two methods of atmospheric correction
for streamer propagation and considered to be 500 kV/m namely:
for positive polarity and for negative polarity is given by
equation (18)
1. The standard procedure- This procedure is used to
Eso = 1476.4 × 1121.91d (18) correct voltages from non-standard conditions to
standard conditions using the relationship
The models presented by Ramirez et al [18] and Calva et al
U
[9] were not considered as inputs to the latest version of the U0 = (19)
IEC 60060-1(2010) and it would therefore be of interest Kt
to compare these models. It is now generally agreed that where U0 is the disruptive-discharge voltage
there are shortfalls in IEC 60060-1 with regard to altitude under standard conditions, U is the measured
correction factors. In that regard, a Cigre work group disruptive-discharge voltage under non-standard
(WG D1.50) “Atmospheric and altitude correction factors conditions and Kt is the atmospheric correction
for air gaps and clean insulators” has been established factor [17].
to coordinate further studies and investigation on altitude
correction. It has also been proven by [3] that the 2. The converse procedure- This is an iterative method
IEC 60060-1 is still inconsistent in its calculated results used to correct voltages from standard conditions to
compared to the experimental results obtained as altitude non-standard conditions using the relationship stated
increases for air gaps under DC voltage stress and that the in equation (19) with U as the subject of the formula.
The determination of Kt in [17] involves the calculation (Ucw ) cannot be used [5]. The curves relating m and
of the g-factor to determine the exponents m and w. The (Ucw ) in the standard are for phase-to-earth insulation,
function that relates m to the g-factor is derived from a longitudinal insulation, phase-to-phase insulation and
curve that averages the results obtained during the work rod-plane gap. These cannot fully describe the dielectric
done by [15] of different gap geometries and positive strength and and discharge characteristics of different gap
polarities of different impulse shapes [22]. This gives rise configurations, rod-to-rod for instance. Also, a value for
to the possibility of over-correction or under-correction the m to be used for AC voltages is recommended but no
voltage [22]. With regards to negative polarity impulses, recommendation is given for DC voltage.
the value m=1 was assumed by [15] and incorporated into Even though this standard is assumed to be able to correct
the standard. This implies that the value of m for negative for altitudes greater than 1 800 m, there are irregularities
impulses is independent of the air density. Since the in the determination of m and no recommendation for DC
variation of air density has an effect on the breakdown voltages is given.
voltage, the assumption of a constant value of m is an
approximation that cannot be considered accurate. 3.3 IEEE Std 4
The relationship between m, w and g in [17] is only valid
when δ x k ≈ 1 [5]. This condition is valid for high This standard has two methods for atmospheric correction
voltage laboratories located at altitudes near sea level. [21].
Laboratories at much higher altitudes have the relationship
δ x k < 1 [5] implying that the application of the correction
method in [17] becomes invalid for results obtained in such 1. Method 1- This method uses the factor K and is the
laboratories. same as the one used in IEC 60060-1
For one to be able to use the g-factor, the U50 of the
2. Method 2- This method uses the factors kd (air density
apparatus/gap is assumed to be known. However, for high
correction factor) and kh (humidity correction factor)
voltage apparatus, the type test withstand voltage is the
for correction and is applicable to gaps less than 1 m.
one known and cannot be assumed to be the U50 of the
apparatus [5]. This implies that the g-factor calculated
using the type test withstand voltage is invalid. Also, as Since Method 1 is similar to the one mentioned in IEC
earlier mentioned, the g-factor incorporates the average 60060-1, it will not be discussed again in this section.
electric field strength required for streamer propagation Method 2 stated that the following relationship,
(500 kV/mm), therefore implying that it is more accurate
for streamer dominated discharges. Otherwise, the kd
U= U0 (22)
expression given for the g-factor cannot be taken as an kh
accurate approximation.
These irregularities make the IEC 60060-1 standard quite where U0 and U have the same definition as mentioned in
unreliable for correction of voltages at altitudes greater Section 3.1.
than the maximum altitude at which the tests were The values of m and n used for evaluating kd are assumed
conducted (1 800 m) and for discharges that are not to be the same for all electrode configurations for DC
streamer dominated. voltages and lightning impulse voltages despite the air
density that the configuration is set. For other electrode
configurations and voltage types, the values of m, n and w
3.2 IEC 60071-2 (1996) are determined from a graph that relates the three variables
to the gap length [21]. The values of m, n and w will also be
This standard was based on the same data used to develop the same for the same gap length, configuration. This does
IEC 60060-1. This therefore makes the standard to be valid not agree with the variation of the air density correction
only up to 1 800 m, but is assumed to be applicable for factor with the type of discharge that governs breakdown
higher altitudes. The relationship between the required in the air gap.
withstand voltage (Urw ) and the co-ordination withstand
voltage (Ucw ) is given by equation (20) 4. ONGOING WORK TO RESOLVE THE
DISCREPANCIES IN CURRENT ALTITUDE
Urw = Ucw × Ka (20)
CORRECTION METHODS
where K a is the atmospheric correction factor given by
equation (21) Due to the discrepancies in altitude corrections methods, a
H Cigre work group (WG D1.50) “Atmospheric and altitude
Ka = em( 8150 ) (21)
correction factors for air gaps and clean insulators ”has
where H is the altitude in m. In this standard, the value been established to develop appropriate correction factors
of m is defined as 1 for lightning impulses and short that can be used. This group involves members from
duration AC voltage tests. For switching impulses, the Australia, Brazil, China, Germany, Italy, Japan, South
value of m is taken from a plot of curves that relate m to Africa, Sweden and the USA. Among the work being done
(Ucw ). However,(Ucw ) is defined as the 10% probability by this group, they are looking at deriving a method that
breakdown voltage therefore, when designing insulation uses a single value of m over the g curve for various
for a lower breakdown voltage probability, the value of electrode arrangements. Work is also being done in South
Figure 1: Conceptual set up of experiment
Africa, in collaboration with the Chinese to produce data • The breakdown voltages are then plotted versus gap
to be used by the Cigre group. length and then compared to the ones obtained from
At Wits University, there is ongoing work to simulate the three models to be verified.
altitude conditions by varying the air pressure in a chamber
and doing rod-plane breakdown tests under DC voltages. • Using statistical methods, the model that corrects the
