Beruflich Dokumente
Kultur Dokumente
Abstract
J. Jaspar and C.J.G. Spoorenberg
SGB-Smit Transformers Nijmegen
30 march 2011
Title : Practical aspects of determining the hot-spot (factor) in large power transformers
In the new IEC 60076-2 [1] and the loading guide IEC 60076-7 [2] there is a lot of
emphasis on the hot-spot temperature rise, how to estimate this and how to measure it? This paper
focuses on these practical aspects and as a first step the different cooling modes, including ON and
OD, are illustrated with the temperature rise as function of height, as according IEC 60076-2 figure B1.
As a next step the hot-spot factor H is determined by the local loss factor Q and the local
cooling factor S ( H = Q * S ). These two factors are usually calculated separately by using 2D FE
programs and the fact that the transformer is a 3D construction is neglected. By performing 2D
magnetic field calculations on four, well chosen, cross sections in the circumference of the winding,
one is able to determine an accurate Q factor for the transformer as a 3D configuration. By performing
2D CFD calculations one can calculate the oil flow in the cooling ducts. In the case of axial cooling
ducts, this is very straightforward for ON, but in the case of OD the pressure drop in the circumference
and the hydraulic resistances of the separate windings have to be taken into account. Pressure drop
measurements during heat run tests of large power transformers illustrate that effect.
In several OD cooled GSU transformers hot-spot sensors are placed and the results of the
temperature rise test are presented and compared with calculated values.
It is recommended that the temperature rise model, according Figure B.1 of IEC 60076-2, should
include both LV and HV winding, the return flow through the cooler and those points that are
measured or directly derived from measurements. At the same time two similar figures should be
added regarding the OD and the OF cooling mode.
Based on the test results one can conclude that the hot-spot factor is a good tool for analysis
but is difficult to validate due to tolerance build up due to measurement errors. One has to use the hotspot temperature rise for validation, although this is less sensitive for validation.
J.Jaspar
SGB-Smit Transformers
Nijmegen
The Netherlands
j.jaspar@smit-trafo.nl
2
1.
Introduction
In the new IEC 60076-2 [1] and the loading guide IEC 60076-7 [2] there is a lot of
emphasis on the hot-spot temperature rise, how to estimate the hotspot factor H and how to measure
it. This paper focuses on the practical aspects, and as a starting point one should first consider how the
oil flow is distributed inside the transformer, taking into account two windings and the different types
of circulation mechanism of the oil through the winding.
The hot-spot factor H is a very good principle for analysis, but how to calculate the Q-factor
and S-factor for each individual paper insulated conductor and then how to come to a good estimate is
a point which is discussed in this paper.
To compare calculated values with measurements several ODAF large power transformers had
fibre optic (FO) temperature probes fitted. The hot-spot factor has a large measurement error so
validation has to made by comparing the calculated and measured the hot-spot temperature rises.
2.
The temperature-rise as function of the height is very well presented in Figure 2 [2] and Figure B.1
( [1] annex B ) which includes at the same time the definitions. ([1]- chapter 3).
Figure 1
To explain the temperature-rise in a transformer with two windings and different types of oil
circulation mechanism, this picture needs to be extended in several steps.
As a first step, the difference in the oil distribution between axial and radial ducts needs to be
illustrated ( figure 2 ). In the case of axial cooling ducts one can see that for ducts with identical
hydraulic resistances put in parallel, an almost equal oil flow distribution over the ducts is guaranteed.
In the case of radial cooling ducts there will be a kind of series connection of hydraulic resistances (
axial part of cooling duct ) and parallel connection of hydraulic resistances ( the radial cooling ducts ).
3
This will result in an unequal oil flow distribution in the radial cooling ducts, with very low velocities
in some ducts in the case of ON
Remark : This series and parallel network of hydraulic resistances looks similar as the series
and parallel network of capacitances, which determines the non-linear capacitive
voltage distribution in a winding during lightning impulse test.
Remark : In the case of OD the oil velocities in the axial ducts are in the order of magnitude of
0.1 to 0.2 m/sec.
