You are on page 1of 26

A Technical Report

ON

IMPROVING THE ACCURACY OF SF6 LEAKAGE DETECTION FOR HIGH


VOLTAGE SWITCHGEAR
Is submitted in the particular fulfillment for the award of the degree of
BACHELOR OF TECHNOLOGY
IN
ELECTRICAL AND ELECTRONICS ENGINEERING
By
ANJUM

11B71A0202

UNDER THE GUIDANCE OF

Mr. U.KRISHNA PRASAD


ASST.PROFESSOR

DEPARTMENT OF ELECTRICAL & ELECTRONICS ENGINEERING


SINDHURA COLLEGE OF ENGINEERING & TECHNOLOGY
Approved by AICTE New Delhi, affiliated to JNTUH, Hyderabad.
Medipally, Godhavarikhani, Ramagundam (M), Karimnagar (D), T.S.
During the academic year 2011-2015.

SINDHURA COLLEGE OF ENGINEERING & TECHNOLOGY


(Approved by AICTE New Delhi, affiliated to JNTUH, Hyderabad)

DEPARTMENT OF ELECTRICAL & ELECTRONICS ENGINEERING

CERTIFICATE
This is to certify that the technical report entitled IMPROVING THE
ACCURACY OF SF6 LEAKAGE DETECTION FOR HIGH VOLTAGE
SWITCHGEAR is submitted in the particular fulfillment for the award of the degree of
BACHELOR OF TECHNOLOGY in ELECTRICAL AND ELECTRONICS
ENGINEERING.
By
ZEBA ANJUM

11B71A0202

Bonafide students of SINDHURA COLLEGE OF ENGINEERING & TECHNOLOGY


During the academic year 2014-2015

Project guide

Head of the Department

Mr. U.KRISHNA PRASAD

Mr. J.MADHUKAR REDDY

DECLARATION

We the students of B.Tech in Electrical & Electronics Engineering, Sindhura


college of Engineering & Technology, Ramagundam, hereby declare that the technical report
entitled IMPROVING THE ACCURACY OF SF6 LEAKAGE DETECTION FOR
HIGH VOLTAGE SWITCHGEAR is the original work carried out by us to the best of my
knowledge and belief. We hereby declare that this mini project bears no resemblance to any
other project submitted at Sindhura college of Engineering & Technology, Ramagundam or
any other colleges affiliated JNTUH for the award of the degree.

By
ZEBA ANJUM

11B71A0202

ACKNOWLEDGEMENT

We are very grateful to Mr. R.NARAYAN DAS, Principal, Sindhura College of


Engineering & Technology for providing the required facilities in the college campus.
We express our sincere thanks to Mr. J.MADHUKAR REDDY, Associate Professor
& Head of the Department of Electrical and Electronics Engineering for the constant
cooperation and constructing, criticism, throughout the project.
We express our sincere thanks to our guide Mr. U.KRISHNA PRASAD, Assistant
Professor for his valuable guidance, involvement and the interest shown by him on me has
been the main inspiration for the successful completion of the project.
We would also thank all the staff of Department of Electrical & Electronics Engineering and
Project Review Committee (PRC) members, who are helped me directly or indirectly for the
successful completion of the project.
We earnestly thank my Parents, Family and Friends for their constant encouragement and
moral support, which made the project work successful.

By
ZEBA ANJUM

11B71A0202

ABSTRACT

Even though average leakage rates of SF6 in high voltage equipment are very low,
there is no guarantee for a 100% leakage-free system, especially when considering the
generally long lifespan of high-voltage equipment. Taking in view the rather strict maximum
allowed leakage rate of SF6 per year, regulated by international standards, these facilities
should be equipped with suitable monitoring devices. However, accurate detection of low
rates of gas leakage is not an easy task. This paper sheds light on the challenges of leakage
detection by thorough analysis of physical aspects. Furthermore, it proposes two methods for
high-fidelity leakage monitoring and detection with special emphasis on their suitability for
use in substations. Application of these methods in the laboratory environment revealed an
accuracy of detection levels of 0.3% per year. This is a substantial improvement from the
present monitoring systems that show fluctuation levels of the indicator to be as high as up to
several percent, which is clearly above the acceptable level.