This project looks to add information, especially for voltages to the closest experimental value obtained
DC voltages, to compliment other efforts being done to will be justified.
develop much more accurate correction factors. Currently, • Using the recordings of the breakdown processes
the variable pressure chamber to house the set up is is during the experiment, the model that yields the best
under construction. The gap sizes will be limited to results will also be analysed and justified.
approximately 0.8 m due to voltages supply limitations. A
conceptual diagram of the experimental set up is shown in
Figure 1. The results of the experiment will also be used to A major challenge on how to bring in 300 kV HVDC
validate the models mentioned in the background section into the metallic walled pressure chamber is being faced.
for DC voltage stress as they were mostly developed The bushings required for DC voltages around 300 kV
from data for lightning and switching impulse tests. The are very large, exceeding lengths of 2 m, and may cause
major steps to be conducted during the experiment are structural problems on the chamber. Besides the structural
highlighted below: problem, acquiring an already made bushing can also
be very expensive and very difficult as the bushings are
hard to find. To counter this problem, an HVDC cable
• The first experiment will be conducted under standard can be used. Alternatively, a bushing fit enough to use
pressure. Once the breakdown voltages of the only for the experiments can be constructed in the high
different gap lengths up to 1 m are recorded, a graph voltage laboratory. This option is currently under rigorous
will be plotted of breakdown voltage versus gap investigation.
length.
5. CONCLUSION
• The breakdown voltages recorded in the first
experiment will then be corrected to different The use of HVDC transmission technology in developing
altitudes up to 4 500 m using the IEC 60060-1 countries is on the rise. This has seen transmission
(2010), the Ramirez model and the Calva model. The lines being set up in areas reaching altitudes of 4 500
breakdown voltages at the different altitudes will be m. However, it has been established that the altitude
plotted against the gap lengths. correction factors have discrepancies when applied to
altitudes higher than 1 800 m. Further discrepancies in
• Experiments will be conducted again at the pressure the current altitude correction methods have been laid out
levels that correspond to the air densities of the and further work is required to determine the appropriate
altitudes used to correct the voltages as mentioned in correction factors. A Cigre working group has been set
the previous bullet point. up to work towards resolving this issue and a lot of input
is still required. Some research efforts at Wits University [10] N. L. Allen. “Correction procedures for High
aim at contributing to the knowledge in that regard. voltage measurements under variable air density.”
IEE Proceedings on Science, Measurement and
6. ACKNOWLEDGEMENTS Technology, vol. 153, no. 5, pp. 194–200, September
2006.
The authors would like to acknowledge with gratitude
[11] M. Moreno, A. Pigini, and F. Rizk. “Influence
Eskom for their support of the High Voltage Engineering
of air density on the dielectric strength of external
Research Group through TESP. They would also like to
insulation.” Guidelines for the evaluation of
express gratitude to the Department of Trade and Industry
dielectric strength for external insulation, 1993.
(DTI) for THRIP funding and to thank the National
CIGRÉ Working Group 33.07.
Research Foundation (NRF) for direct funding.
[12] A. Robledo-Martinez and P. A. Calva. “DC Break-
REFERENCES down Characteristics of Rod-Rod and Rod-Plane
Gaps in Reduced-Density Air.” In Sixth International
[1] HVDC Power Transmission Part 1: Basic Principles, Symposium on High Voltage Engineering, vol. 3.
Planning and Converter Technology, vol. 9 of Eskom New Orleans, USA, 1989.
Power Series. Crown Publications, 2012.
[13] M. Boutlendj and N. L. Allen. “Assessment
[2] D. Yujian, L. Weiming, S. Zhiyi, L. Qingfeng, of Air-Density Correction for Practical Electrode
G. U. Chen, and S. Zhaoying. “Switching Impulse Systems.” European Transactions on Electrical
Voltage Flashover Characteristics of Air Gaps in Power, vol. 6, no. 4, pp. 267–274, July/August 1996.
±800 kV UHV DC Transmission Tower and Altitude
Correction.” Cigre A3-104, 2012. [14] H. J. Geldenhuys. “The breakdown voltage of air in
a 50 cm rod-plane gap over a practical range of air
[3] N. Parus, I. R. Jandrell, N. Mahatho, T. Govender, density and humidity.” 5th International Symposium
and H. A. Roets. “Live work under HVDC voltage: on High Voltage Engineering, 1987.
An investigation into altitude effects of flashover
[15] A. Pigini et al. “Influence of air density on the
voltage for rod-plane air gaps.” submitted for
impulse strength of external insulation.” IEEE
publication.
Transactions on Power Apparatus and Systems, vol.
[4] P. A. Calva and F. P. Espino. “Correction Factors PAS-104, no. 10, pp. 2888–2900, October 1985.
for Positive DC Voltages.” IEEE Transactions on
[16] SANS 60060-1. “High-voltage test techniques. Part
Dielectrics and Electrical Insulation, vol. 5, no. 4,
1: General definitions and test requirements.”, 1989.
pp. 541–544, August 1998.
[17] SANS 60060-1. “High-voltage test techniques. Part
[5] J. Rickmann et al. “Analysis and comparison 1: General definitions and test requirements.”, 2011.
of atmospheric and altitude correction methods for
air gaps and clean insulators.” 18th International [18] M. Ramirez, M. Moreno, A. Pigini, G. Rizzi,
Symposium on High Voltage Engineering, , no. and E. Garbagnati. “Air density influence on
OC2-03, pp. 542–547, August 2013. the strength of external insulation under positive
impulses: Experimental investigation up to an
[6] F. W. Peek(Jr). “Effect of altitude on the spark-over altitude of 3000m a.s.l.” IEEE Transactions on Power
voltages of bushings, leads and insulators.” AIEE Delivery, vol. 5, no. 2, pp. 730–737, April 1990.
transactions, vol. 33, pp. 1721–1730, December
1914. [19] K. Feser and J. Schmid. “Temperature effect on
the streamer breakdown of air.” 6th International
[7] T. Udo and Y. Watanabe. “DC High-voltage Symposium on High Voltage Engineering, vol. 3,
Sparkover Characteristics of Gaps and Insulator August/September 1989.
Strings.” IEEE Transactions on Power Apparatus
and Systems, vol. pas-87, no. 1, pp. 266–270, January [20] IEC 60071-2. “Insulation co-ordination- Part 2:
1968. Application Guide.”, 1996.
[8] N. L. Allen, M. Boutlendj, and H. A. Lightfoot. [21] IEEE Std 4-2013. “IEEE Standard for High-Voltage
“Dielectric Breakdown In Nonuniform Field Air Testing Techniques.”, 2013.
Gaps: Ranges of Applicability to dc Voltage
[22] I. Gutman et al. “Atmospheric and altitude
Measurement.” IEEE Transaction on Electrical
correction of air gaps, clean and polluted insulators:
Insulation, vol. 28, no. 2, pp. 183–191, April 1993.
state-of-the-art within CIGRÉ and IEC.” CIGRÉ, ,
[9] P. A. Calva, V. del Moral, M. G. Márquez, and G. P. no. D1-213, August 2014.
Cabrera. “New Proposal of Correction Factors for
DC Voltages.” Conference on Electrical Insulation
and Dielectric Phenomena, pp. 455–458, 2003.
Appendix C
ISH 2015 Publication
This appendix contains a publication handed in to the 19th International Symposium