In the case of radial cooling ducts, the axial oil velocity will be higher for the same oil flow
[m3/ sec] through the winding. ( compare figure 2a with figure 2b )
In the case of radial cooling ducts the straight blue ( oil rise ) line in figure 1 needs to modified
slightly by superimposing some kind of small variation on it. There are many publications referring to
the temperature distribution as function of height for ON [8] with radial cooling ducts. The experience
of Smit Transformers is related to windings with axial cooling ducts, so this paper is limited to this
experience
As a second step Figure B.1 [ 1] will be extended for two windings, because an LV winding
and HV winding will differ in oil flow, average winding gradient (
wo
This can be expressed by a temperature rise line for the LV and HV winding, starting from
the same bottom oil temperature rise (
temperature rise test ( or directly derived from the test data ) are marked with a star ( Figure 2 [2] ).
From figure 3 it is clear that the temperature rise of the top oils that exit each winding are different and
that the top-oil temperature rise, as measured by the sensors on the top of the tank (
), is a kind of
weighted average of these two temperature rises. One of the basic assumptions of a temperature rise
test is that the temperature of the average oil in the cooler is the same as the temperature of the average
oil in the windings, which is usually, but not always, a valid assumption. This difference is also
expressed in figure 3.
As a third step the figure will be extended with an oil flow parallel to the windings, as in the
case of OD and OF. In the case of OD, the oil velocity in the windings will increase, which will result
in a much lower temperature increase of the oil in the height of the windings, which is expressed by a
steeper temperature distribution line ( figure 4 ) . The oil velocities are now determined by the
hydraulic resistances of the cooling ducts in the separate windings and can be quite different from each
other. The oil that exits the windings is mixed with the oil that bypasses the winding (a thin vertical
arrow figure 4 ), which is relatively small compared to the flow through the windings. This effect is
expressed in the temperature rise of that mixture of oils. This mixture is measured by the top-oil sensor
and might be even somewhat lower than both oil temperatures rises that exit the windings.
Due to the higher oil velocity, the heat transfer from the winding to the oil improves, so the winding
gradient (
wo
) decreases.
Combining figure 3 regarding the thermosiphon flow in the winding with the corresponding
winding gradients and figure 4 with respect to a very large oil flow bypassing the winding (a thick
vertical arrow ), will result in figure 5, which represents the OF cooling.
In the case of OF, there is no representative measurement of the top oil temperature and so the
basis to determine the hot-spot temperature rise is point of debate. Smit offers the OD solution as
alternative for OF, because of that reason.
6
3
The Q and S-factor determine the hot-spot-factor H ( H = Q*S ) [1] and are defined as:
Q factor : A dimensionless factor to estimate the increase of the average winding gradient due to the
local increase of the additional loss
S factor : A dimensionless factor to estimate the local increase of the average winding gradient due to
the variation in the liquid flow stream.
Based on these definitions the Q-factor and S-factor are not just a single number, but should
be considered as a function of location. The hot-spot factor should then also be considered as a
function, but in the discussions one usually refers to the hot-spot factor as the maximum value of this
function in the winding and the hot-spot is the location of that maximum, since the hot-spot is the area
of most interest for the designer.
Cz
Cr
r
Figure 6
The losses ( presented by the numerical value P ) and the cooling ( presented by
the vector components Cr, Cz and C ) for one paper insulated conductor
Heat transfer ( = cooling ) is always in three directions and can be considered as a vector and the
additional losses are located in one conductor and can be considered as a scalar . That makes the
calculation of the S factor more complicated than the calculation of Q-factor. In the case of axial
cooling ducts the heat transfer will take place as follows ( see figure 6 ):
In the axial direction due to the contact between the paper insulated conductors.
In the circumferential direction, because the copper conductor will redistribute the temperature.
In the radial direction the heat transfer is through the paper that is in contact with the oil.
As a first step, these factors will be expressed in a mathematical way by using cylindrical
coordinates:
= P( r,z, ,T ) / Pave
Where:
P( r,z, ,T )
: Radial position
: Angle in circumferential position
: Axial position
Pave
: The average loss of the winding at average winding temperature [W/m3 ] or [W/m]
For calculation purposes one can redefine that for one single paper
The additional losses should be calculated for a small part of the winding, and the smallest
element in a winding is a single paper insulated conductor. For that reason one should calculate the Qfactor for each individual conductor. If one would calculate the Q-factor for a single disk or even
several disks together, the Q-factor will be lower because one averages the additional losses. If one
would calculate the Q-factor for one single layer in the case of axial cooling ducts, the Q-factor would
always be about 1.0.