INDEX

TITLE

PAGE NO

1. INTRODUCTION
2. STATE OF ART OF LEAKAGE DETECTION
3. FLUCTUATION OF GAS DENSITY AND PRESSURE
3.1 ORIGIN OF FLUCTUATION
3.2 GAS PRESSURE
3.3 GAS DENSITY
4. SELECTION OF THE QUANTITY TO MEASURE

5.

6.

7
8
10
10
10
10
13

4.1 LEAKAGE DETECTION BY DENSITY MEASUREMENT

13

4.2 LEAKAGE DETECTION BY PRESSURE MEASUREMENT


LABORATORY RESULTS
5.1 FOR DENSITY MEASUREMENT
5.2 FOR PRESSURE MEASUREMENT
CONCLUSION
REFERENCES

16
21
21
23
25
26

1. INTRODUCTION
SF6 gas is widely used in the high-voltage (H.V.) electrical industry because it is an
excellent dielectric and has a very good electrical arc-switching characteristic. This gas was
6

first synthesized in 1900. It is used inside circuit breakers to extinguish the electrical arc
when the contacts inside the circuit breaker are pulled apart, thus enabling the electrical
circuit to be broken. The gas is also a very good insolent to withstand electrical dielectric
stresses. It is not the only gas that can be used, but it has become the industry's standard over
the past 30 years as it allows equipment to be much smaller and more efficient. The IEC
(International Electrotechnical Commission) standard 62271-203 specifies the maximum
allowable leakage rate of SF6 in gas-insulated high voltage equipment as 0.5% per year (pro
anno, p.a.). This implies the necessity to measure or even monitor continuously such small
rates of gas leakage.
Reliable monitoring also makes high demands on measurement accuracy and long-term
stability of the sensor chosen. Gas-insulated equipment has shown a remarkably high degree
of reliability, including a very low level of SF6 leakage rates. Current gas-insulated
switchgear (GIS) designs guarantee a leakage rate of significantly below 0.5% p.a. when new.
But then, the long time span in service needs to be considered. Literature shows in certain
cases an increase of leakage rate in GIS after two or three decades of use. This is mainly
caused by corrosion of flanges and deterioration of rubber seals. An important motivation for
the rigid maximum allowed leakage rate is assumed to originate in environmental aspects
with SF6.
Besides its excellent properties for high voltage application, it is also known as being the
strongest greenhouse gas with a global warming potential (GWP) of 22 800 on a time interval
of 100 years. Abandonment of SF6 in high-voltage equipment by replacing GIS with airinsulated switchgear (AIS) would even lead to a net increase of greenhouse gas due to higher
ohmic losses and elevated material usage for AIS compared with GIS. A mixture of 30%
CF3I and 70% CO2 delivers an insulation performance of about 75% to 80% of SF6.On the
negative side, discharges in CF3I produce iodine, which reduce the dielectric strength. For
high-voltage circuit breakers, there is no replacement gas yet that delivers the same
performance as SF6 without having important disadvantages. The goal of this paper is to
better understand the challenges and improve the quality of SF6 leakage monitoring. It
reviews current technology and pinpoints its limitations, characterizes the thermal-induced
pressure and density fluctuations, and proposes two methods to improve leakage detection.

2. STATE OF ART OF LEAKAGE DETECTION


There are two types of instruments for SF6 monitoring available that address leakage
detection. The first measures the concentration of SF6 in the air. This is useful for indoor
7

applications as it ensures personal safety by sending out an alarm if the maximum safe SF6
concentration is exceeded. According to material safety data sheets [18] the threshold limit
value (TLV) of SF6 in ambient air is 1000 ppmv (parts per million by volume) in Europe and
the USA. There are several measurement principles. One of them is infrared spectroscopy,
which is based on the increased absorption of light of certain wavelength in presence of SF6.
However, this type of monitoring cannot measure the individual rate of leakage of each GIS
gas compartment.