on High Voltage Engineering in 2015.
79
The 19th International Symposium on High Voltage Engineering, Pilsen, Czech Republic, August, 23 – 28, 2015
Investigating the Effects of Altitude (Air Density) on the HVDC

Breakdown Voltage of Rod-Plane Air Gaps
T. Gora* and C. Nyamupangedengu and I.R. Jandrell
University of the Witwatersrand, 1 Jan Smuts Ave, Braamfontein, Johannesburg, South Africa
*Email: <tatenda.gora@students.wits.ac.za>
Abstract: This paper is on a study of altitude correction factors on short air gap DC breakdown
voltage. Tests were conducted on small rod-plane air gaps in a pressure vessel. The air gaps used
varied between 100 mm and 300 mm in length and pressure was drawn in the pressure vessel to
give relative air densities as low as 0.53. The pressure was drawn from an ambient pressure of
about 835 mbar. The range of pressures used made it possible to simulate relative air densities
corresponding to altitudes of 1700 m to 4500 m. The results obtained showed a decrease in
breakdown voltage as pressure decreased, but only up to a threshold point beyond which an
increase in breakdown voltage at very low pressures was observed. The results were corrected
using IEC 60060-1 and were also compared to predicted breakdown voltages proposed by Calva et
al. The corrected results according to the IEC 60060-1 standard did not resemble the trend of a
decrease in breakdown voltage as relative air density decreases. When compared to the predicted
breakdown voltages according to the method by Calva et al, the experimentally obtained results
did relate to the predicted breakdown voltages. Even though the predicted flashover voltages
were higher, the trend in the breakdown voltage agreed with the ones obtained using Calva et
al but only until the pressure threshold was reached. A potential exists for the method proposed
by Calva to be improved for small air gaps.
1 INTRODUCTION the University of the Witwatersrand, Johannesburg,