C(r,z, ,T)
[ K/(W/m2) ]
Cave
[ K/(W/m2) ]
The maximum value of Q * S in the winding is usually named the hot-spot factor of a
winding, and that can can be mathematically expressed as:
,T ) * C( r, z,
, T) ) / ( P-ave * C-ave ) ]
From the previous points, it will be clear that one has to make assumptions to calculate the
hot-spot factor in the winding of a transformer. One therefore can better use the word estimate, but that
certainly does not imply inaccurate.
straight forward mathematical procedure, especially when one neglects the magnetic field due to eddy
currents in all the conducting parts. These conducting parts consist of the copper of the windings, the
steel press construction of the windings, the core and the tank wall including the magnetic shunts. In
the past already a comparison was made between the different calculation methods [ 10 ] but that is
outside the scope of this paper. A difference due to the calculation method and another approach is not
taken into account.
4.1
but the surrounding steel parts like core, pressing structure and tank are not. ( see fig 7 and 8 )
Figure 7 3-D picture of a three phase transformer with 5-legged core and LV winding with large pitch
Figure 8
9
In the case the inner low voltage winding has a big pitch, the winding itself is, not rotationally
symmetric. One is able to determine the magnetic field in the winding more accurately by making a
number of 2-D calculations. If one looks at the top view of this active part ( see figure 8 ) , one can
limit the calculation to 2D- cross sections as A, B , C and D ( see figure 9 ).
cross section A
cross section B
cross section D
cross section C
Average
Figure 9
Four 2-D cross sections of the active part and the magnetic average.
10
One is allowed to make these 2-D calculations, because the dimensions that determine the
radial magnetic flux density at the top of the winding, are much smaller relative to other dimensions,
like winding height, average diameter and core diameter [7] .
4.2
Calculation example
This previous calculation procedure can be illustrated by using the data of a transformer that is
already described in literature [8]. The windings are rotation symmetric and the core is three legged, so
the magnetic field calculation can be limited to two 2D cross sections, inside the core window and
outside the core window.
The core diameter
587 mm
: 2700 mm.
: 1499 mm
100 mm
The magnetic flux width, defined as the width of the main barrier
and half of the winding thicknesses , is proportional to the total
magnetic flux of the winding set
170 mm
A 2-D magnetic field calculation in the core window ( like cross section B ) is a good
approximation because the part of the circumference inside the window ( 587 mm ) is much larger
than the end distance ( 100 mm ) and the magnetic flux width ( 170 mm ).
About one quarter of the circumference ( 2700/4 = 675 mm ) faces the tank.The radial
clearance of the HV winding to the tank is 295 mm, which is far less than 675 mm. So a 2-D magnetic
field calculation outside the core window (like cross section C or D ) is a good approximation, because
the part of the circumference facing the tank is ( 675 mm ) is much larger than radial clearance ( 295
mm ) and the magnetic width ( 170 mm )
P[ W/m ]
Average value of winding
P [W/disk ]
18,5
734
Pmax [W/m ]
Pmax [W/disk]
40,5
2,2
1506
2,1
32,9
1,8
1230
1,7
HV 20 mm higher *)
46,4
2,5
1713
2,3
HV 20 mm lower *)
39,8
2,2
1476
2,0
Table 1 Losses Pmax per conductor and per disk for 66 MVA and 75 oC
*)
Calculations are made outside the window, but with the height of HV winding either 20 mm
11
**)
The hot-spot factor is calculated in one LV winding and the effect of the other phase inside the
Remark : The Q-factor of this LV winding is relatively high, so the variation ( table 1 ) for different
cross sections and for height tolerances is more pronounced. In the case the axial dimension
of the conductor would be half, the Q- factor would then have a variation between 1.2 and 1.4.
Conclusion : A Q-factor of 2.0 in for this LV winding of this transformer is a proper estimate.
This is based on the volume of one conductor, the calculation method and the boundary
conditions.
4.3
figure presents the Q factor as .average of a number of cases for which the calculation was
made by means of computer programs and confirmed by experimental results. The graph is limited
with respect to MVA, short-circuit impedance and 50 Hz. There are many other considerations that
should be taken into account and it requires the use of magnetic field calculations to calculate the
maximum value of the Q factor in each case.
12
5.
Calculation of the S factor in a core type transformer with axial cooling ducts.