The second type of SF6 monitoring device measures the gas content in each individual SF6
volume. Therefore, a pressure or a density sensor records the (continuous) value of pressure
or density in short intervals and compares it with a threshold value. A more basic variant is
the pressure switch that deactivates the drives of circuit breakers when the pressure goes
below a certain limit so as to avoid damage to the asset. Figure 1 shows a continuously
recorded SF6 density in a high-voltage circuit breaker over 365 consecutive days. The visible
fluctuation in the measurement is a result of the inhomogeneous, non stationary temperature
field in the gas volume. The inhomogeneity in the temperature field originates from heating

certain components through ohmic losses as well as in time-varying ambient conditions such
as ambient temperature and solar irradiation.

3. FLUCTUATION OF GAS DENSITY AND PRESSURE

3.1 ORIGIN OF FLUCTUATION:


Natural fluctuations of pressure and density originate in the changing temperature
distribution in the gas volume. The main factors that impact the non homogeneous and timevarying temperature distribution are ohmic heating of conductors carrying substantial current
as well as external impact due to changing ambient temperature, solar irradiation, wind, and
precipitation. For indoor GIS, ambient conditions can be assumed to be fairly constant, which
therefore do not contribute to fluctuations. This results in reduced complexity of the physical
problem but does not reduce the level of fluctuation as load current-induced fluctuation
seems to be the most pronounced of all according to experiments in the laboratory.
3.2 GAS PRESSURE:
According to Amonton's Law [22] for ideal gases and a constant volume, the pressure
p is proportional to the average temperature T for a constant amount of substance n and a
constant volume V :

Strictly speaking, the condition for an ideal gas is not fulfilled. Nevertheless, practice has
shown good agreement with Amonton's law as long as there is no condensation of SF6. The
surface temperature of certain elements in GIS can reach 90 C (363 K), which corresponds
to an increase of 23.9% compared to the ambient temperature (293 K). Even though some
elements become that hot, the average gas temperature does not even come close to this
value. Figure 2 (top curve) shows the pressure during a 24 h experiment on a horizontal 550
kV GIS bus-bar element. The pressure at rest (no current; ambient temperature) is equal to
100%. During the experiment, it exceeded 113%, which is equal to a pressure rise by 13%.
The pressure is a measure for the average gas temperature.
3.3 GAS DENSITY:
Gas density has a major advantage over pressure if it comes to low dependency on
temperature. Average gas density is defined as the ratio of mass m by volume V, both being
independent of temperature:

10

However, most density sensors measure the local gas density


of At the location r

( r )

at the location

of the sensor, which is generally different from the average value

. Assuming ideal gas conditions, the local density can be expressed by gas

11

Equation (5) demonstrates that for any given time the temperature field is inversely
proportional to the density field. Therefore, it is necessary to calculate or measure only one of
the distribution functions, while the other can be easily derived by inversion. An increase of
local density by 5% is equal to a decrease of temperature by 5% at the same location.

12

4. SELECTION OF THE QUANTITY TO MEASURE


Here two quantities are recommended for the precise measurement/monitoring of
leakage rate: Gas pressure and gas density. Both of them have their inherent advantages and
disadvantages (Table 1). Both quantities correlate well with the leakage rate, which is a
prerequisite. Although pressure shows more pronounced noise, fluctuation caused by
temperature (equation 1), it benefits from a large selection of available sensors that show a
high measurement accuracy. Furthermore, pressure is independent of the exactlocation of the
sensor.