ongoing work is being done to contribute towards
reviewing the altitude correction methods. This paper
The desire to meet electricity demands, especially in presents some of the results obtained so far.
emerging economies, has led power utilities to embark Before presenting the results, this paper starts with
on projects that involve the transmission of power a review of the evolution of the altitude correction
from sources that are far away from the load centres. methods.
The number of High Voltage Direct Current (HVDC)
schemes around Europe and Asia and the lengths of
the transmission lines show the extent at which HVDC 2 EVOLUTION OF ALTITUDE CORRECTION
is being used to connect the load centres to remote METHODS: A CRITICAL REVIEW
generation sources [1]. In China, the Xiangjiaba-
Shanghai HVDC link connects the load centre from This section discusses proposed models of altitude cor-
the generation source which is approximately 4 000 km rection factors.
away. The Xiangjiaba-Shanghai link passes through
the Tibet region which is quite mountainous, resulting 2.1 Peek Model
in the line passing through altitudes above 4 000 m.
Research is still being done on insulator performance at Peek [3] investigated the effect of altitude on the spark-
such high altitudes [2]. In Africa, Eskom is considering over voltages of bushings, leads and insulators by plac-
possibilities of importing power from the Congo. This ing them in a cask and varying the pressure. Pressure
project will involve the transmission of power over was varied to obtain relative air densities between 0.5
distances of approximately 4 000 km. For this project and 1.0 while the temperature variations due to the
to be economically viable, the power will have to be pressure changes were recorded. In his research, Peek
transmitted using high voltage direct current (HVDC) [3] was able to produce tables of correction factors to
technology. It has, however, been generally agreed that be used on specific insulators by reading off the relative
the available standards for altitude correction have air density. Therefore, the correction factors for insu-
discrepancies that can hinder the design of a reliable lators were different from those of bushings and leads
system at high altitudes. under the same relative air density [3]. Peek attributed
these to the uniformity of the field suggesting that the
It is generally agreed that there is need to de- flashover voltage for a uniform field decreases directly
velop altitude correction methods applicable to with δ, whereas for non-uniform fields, the flashover
altitudes beyond 1800 m. In that regard, there is a voltage decreases at a lesser rate than δ [3]. This means
Cigré working group (D1.50) that is currently working that given that the flashover voltage of an insulator at
towards the development of such a standard. At sea level (δ=1) is v1 , then the breakdown voltage v at
δ=x was given by where d is the length of the discharge gap and k is a
v = xv1 (1) factor dependent on the type of voltage used and δ is
defined as:
where δ is the relative air density at a specific altitude P 273 + t0
and read from the tables derived by Peek [3]. This δ= (7)
P0 273 + t
therefore implied that the breakdown voltage was di-
To determine n in equation (5), the G factor, equa-
rectly proportional to the air density.
tion (6), was used and n took the following values [5]
depending on the range of values of G
2.2 IEC 60-1
G0 (G0 − 0.2)
In 1973, the IEC 60-1 was amended. In the 1973 version n= (8)
of IEC 60-1, it was stated that the breakdown voltage 0.8
3 − G0
of an air gap or insulator at relative air density δ, uδ is n= (9)
related to the breakdown voltage of the same air gap 2G0
or insulator under standard conditions, u0 by [4] where G0 is the G factor under standard conditions (δ
and k=1). Equation (8) was valid for 0.3≤G≤1 and
uδ = kd u0 (2) equation (9) was valid for 1≤G≤2. These results by [5]
where kd is the air density correction factor defined as were incorporated into the IEC 60060-1 (1989), includ-
ing the graphs of equation (8) and equation (9) as the
P m T0 n air density correction factors [6]. The results still play
kd = ( ) ( ) (3)
P0 T a major role in the recent IEC 60060-1(2010) which,
instead of using graphs for determining G as in IEC
where P and P0 are the pressure and standard pressure 60060-1(1989), uses equations that are quite similar to
respectively while T and T0 are the temperature and equations (8) and (9) [7]. It is important to note that
standard temperature respectively. The exponents m the model derived by Pigini et al takes into account
and n are the variations of pressure and temperature the voltage type through the factor k in equation (6)
respectively that are supposed to be determined before whereas IEC 60-1 (1973) did not include any factor
kd is evaluated. Since the air density is dependent on that took into account the type of voltage used. Since
pressure, the determination of the exponent m became the standard was produced from the work conducted
the area of concern regarding atmospheric corrections by Pigini et al, the standard is limited to an altitude
and was defined in IEC 60-1 (1973) as of 1 800 m. Given that the only difference in the al-
titude correction procedures highlighted in the current
m = −0.12d + 1.12 (4)
version of IEC 60060-1 (2010) with those of IEC 60060-
where d is the air gap length between 1 m and 6 m. 1 (1989) is the use of equations to determine the index
For d≤1 m, m=1.0 and m=0.4 for d≥6 m. The fac- n instead of graphs, it therefore holds that the current
tor m in this standard was therefore independent of the version of IEC 60060-1 is also limited to an altitude of
type of voltage used and the relative air density being 1 800 m.
corrected for. This therefore implied that the variation
of pressure was only dependent on the length of the air 2.4 Ramirez Model
gap. Compared to the model proposed by Peek, the A year after IEC 60060-1 (1989) was published, Ramirez
model in the IEC 60-1 does not give a linear relation- and his counterparts [8] proposed a model claimed to be
ship between the flashover voltage and the relative air more accurate than the one incorporated in IEC 60060-
density. For DC and impulse voltages of either polarity, 1(1989) and takes account of the influence of air density
the standard specifies that m and n can be considered on rod-plane gaps. The model is expressed in equation
equal. (10)
2.3 IEC 60060-1(Pigini Model) U 0.8[1 + T (1 − δ)](δ − 0.2G00 )