The temperature rise over ambient of a paper insulated conductor inside a winding consists of
The losses in the copper have to be conducted over the paper insulation, which cause a temperature
drop over the paper.
The losses have to be cooled by oil convection, which causes a temperature drop over the oil film
that is between the paper and the main oil in the duct.
The temperature rise of the oil over ambient inside the duct that is in contact with this conductor.
This oil temperature is a function of the height, but might also depend on the location in the
circumference and the location of the cooling duct in radial direction.
Depending on the type of winding there are transpositions and they are handmade and require
skilled workmanship. In every disk winding one has to make a S-bend to change from one disk to the
next disk. In a layer winding one has to change from one layer to the next one. In the case of parallel
paper insulated conductors, one has to transpose the conductors on defined locations to assure an
almost constant current distribution between the parallel conductors. Due to these transpositions the
winding should not be considered rotation symmetric regarding the cooling. The heat transfer in the
circumferential direction close to a transposition cannot be neglected. These transpositions are
sometimes protected with extra paper or Nomex which increases the thermal resistance, but in another
way they are sometimes better surrounded by oil which improves the heat transfer to the oil.
5.1
The conductors in a layer are in good contact with each other, also due to the axial pressure
force on the winding. By conduction there will be a heat transfer between the conductors. This heat
transfer distributes the temperature within the layer. At the same time there is also heat transfer from to
the copper to the oil through the paper.( see figure 6 )
The conductor is made of copper and there will also be heat transfer in the circumferential
direction. This is relevant in the case of a local disturbances in the rotation symmetry , as in the case of
transpositions.
An analytical model of a conductor with different insulation thickness over the length of the
conductor is presented in [12]. The bell shaped temperature profile, as presented in figure 7, is due to
the heat transfer through the paper in the radial direction and the heat transfer through the copper in the
length direction of the conductor.
5.2
The heat transfer between a paper insulated conductor and the oil can be improved by
increasing the cooling surface. At the end of the winding there are higher additional losses due to the
higher radial stray field strength. By adding a radial cooling duct near the end of the winding, one is
13
able to create some extra cooling surface. In that way one is able to reduce the S factor at the location
were the Q factor is higher. The hot-spot factor in this area will then reduce.
A different option would be the increase of the heat transfer coefficient [ W/m2/K ], but the
heat transfer coefficient depends on the oil velocity ( and hence whether the oil flow is laminar or
turbulent), the viscosity of the oil and distance from the entrance of the cooling duct [6]. The heat
transfer coefficient in axial ducts is difficult to influence in the case of ON, but can be increased in the
case of OD.
5.3
the oil flow in the axial cooling ducts more or less uniformally. The winding is rotation symmetric
from that point of view and the S factor for each winding can be estimated as being almost equal to 1.0.
In the case of OD, the oil is directed into the windings at both sides of the core, so the oil
has to flow in the circumferential direction which will result in a pressure drop in that direction. So the
axial oil flow will be a function of the circumference , and as a consequence so will the temperature
drop over the winding and the top-oil that exits the winding. Because the temperature drop over the
winding height is relatively small in the case of OD, the consequences on the temperature distribution
will be limited.
In the case of OD, the oil volume flow distribution between the windings is determined by
hydraulic resistance of the windings. Deviations between the hydraulic resistances of the separate
windings will create in an unequal oil flow distribution. These resistances also include the electrical
insulation construction at the top and bottom of the winding, because they create a zig-zag pattern to
prevent long creepage paths but they also create a hydraulic resistance. These insulation constructions
will be different for the LV and the HV winding ( and generally higher for the HV ) and so will have an
influence on this oil flow distribution.
5.3.1
A long axial cooling duct with an almost rectangular cross section, has a well defined
hydraulic resistance. By measuring the pressure over the duct height, one can calculate the oil flow and
so the oil velocity in the duct. To measure the pressure drop in the circumference, one makes a small
hole in the pressure ring to connect a small teflon tube that can be inserted into the oil duct of the
pressure ring. This tube is connected to a differential pressure meter( figure 11) at the outside of the
tank. The other side of this pressure meter is in contact with the pressure of the bottom oil. The meter
is calibrated to zero to compensate for the difference in static pressure due to the different height
locations inside the tank.