It can be concluded that both quantities, pressure and density, have their specific
benefits and drawbacks. The following two sections show how to make the most of each to
meet the requirements in accuracy.
4.1 LEAKAGE DETECTION BY DENSITY MEASUREMENT:
Most impacting drawback of gas density measurement is the limited accuracy of
available density sensors. Thus, some explanations shall be given for the underlying
technology. Most sensors are based on one of the following two main principles: temperature
compensated pressure measurement and measurement of viscosity, the latter also with
compensation of temperature to a smaller extent. Density sensors, based on pressure
measurement, also incorporate a temperature measurement to compensate for the effect of
different gas temperatures (equation (5)). Gas temperature is typically measured at the
location of the sensor, whereas it finally outputs the gas density at the very same location
(combination sensor). While pressure measurement can be performed with excellent
accuracy, it is more difficult to measure the gas temperature accurately.
13

The built-in temperature sensor is mechanically connected with the housing of the
sensor. This generally results in heat flux through this connection. In most cases, the sensor
housing is approximately on the same temperature potential as the SF6 tank, whereas the
actual sensor is on the local gas temperature potential. Experimental tests by a manufacturer
of GIS confirmed the limited accuracy of this type of density sensor. However, future
generations of density sensors are expected to address these limitations to deliver
substantially higher accuracy. The second type of density sensor is based on the relationship
between dynamic viscosity and density of a specific gas (SF6) at a specific temperature. The
viscosity is measured by the deviation of the oscillation frequency of a quartz oscillator
whose quartz is resonating in the gaseous medium to be measured. Therefore, the sensor
contains two identical quartz oscillators where one of the quartz crystals is in vacuum
(oscillation frequency fr) and the second is in the medium to be measured ( fm). The density is
internally calculated based on an empirical formula, given in the datasheet from the
manufacturer of a certain sensor:

Furthermore, the dominant error seemed to be a constant offset, which could be easily
avoided by relative measurement where only the change from the initial value is evaluated. It
can be concluded that relative measurement is able to increase significantly the accuracy and
is therefore recommended.
DETERMINATION OF THE OPTIMAL LOCATION
The non-uniform and time-varying density distribution yields temperature fluctuation
(equation (5)) at the location of the sensor. These locations would be preferable for a density
sensor if it comes to minimum fluctuation. Computational fluid dynamics (CFD) simulation
is an appropriate tool to find such locations. Figure 9 shows such a simulated steady-state
density field for a horizontal GIS bus-bar element with nominal load current, constant
ambient temperature (20C), and without incident solar radiation. Even though this
simulation was steady state only, it can be assumed that density fluctuation will be high at the
bottom position (BOT in Figure 9). The location with minimum fluctuation is on the side at
half height (SIDE). Density recording simultaneously at all four locations with hair crosses
14

confirms this interpretation (Figure 2). Measurement as in Figure 2 would be an alternative


method to determine the optimal location with respect to density fluctuation. Interestingly,
this seems to be of universal character: In all tested configurations (GIS bus-bar horizontal,
vertical, or slightly askew) and all tested impact factors (load current, incident solar radiation,
and their combination) the density fluctuation is minimal or close to minimal at half height
(SIDE).

It can be concluded that leakage monitoring equipped with density sensors can be improved
by an optimal choice of the sensor location. This location can be determined by CFD
simulation or measurement. As a rule of thumb, a location at half height of the gas volume is
a good choice for minimum density fluctuation. In real applications, the location of least
density fluctuation is not always easily accessible. Such a case would be an SF6 bushing
where a location in the grounded socket would be chosen for practical reasons. For vertical
bushings, this would be a disadvantageous location due to the significant level of density
fluctuation. In such cases, implementing pressure monitoring as described later may be
considered.

15

4.2 LEAKAGE DETECTION BY PRESSURE MEASUREMENT:

MODEL-BASED FILTER
Pressure measurement might look less advantageous for an accurate leakage
monitoring system because pressure shows even more pronounced fluctuation than density.
Assuming pressure fluctuation is deterministic, it can be removed from the signal utilizing a
filter based on first principles (i.e., thermodynamics, fluid dynamics, radiation law, and
electrodynamics). Therefore, an internal model evaluates the ambient conditions and
calculates the prospective gas pressure. Deterministic fluctuations in the measured pressure
cancel out if it is compared to this internally calculated value instead of comparison with an
absolute number. Such a model-based filter approach can also be developed for density
instead of pressure. The problem with such an approach would be the high complexity of a
density model as it will need to calculate the exact temperature distribution in the gas. That
task would be much more complex than for a pressure model where the model only needs to
calculate the average gas temperature and not its exact distribution.