= + 0.2 (10)
U0 (1 − 0.2G00 )
The IEC 60060-1 (1989) standard stated that
where δ is the relative air density calculated by equation
U50 = U50(std) (δ.K)n (5) (7) and
1 − 0.8G00
where K is the humidity correction factor, n is the air T = 1.4 (11)
1 − 0.2G00
density correction factor and U50 is the breakdown volt-
age under non-standard conditions. A new factor, G, with G0 defined as
was introduced into this correction. G was determined U0
on the assumption that the major process during the G00 = 1+(F0 −1)
(12)
breakdown of the gap was streamers [5]. G was defined 500[ 3 D]
as where F0 is the gap factor. This proposed model satis-
U50
G= (6) fies corrections for altitude up to 3000 m [8].
500δkd
2.5 Calva Model

Calva et al [9] proposed a method whereby one could
predict the breakdown voltage of an air gap under DC
stress at any relative air density. This model determines
the breakdown voltage Vb as
Vb = Eso d[k1 + k2 ]S (13)
where k1 is the air density correction factor
k1 = δ m (14)
and m=1.4 for positive polarity and 0.44 for negative

polarity. k2 is the humidity correction factor
Figure 2: Picture of experimental set up
−0.83 h − 11
k2 = 1.3δ (15)
100
and S is the gap factor. Eso is the electric field re-
quired for streamer propagation and considered to be and used as the rod electrode. The rod protruded
500 kV/m for positive polarity and for negative polarity into a 550 mm chamber with a borate glass wall with
is given by equation (16) an inner diameter of 120 mm as shown in Figure 3.
To adjust the distance between the rod tip and the
Eso = 1476.4 × 1121.91d (16) grounded plane set at the bottom of the chamber, an
adapter was machined which allowed the user to undo
It is apparent that the limitations of the current alti- the rod, set the appropriate length and then hold the
tude correction methods for altitudes beyond 1800 m rod in place.
need to be addressed. This paper presents results for
tests performed under DC voltage stress for simulated
altitudes between 1700 m and 4500 m.
3 EXPERIMENTAL SET UP
The single line diagram of the set up of the experiment

is shown in Figure 1 while Figure 2 shows a picture
of experimental set up used in the laboratory. The
set up consisted of a 400 kV rated Walton-Cockroft
generator, a 2 MΩ current limiting resistor and the
electrode set up in a test cell.
Figure 3: Picture of chamber and electrode set up
The chamber was equipped with two valves used to

either draw a vacuum or compress air by connecting
Figure 1: Single line diagram of experimental set up either a pump or compressor. Both valves had a gauge
to show how much pressure has been drawn from the
ambient atmospheric pressure or compressed above the
The Walton-Cockroft generator used is equipped with ambient atmospheric pressure. For either a vacuum to
an inbuilt voltage divider connected to an LED to drawn or air to be compressed, nozzles from the valves
display the voltage reading. All voltage measurements were located at the base of the chamber that allowed
were obtained using the same voltage divider. air to flow either way. An Edwards™vacuum pump was
used and the lowest pressure achieved was 300 mbar
An aluminium electrode 700 mm long with a 15 below ambient.
mm diameter flat tip was connected to the resistor
4 THE TEST PROCEDURE corona became more prominent. For breakdown points
beyond the knee point, the pre-breakdown corona was
The gap length was varied between 100 mm and 300 observed to periodically appear and disappear before
mm. Air gaps larger than 300 mm could not be tested the complete breakdown of the air gap.
due space limitations in the vessel. In order to adjust It has already been established that the breakdown volt-
the air gap length to a required value, the rod was ini- age of an air gap is dependent on the mean free path
tially set in such a manner so that the rod tip would between the air molecules [10]. As the pressure de-
be in contact with the ground plate. A zero point was creases, the mean free path between the molecules in-
then marked on the rod. When the rod is pulled up, the creases allowing greater mobility for there to be ioni-
distance between the zero mark and the cap surface of sation with less energy required compared to when the
the vessel would be equivalent to the length between mean free path is small. However, there is a point
the rod tip and the ground plate. whereby the quantity of molecules present is so small,
Once the required gap length was set, tests were first reducing the probability of ionization. This whole pro-
done at the ambient atmospheric pressure. Voltage was cess is governed by Townsend’s first ionization coeffi-
increased at a rate of approximately 2 kV/s until there cient, α, which is a function of the electric field (E) and
was breakdown in the air gap. Pressure was then re- the pressure (p) according to equation (17)
duced by 100 mbar and the voltage was raised again
Bp
until breakdown occurred. The process would then be α = Ape− E (17)
repeated for the specific gap length until the test pres-
sure was 300 mbar below the ambient pressure. Each where A and B are constants dependent on the
breakdown test was repeated five times per gap length temperature and gas type [11]. This relationship states
per pressure level. that at very low and very high pressures, Townsend’s
Before increasing the voltage, the high speed camera first ionization coefficient decreases therefore implying
was manually triggered to start recording and monitor an increase in the breakdown voltages. It is believed
any visible activities that could occur before breakdown. that this may be the cause of the increase in the
breakdown voltage as pressure continues to decrease.
5 RESULTS AND DISCUSSION