14
Before start of the test the oil temperature is measured, because the oil viscosity is strongly dependent
on the temperature. In steps the pumps are switched on, so the pressure and the oil volume flow are
measured at each step. The result is a graph with pressure as function of total oil flow, with the
temperature as a parameter.
As a first case the measurement was performed on a 375 MVA single phase generator step up
transformer, with a LV winding with radial cooling ducts and a HV winding with axial cooling ducts.
The measurements were made before and shortly after the temperature rise test, so results are available
for two different temperatures ( figure 12 ) . The decrease of the viscosity of about a factor 4 to 5
results in a higher oil flow and in lower pressure drop.
Pressuremeasurement
Pressuremeasurement
60
25.00
50
20.00
Pressure [ mbar ]
pressure [mbar ]
40
30
20
15.00
10.00
5.00
10
0.00
100
200
300
400
Volumeflow[ m^3/hour ]
500
600
50
100
150
200
250
Measurement 48 oC
Calculated 48 oC
Measurement 13 oC
Calculated 13 oC
90 degrees
45 degrees
0 degrees
Calculated
As a second case the measurement was performed on a 224 MVA transformer with only axial
cooling ducts. Based on the test result one can estimate that the pressure at the oil inlet ( reference
position 0 degrees ) is about 30 % higher than average and in the middle of the core window ( 90
degrees ) about 30% lower than average. ( Figure 13 ) In the case of an average temperature drop
of 10 K over the height of the windings, the effect on oil exit temperature of the winding will be limited
to -3K at the inlet of the oil and +3K in the middle of the core window.
15
The test results are compared with the hydraulic calculations based on rotation symmetry and
there is a reasonable agreement. Based on these test results, one may conclude that the duct in the
circumference should have an adequate cross section to limit this pressure drop. In this way the oil
velocities in all (axial ) cooling ducts in the circumference of the windings do not deviate too much
from the average oil velocity .
5.3.2
manufacturing stage a quality check is done, by circulating air through the windings. There are two
aspects that needs to be checked:
Is the division of volume flow between the LV- and HV winding in accordance with the
calculation?
Is the flow distribution between all the cooling ducts inside a winding both in the circumferential
direction and radial direction fairly constant, especially close to the transpositions?
Figure 14
Manufacturing tolerances have an effect on the flow distribution, but due to the very strong
dependence of the viscosity of the oil to the oil temperature, the effect on the oil temperature gradient
over the height is limited.
Remark :
The kinematic viscosity is temperature dependent and depends also on the type of oil. It can be
expressed by the following typical mathematical function:
= 65 10-9 *e (676/(T + 96) [ m2/s ]
with T : Oiltemperature in oC
In the relevant temperature range, this results in about a factor 1.7 to 1.4 decrease in viscosity for
every 10oC rise of the oil temperature.
16
6.
In principle, there are two methods to fix a FO sensor thermally to the conductor. In one case
the FO sensor is fixed under the conductor paper in direct contact with the (enameled) copper
( figure 16 ] ) or in an axial spacer ( figure 17 ) as according IEC [ 1 ]. As a first step the difference
between these two methods is investigated.
A small FE model ( Figure 15a ) is made in 2D, consisting of two copper conductors on top
and bottom. The copper conductors have a paper thickness of 0.6 + 2.0 = 2.6 mm. The 2.0 mm paper
simulates the heat transfer from the paper to the oil and the oil temperature in the oil duct is set to zero.
The boundary condition on top and bottom are set to a heat flux of 1000 W/m2 and the left and right
boundary are set to be symmetric. The detail of temperature profile ( figure 15b ) shows that there is
no difference in temperature between the centre of the spacer and the copper. It can thus be concluded
that both methods to fix the sensor represent the temperature of the copper.
Figure 15a
The outermost paper of the conductor which is in contact with the spacer has the same temperature as
the innermost paper that is in contact with the copper. This complies with the fact that, in old
transformers, the paper under the spacer is as much aged as the paper in contact with the copper
17
7.
The hot-spot factor needs to be validated by measurements during temperature rise tests, but
looking at the definition from a measurement point of view
H=
one already may expect a very large tolerance due to the individual measurement tolerances of
the 4 separate temperature measurements. ( [9] + [11]). One can illustrate this with a small numerical
example:
: (
80 +/- 1
: (
55 +/- 2
: (
50 +/- 1
: (
om
30 +/- 2
The HV winding of the middle phase had 8 FO sensors assembled in about one quarter of
the circumference ( figure 17 ). In the case that several sensors are assembled in the winding at about
the same height, one is able to reduce the measurement error of the individual FO sensors of +/- 1 K
and the measurement error of the individual channels of the test equipment. Based on the assumption
that there are no temperature differences at that height in the winding one determines the average value
of the measured temperatures. The offset between each sensor and this average value is used to correct
the values during the temperature rise test.