Figure 13 illustrates the model-based method for leakage detection based on pressure.
Therefore, the gas volume to monitor is equipped with a pressure sensor (Block 1). The
model calculates the prospective pressure in parallel (Block 2). For this purpose, this sub
model needs the surface temperature distribution as its input parameter. This is either
16

calculated by means of a thermal-network model (Block 3) or measured directly by sensors


mounted on the GIS (dashed arrow). The difference between the calculated pressure value
and the pressure reading from the sensor is the real pressure deficit (assuming the predicted
value is accurate enough). In practice, this difference also contains all nondeterministic
pressure fluctuation as well as inaccuracies from both the model and the sensor.

DETERMINATION OF TEMPERATURE DISTRIBUTION ON THE


SURFACE
There are basically two different approaches to determine the temperature distribution on
the surface of the gas vessel: by measurement or by calculation. Measurement can be done by
noncontact methods using pyrometers or by sensors mounted on the gas vessel, e.g.,
thermocouples or resistance thermometers. One disadvantage of the determination of
temperature through measurement is the necessity to modify existing hardware. It also
increases costs and causes a potential source of malfunction. Additionally, it is technically
demanding for parts that are not on ground potential such as live-tank circuit breakers,
bushings, or active parts in the GIS.
On the other hand the calculation of the temperature distribution can be very demanding
if it is based on CFD. A substantially more efficient calculation method is based on thermalnetwork models [25]. Besides their computational efficiency, there is a second reason why
thermal network models are particularly useful for this purpose: Some manufacturers of GIS
develop such models of their products in order to provide a tool allowing dynamic calculation
of maximum load under certain ambient conditions. This could substantially reduce the effort
to develop a SF6 monitoring system. A thermal network model builds on the analogy of
thermal and electrical fields. Therefore, thermal fields can be modelled in a similar way as
electrical fields, i.e., by a network consisting of lumped elements. Temperature potentials can
be modelled by voltage sources, radiation and ohmic losses by current sources, thermal
conductivity by resistors, and heat capacity by capacitors (Figure 14). Ohm's law, Kirchhoff's
circuit laws, and the superposition theorem are valid. Thus, the same numerical tools can be
used to solve the thermal network as for electric network analysis.

17

The system to model needs to be segmented into subparts, each having its proper heat
capacity as well as thermal resistances to the neighbouring subparts. Depending on the
situation, the resistances can be of conducting, convecting, or radiating nature. Furthermore,
some subparts may include a source. Figure 14 shows the process on the example of a
horizontal GIS bus-bar element. A numerical network analysis tool can be used to calculate
the potential at each node. The potential corresponds to the temperature of the subpart. The
resolution of the obtained temperature distribution is therefore determined by the
segmentation process described earlier.
For static models, the capacitances can be omitted. The first method (M1) to
determine the weighting coefficients is based on the rather simplistic idea that all subparts
should have the same weight independent of the underlying number of known temperature
locations. In the test case there are two subparts: The inner conductor and the outer
encapsulation. As there are five locations of known temperature on the encapsulation but only
two on the conductor, the former get a weighting coefficient of 10% while the latter get 25%
(Table 3). The second method (M2) assumes that heat conduction is the main heat transfer
mechanism in the test case (i.e. neglecting convection). Therefore, the weighting coefficients
are chosen according to the covered surface of the corresponding subpart. The surface areas
18

A1 to A7 were assigned to the known temperature values T1 to T7. It should be noted that the
temperature values T2 to T4 apply symmetrically to both sides of the conductor and therefore
appear twice (red, green, and purple surface parts in Figure 15).