When applied to altitude correction factors, the
The five breakdown voltages obtained for each pressure breakdown voltage trend observed may be of major
level and gap size were averaged. The results of the concern. A factor may need to be applied to cater
average breakdown voltages against relative air pressure for the increase in voltage at very low relative air
for each gap size are shown in Figure 4. densities and small air gaps. The breakdown voltages
obtained during the testing were corrected to standard
conditions using IEC 60060-1 (2010) [7]. The method
proposed by Calva [9] which can be used to predict
the breakdown voltage of an air gap at any relative air
density was also used and the results obtained were
compared to the experimental results. Figure 5 to
Figure 7 show the obtained results corrected using [7].
Figure 4: Plot of experimental DC flashover voltages
It was noted that the breakdown voltages of the air

gaps decreased with a decrease in pressure (relative air
density), but only to a knee point. After a threshold in
pressure which seems to be related to the gap length,
the breakdown voltage begins to increase with a further
decrease in pressure. The pressure threshold becomes
lower with an increase in gap length. It was also ob- Figure 5: DC breakdown voltage vs relative air density
served that as the pressure decreased, pre-breakdown for 100 mm air gap
The method proposed by Calva [9] to determine

the breakdown voltage at any relative air density was
applied and the results were plotted on the same
axis as the uncorrected data. The plots are shown in
Figures 8 - 10.
Figure 6: DC breakdown voltage vs relative air density

for 200 mm air gap
Figure 8: Comparison of results to method proposed

by Calva for 100 mm air gap
Figure 7: DC breakdown voltage vs relative air density

for 300 mm air gap
Ideally, the correction factors applied to the experimen-

tal results were supposed to map the experimental plot
to a straight horizontal line that corresponds to the
actual breakdown voltage under standard conditions. Figure 9: Comparison of results to method proposed
However, the IEC 60060-1 correction method maps the by Calva for 200 mm air gap
results onto a diverging graph above the breakdown
voltage under standard conditions. This confirms the The comparison of the experimentally obtained break-
limitations in the standard for results obtained for air down voltages to the predicted breakdown voltages ac-
densities less than 0.8. The same trend was observed cording to the method by Calva only agree in trend
for the 200 mm and 300 mm air gaps as shown in that as relative air density decreases, the breakdown
Figures 6 and 7. The failure of the IEC 60060-1 stan- voltage for a specific gap length also decreases. The
dard to correct at higher altitudes can be attributed trend was observed to agree more as the gap length
to the fact that the IEC 60060-1 standard was derived increases. This trend is only consistent up to the knee
empirically using data obtained at altitudes up to point where divergence begins. It was also observed
1 800 m [5]. Therefore, the changes in breakdown that the method proposed by Calva predicted the volt-
mechanisms that occur at higher altitudes (thus lower age to be less than the experimental results. By as-
relative air density) were not incorporated into the sessing the way in which the trend agrees, there is
standards. a potential to improve Calva’s method to predict the
breakdown voltages for small air gaps at altitudes higher
REFERENCES
[1] HVDC Power Transmission Part 1: Basic Prin-

ciples, Planning and Converter Technology, ser.
Eskom Power Series. Crown Publications, 2012,
vol. 9.
[2] D. Yujian, L. Weiming, S. Zhiyi, L. Qingfeng, G. U.
Chen, and S. Zhaoying, “Switching impulse volt-
age flashover characteristics of air gaps in ±800
kv uhv dc transmission tower and altitude correc-
tion,” Cigre A3-104, 2012.
[3] F. W. Peek(Jr), “Effect of altitude on the spark-

over voltages of bushings, leads and insulators,”
AIEE transactions, vol. 33, pp. 1721–1730, De-
cember 1914.
Figure 10: Comparison of results to method proposed [4] M. Boutlendj and N. L. Allen, “Assessment of air-
by Calva for 300 mm air gap density correction for practical electrode systems,”
European Transactions on Electrical Power, vol. 6,
no. 4, pp. 267–274, July/August 1996.
than 1800 m by incorporating a factor that can allow
the Calva method to precisely predict the breakdown [5] A. Pigini et al., “Influence of air density on the im-
voltage of a small air gap. pulse strength of external insulation,” IEEE Trans-
actions on Power Apparatus and Systems, vol.
PAS-104, no. 10, pp. 2888–2900, October 1985.
6 CONCLUSION
[6] SANS 60060-1 , “High-voltage test techniques.
Laboratory experiments were conducted using a pres- part 1: General definitions and test requirements,”
sure vessel to simulate different relative air densities and SABS, Pretoria, South Africa, 1989.
DC breakdown tests were done in the simulated con-
ditions. Rod-plane gaps of sizes between 100 mm and [7] SANS 60060-1, “High-voltage test techniques.
300 mm were used. From the results, it is confirmed part 1: General definitions and test requirements,”
that the current IEC 60060-1 standard does not apply SABS, Pretoria, South Africa, 2011.
for altitudes above 1800 m. The method proposed by
Calva, however, is more accurate in predicting break- [8] M. Ramirez, M. Moreno, A. Pigini, G. Rizzi,
down voltages at high altitudes and as far as trend is and E. Garbagnati, “Air density influence on the
concerned. The error in the trend becomes less with an strength of external insulation under positive im-
increase in gap length. pulses: Experimental investigation up to an alti-
tude of 3000m a.s.l.” IEEE Transactions on Power
ACKNOWLEDGMENTS Delivery, vol. 5, no. 2, pp. 730–737, April 1990.
The authors sincerely acknowledge the ISH 2009- [9] P. A. Calva, V. del Moral, M. G. Márquez, and
SAIEE Postgraduate Research Scholarship in High G. P. Cabrera, “New proposal of correction fac-
Voltage Engineering for directly funding this research tors for dc voltages,” Conference on Electrical In-
work. The authors would also like to acknowledge, with sulation and Dielectric Phenomena, pp. 455–458,
gratitude, Eskom for their support of the High Voltage 2003.
Engineering Research Group through TESP. Authors [10] E. Kuffel, W. S. Zaengl, and J. Kuffel, High Volt-
would also like to express gratitude to the Department age Engineering: Fundamentals, 2nd ed. Oxford,
of Trade and Industry (DTI) for THRIP funding and Boston: Butterwoth-Heinemann, 2000.
to thank the National Research Foundation (NRF) for
direct funding of the High Voltage Research Group of [11] F. Kreuger, Industrial High Voltage 1, 2, 3. Delft,
the University of the Witwatersrand, Johannesburg. Netherlands: Delft University Press, 1991.
Appendix D
Cigré WG D1.50 Contribution
This appendix contains a contribution to the Cigré WG D1.50 Task Force on short air
gaps.
86
Contribution to Cigré D1:50 - short gap
task group
__________________________________________________________________________________
Prepared for Cigré D1-50 by Tatenda Gora and Cuthbert Nyamupangedengu; University of the
Witwatersrand, Johannesburg, South Africa.
1. Introduction
This report presents the results obtained in South Africa for short rod-plane gaps of lengths up to 0.4
m. The tests were to verify the effects of pressure (thus relative air density) on the breakdown voltage
on short gaps. Two types of tests were conducted in this work:
 Open air tests - where simple rod-plane tests were conducted in laboratories situated at 1 700
m above sea level (asl) and 210 m asl for high and low altitude conditions respectively.
 Pressure vessel tests- where rod-plane tests were conducted in a vessel under reduced pressure
so as to simulate different relative air densities.
Some of the results were published during the recently held ISH 2015 conference in Pilsen [1]. All the
tests conducted in this work were for positive polarity HVDC.
2. Brief description of experimental set up and procedure