The third method (M3) to determine the weighting coefficients is based on the
simulation of a stationary temperature field in the cross section that can be obtained by CFD.
The resulting histogram of the surface area by temperature can be mapped to the maximum
temperature values at all i = 7 positions. This leads to weighting Coefficients as in the M3column of Table 3

The fourth method (M4) evaluates the temperature at one single position only. The
user should select the position which shows best correlation with the pressure. In the chosen
example, this is T1 in Figure 15. It turned out that a linear approximation delivers excellent
results:
19

20

5. LABORATORY RESULTS
5.1 FOR DENSITY MEASUREMENT:
Laboratory experiments are used to verify the findings stated earlier. The test setup
consisted of a horizontal GIS busbar element consisting of components from the ABB ELK-3
(550 kV) series. The gas volume under test is approximately 2.8 m long and 0.5 m in
diameter. It is closed on both sides by barrier insulators (Figure 10). The configuration is
normally in horizontal position. However the complete arrangement can also be tilted by any
angle up to complete upright position (standing on four posts). A computer controlled lowvoltage transformer delivers currents up to 4500 A (on the right in Figure 10). The short
circuit is at the left barrier insulator between the conductor and the encapsulation. This
configuration somewhat differs from nominal operation because of the permanent load
current in the encapsulation, which results in circumferential heating of the gas chamber.
Infrared heaters and a pyrradiometer were utilized to simulate the effect of solar irradiation
such as shown earlier in Figure 2. The heaters were not powered for the studies shown below.
The setup involves 1 pressure sensor (square in Figure 10), 12 density sensors (circles), and
19 thermocouples (triangles).Gas-tight connectors allow their connection to measurement
equipment outside the GIS. The test duration for this first experiment was 24 hours. Before
conducting the test, the load current was zero for a long period of time to ensure a steadystate condition. The ambient temperature was kept constant and there was neither solar
radiation nor wind nor precipitation. During the first 9 hours the load current was kept
constant at 4000 A, while for the remaining 15 hours it was zero (Figure 11 bottom).

21

Figure 11 shows the resulting density sensor readings at the same four locations as in
Figure 9. The dominating density characteristics are of exponential type. During the initial 10
minutes, the density increased at the three peripheral sensor locations (TOP, SIDE, and
BOT), but decreased at the inner location (COND). This can be explained by a pure
diffusive behaviour dominated by heat conduction. With an increasing temperature gradient,
the addictive behaviour gained in influence corresponding to the increased heat transport by
gas convection. After 1 hour, it became obvious that the bottom position was disadvantageous
for density monitoring. The position at the conductor was not recommended either. The
position with the least fluctuation was at the side. The pressure reading in Figure 11 is only
for comparison purposes and shows clearly the advantage of density monitoring. The case of
vertical elements has also been tested. As expected, the characteristics look very similar as for
the horizontal bus-bar except that the maximum deviations are slightly less pronounced.

5.2 FOR PRESSURE MEASUREMENT:


22

The pressure was measured in the horizontal GIS bus-bar during a 24 h load cycle as
in Figure 2 but without solar radiation. The temperature values were obtained by
simultaneous measurement at the same experimental setup as for the pressure. All four
variants of the pressure model and the pressure measurement show similar characteristics.
Some of the weighting methods tend to deliver exaggerated pressure values consistently
(especially M3), while others deliver values that are too low (especially M2). M1 and M4
produce pressure values very close to the measurement. Method M1 shows good agreement
with pressure measurement, which leads to substantial attenuation of fluctuation for the
presented GIS geometry volume and the chosen load profile. Method M4 takes an
exceptional role among the four methods. It only considers the temperature at one location
(T1 in Figure 15) but allows (linear) adaptation of its impact. This leads to the best
attenuation of fluctuation. On the downside, it needs individual tuning parameters (T0 and k
in equation (14)) that require additional effort for determination. This tuning for a certain
configuration might lead to unintentional specialization to a certain load profile with reduced
accuracy for very dissimilar load profiles from the one that was used for parameter tuning.
Refinement of this method by adaptive tuning parameters based on neuronal network might
be considered in future. Two factors are considered to quantify the quality of the pressure
model: the maximum deviation from the measurement and the corresponding filter
attenuation factor A. The motivation for this rather atypical criterion of quality originates in
the application of this model: The model output is assumed to be compared with an internal
threshold for the maximum leakage rate. The maximum deviation is defined as,