2.1 Open Air Laboratory Test
Experimental Set up
The open air laboratory tests were conducted at altitudes of 210 m and 1 700 m respectively. A
diagram of the experimental set up is shown in Figure 1. A 15 mm copper rod with a fitted flat
end cap was used as the rod electrode in the experiment. The same electrode was used for both
high and low altitude experiments.
Figure 1: Diagram of open air experimental set up

Figure 2: Image of test set up at 1 700 m asl
Experimental Procedure
The air gap length during the tests was adjusted in steps of 0.05 m from 0.10 m to 0.40 m. For every
gap length, 5 breakdowns were conducted by slowly increasing the voltage. Between each breakdown,
a 5 minute window was observed to allow any accumulated space charge to dissipate. The voltage
divider of the HVDC generator was utilised for recording the breakdown voltages.
2.2 Pressure Vessel Tests

Experimental Set Up
The pressure vessel tests were conducted in a 0.55 m cylindrical tube with a 0.12 m inner radius borate
glass wall. The vessel could only vacuum pressure due to alterations made to the lid so that a rod could
be fitted. An aluminium rod machined to the same tip dimensions as the copper rod used in the open
air tests was used as the rod electrode. A diagram of the experimental setup is shown in Figure 3
accompanied by a picture of the laboratory setup in Figure 4.
Experimental Procedure
Due to limitations of the pressure vessel, the maximum air gap size tested was 0.3 m. The air gap size
was adjusted in 0.1 m intervals from 0.1 m to 0.3 m. The procedure was as follows:
 Set gap length and record ambient pressure, temperature and humidity
 At ambient pressure, conduct 5 breakdown experiments by and record results. Observe 10
window to allow space charge dissipation between each breakdown.
 By maintaining the gap length, draw 100 mbar pressure from the pressure vessel. Record
temperature and humidity and repeat the step in the previous bullet. This was done until
breakdown results at 300 mbar below ambient pressure were obtained.
 Increase gap length by 0.1 m and repeat bullets 1 to 3 until results for 0.3 m are obtained.
Figure 3 : Pressure vessel experimental set up
Figure 4: Pressure vessel laboratory set up

3. Results
3.1 Open air tests in laboratory
Low altitude results (210 m asl)
Table 1 and Figure 5 show the results obtained at low altitude. There is a close relationship between the
IEC corrected and uncorrected results as expected since the tests were done close sea level.
Table 1: Low altitude open air test results
Gap Temperature Pressure Humidity Uncorrected IEC 60060- % Variance

Length (0C) (mbar) (%) breakdown 1 (2010)
(m) voltage (kV) Corrected
breakdown
voltage
(kV)
0.10 27.6 987.8 48.6 56.92 59.61 4.1
0.15 26.2 986.3 56.0 82.24 84.93 3.2
0.20 27.8 986.1 51.8 107.62 111.61 3.6
0.25 26.5 985.4 54.8 131.48 136.14 3.4
0.30 27.2 985.2 54.9 157.1 162.04 3.0
0.35 24.9 985.8 56.0 174.92 181.94 3.9
0.40 25.5 985.5 46.0 196.76 210.23 6.4
Figure 5: Plot of IEC 60060-1(2010) corrected and uncorrected laboratory gap tests at 210 m asl
High altitude results (1 700 m asl)
Table 2 and Figure 6 show the results obtained for the open air tests at high altitude of 1 700 m asl. The
variance between the corrected and uncorrected at high altitude is bigger than at low altitude implying
that the correction factors are more accurate at lower altitudes than higher altitudes. Figure 6 also
confirms the trend.
Table 2: High altitude open air test results
Gap Temperatur Pressur Humidi Uncorrecte IEC 60060-1 % Variance

Length e (0C) e ty (%) d (2010)
(m) (mbar) breakdown Corrected
voltage (kV) breakdown
voltage (kV)
0.10 15.3 844.7 41 52.50 64.43 18.5
0.15 14.4 841.7 45 67.75 82.77 18.1
0.20 15.4 840.2 34 81.18 102.81 21.0
0.25 14.9 841.1 42 100.15 122.89 18.5
0.30 15.1 842.3 46 118.2 140.33 15.8
0.35 15.8 839.7 33 129.92 163.04 20.3
0.40 14.8 842.3 47 156.8 187.09 16.2
Figure 6: Plot of IEC 60060-1(2010) corrected and uncorrected laboratory gap tests at 1700 m asl
Comparison of corrected low and high altitude air test results
Table 3: Comparison of High and low altitude corrected results
Gap length (m) High altitude Low altitude Percentage absolute

corrected voltage corrected voltage error (%)
(kV) (kV)
0.10 64.43 59.61 8.09
0.15 82.77 84.93 2.54
0.20 102.81 111.61 7.88
0.25 122.89 136.14 9.73
0.30 140.33 162.04 13.40
0.35 163.04 181.94 10.39
0.40 187.09 210.23 11.01
Figure 7: Comparison of IEC 60060-1(2010) corrected results for high and low altitude
Remarks:
 The two plots in Figure 7 are supposed to coincide, or within at least 5% error, due to the fact
that they are being corrected to the same standard conditions. This is what the IEC 60060-1
(2010) correction procedure is supposed to achieve ideally.
 Trend in plot leads to speculation that the bigger the air gap, the larger the error to be
observed
 There is a need for baseline results obtained at sea level (0 m asl) at standard temperature
(20 0C) to compare how the correction varies from the actual baseline.

Historia de Las Correcciones de Aire

Uploaded by

Copyright:

Available Formats

Historia de Las Correcciones de Aire

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Historia de Las Correcciones de Aire

Uploaded by

Copyright:

Available Formats

University of the Witwatersrand, Johannesburg

Investigating the Effects of Altitude (Air Density) on the HVDC

Author: Tatenda Gora

List of Tables viii

2 Historical evolution of Altitude Correction Methods 4

3 Breakdown Mechanisms in Air 18

4.2.2 Equipment Used in the Tests . . . . . . . . . . . . . . . . . . . . . 26

5 Experimental Results, Analysis and Discussion 33

6 Conclusions and Recommendations 62

A Experimental Data And Results 69

B SAUPEC 2015 Discussion Paper 72

C ISH 2015 Publication 79

D Cigré WG D1.50 Contribution 86

2.1 Summary of the development of altitude correction methods in IEC 60060-

3.1 Electric field distribution in a rod-to-plane gap . . . . . . . . . . . . . . . 19

4.1 Open air test set-up in the HV lab at Wits University . . . . . . . . . . . 26

5.1 Sea level test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

5.18 Comparison of experimental results with Calva’s prediction method for

4.1 Variation of temperature due to pressure in the chamber . . . . . . . . . . 31

5.1 Environmental conditions during sea level experiments . . . . . . . . . . . 34

A.1 Low Altitude Experimental Data . . . . . . . . . . . . . . . . . . . . . . . 69

asl Above Sea Level

It is already known that changes in environmental conditions (temperature, humidity

The structure of this dissertation is as follows:

• Chapter 2 presents the evolution of the atmospheric correction methods. This

• Chapter 3 is a theoretical section on the mechanisms of the breakdown processes

• Chapter 6 concludes the dissertation and presents recommendations for future

• Appendix A contains the tables of all the experimental test data.

• Appendix B is a copy of a discussion paper presented at the SAUPEC 2015 by

• Appendix C is a copy of a paper published in the proceedings of the International

• Appendix D is a report submitted as a contribution to the Cigré WG D1.50 task

Historical evolution of Altitude

2.1 Peek Model

Peek (1914) investigates the variation of flashover voltage with

Pigini et al (1985) introduce the G-factor that determines the value of

Figure 2.1: Summary of the development of altitude correction methods in IEC

2.2 IEC 60-1 (1973)

where kd is the air density correction factor defined as

and kh is the humidity correction factor defined as

The evaluation of m, n, w and the restriction of the correction method to specific

2.3 IEC 60060-1:1989(Pigini Model)

U50 = U50(std) (δ.K)n (2.8)

where U is the non-standard breakdown voltage in kV, U0 is the standard breakdown

2.4 Ramirez Model

is expressed in equation (2.17)

U 0.8[1 + T (1 − δ)](δ − 0.2G00 )

2.5 Calva Model

Vb = Eso d[k1 + k2 ]S (2.20)

where k1 is the air density correction factor

2.6 Some Irregularities In The Commonly Used Standards

2.6.1 IEC 60060-1 (2010)

where U0 is the disruptive-discharge voltage under standard conditions, U is the

2.6.2 IEC 60071-2 (1996)

Urw = Ucw × Ka (2.25)

where K a is the atmospheric correction factor given by equation (2.26)

2.6.3 IEEE Std 4

Breakdown Mechanisms in Air

3.1 Non-Uniform Field Breakdown

Critical field line