Table 4 summarizes the quality of the four different methods of determination


of weighting coefficients

23

AN ALTERNATIVE METHOD:
Besides comparison of measured values with an internal model, there is often another
option readily available. If several gas volumes are exposed to identical conditions, viz.
identical load current and ambient conditions, they can be compared with each other. This
condition can often be assumed being met for the different conductors of a three phase system
with single-phase enclosures. Nevertheless, this does not always work: The author once
struggled with a three phase system where the three SF6 circuit breakers were exposed to
different solar radiation. In that particular case, the density readings were by no means
comparable. It is advisable to check for asymmetries before implementing leakage
monitoring based on such an algorithm.

24

6. CONCLUSION
Gas leakage monitoring in SF6 switchgear is challenging, especially if the regarding
unit has significant fluctuation in load current and/or ambient conditions. Among the
multitude of options to improve the accuracy of SF6 leakage detection, two methods proved
to be robust enough for application in substations: density monitoring with optimal sensor
location utilizing CFD simulations and pressure monitoring with temperature compensation
based on a thermal-network model. They both have their specific field of application:
- If there is a high degree of freedom to choose the optimal sensor location and no time to be
spent for developing a digital model, then density monitoring is recommended. CFD
simulations or measurements are advisable to find the best location of the density sensor.
- If the fluctuation in load current and/or ambient conditions is quite substantial, then using a
pressure sensor is recommended in combination with a model based filter algorithm. This is
especially true for vertical SF6 bushings where the sensor needs to be placed at the lowest
location of the gas volume where the temperature fluctuation is highest.
- In both cases, the comparison between conductors L1, L2, and L3 is another method to
distinguish between thermally based fluctuation and a real leakage. However, this method
implies the same load and ambient conditions for all three conductors, which might not
necessarily be true.
- If the gas volume is not opposed to changing ambient conditions and at the same time there
are no substantial load fluctuations, no additional measures might be needed to further
improve monitoring. Both pressure and density monitoring might work well. Application of
the above-mentioned measures to a GIS setup in the laboratory reduced the level of pressure
fluctuation from 11% to 0.3% and the level of density fluctuation from 4% to 0.8%. Future
long term tests on complete indoor and outdoors substations should analyse the quality and
usability of the proposed methods.

25

REFERENCES
[4] M. Kawada, T. Minagawa, E. Nagao, M. Kamei, C. Nishida and K. Ueda Advanced
Monitoring System for Gas Density of GIS, Intl. Conf. on Condition Monitoring and
Diagnosis, pp. 363-368, Beijing, China, 2008.
[9] L. Niemeyer, A Systematic Search for Insulation Gases and their Environmental
Evaluation, 8th Intl. Sympos. Gaseous Dielectrics, pp. 431- 442, Virginia Beach, USA,
1998.
[10] N. H. Malik and A. H. Qureshi, A Review of Electrical Breakdown in Mixtures of SF6
and Other Gases, IEEE Trans. Electr. Insul., Vol. 14, pp. 1- 13, 1979.
[19] Datasheet SF6-Tracer, G.A.S Gesellschaft fr Analytische Sensorsysteme mbH, 11
June 2008. PDF download http://www.gas-dortmund.de [21] J. De Capitani, Influence of
Thermal Expansion on Gas Leakage Detection in Switchgear, Term paper, ETH Zurich,
2007.
[24] Datasheet Trafag 8773 Density Sensor, file H72505c, version 5.2.1.1, February 2004.
PDF download http://www.trafag.com
[25] X. Dong, R. Summer, U. Kaltenborn, Thermal Network Analysis in MV GIS Design,
CIRED 20th Intl Conf. on Electricity Distribution, Prague, Czech Republic, paper 0637,
2009.

26

You might also like