The temperature rise test was preformed with nominal current and started at t = 2.9 hours
(figure 18). To limit the time to reach a stable condition, the test started with some radiators closed and
18
the pumps running. In several steps the radiator valves were opened. At t =3.75 hours all coolers are
open and the fans are switched on and the operating condition is then ODAF. The temperature rises
have to reach a steady state to determine the hot-spot gradients.
30
25
2 - Top oil
20
7 - Top oil
5 - Top oil
15
3 - Top oil
10
1 - Top oil
8 - Top oil
5
4 - Top oil
6 - Top oil
0
2.90
3.90
4.90
-5
5.90
6.90
7.90
Figure 18
Overview of sensors top oil during heat run test with nominal current
Sensor 1,3,5 and 7 are assembled under the paper and sensor 2,4,6 and 8 are in a
spacer. The distances between the corresponding sensors (1 2 ; 3 4 ; 5 - 6 and 7 8 ) are less than
100 mm. The temperature differences between the corresponding sensors is limited and if one adds all
these differences together it is almost zero.
A dynamic difference can be determined, because the sensors under the paper respond
somewhat faster in the case of a steep ( switch off ) load change but this difference is only interesting
from a theoretical point of view. ( see figure 19 ------- lines compared to ________ lines ). The
differences are only visible in the time scale of minutes instead of hours.
19
30
2 - Top oil
25
7 - Top oil
5 - Top oil
6 - Top oil
1 - Top oil
20
8 - Top oil
4 - Top oil
3 - Top oil
15
10
8.34
8.36
8.38
8.40
8.42
8.44
8.46
8.48
8.50
Figure 19
wo
: Gradient winding-oil
(Q)
(S)
: Temperature change due to S-factor caused by different axial oil velocity in circumference
(H)
Calculated values
0o
wo
(Q)
(S)
(H)
g+(H)
19
15.6
+5
-1.5
+3.5
22.5
30
18.3
+3.5
-0.5
+3
22
60
+1.5
+0.5
+2
21
90
+1
+1.5
+2.5
21.5
Table 2 Hot-spot rise over top oil in circumference in OD cooled Smit disk winding
*)
**)
Gradient based on cool down curve including decrease of average oil of cooler
***)
Average value of two sensor measurements on that location. First value between brackets
is under paper and second value is in spacer
The corresponding calculated and measured values in the circumference between all the
sensors are limited, and on average it is zero. The differences between the corresponding sensors can
also be due to the assembly tolerance of sensor and lead ( see also figure 17 ).
20
A numerical calculation of the hot spot factor can also be made, but the value of H is
between 1.1 ( 19.7/ 18.3 ) and 1.6 ( 24.6 / 15.6 ). This large range of H creates no added value for
validation of the design.
7.2
The heat exchanger type of coolers of this transformer have a very large hydraulic resistance
and in the case of loss of auxiliary power the pumps are switched off. The oil velocity in the cooling
ducts almost comes to a stop and the windings heat up almost adiabatic.
The specification of this transformer required a time limit of operating at nominal current in
case of loss of cooling pumps. To check this requirement in the specification, the pumps are switched
off during the temperature rise test and the increase of the hot spot temperature is monitored. The FO
sensors take part in the quality check off the specification.
LV hotspot
100
Hot spot HV nr 1
Temperature [C]
110
90
Hot spot HV nr 2
80
Top-oil exit LV
70
Topoil exit HV
60
Pressure plate
50
Topoil at cover
40
17.500
17.583
17.667
17.750
17.833
17.917
18.000
18.083
18.167
18.250
21
The top-oil exit temperatures of the HV - and LV winding also increase but it is less steep (
about 2 K/minute ) but also not such a smooth line. The top-oil temperature at the cover and the
pressure plate increase slow ( about 1 K/minute ), because they are thermally not so directly linked to
the windings. At t = 18.02 the load is switched off and the temperatures decrease slow. At t = 18.1 the
pumps are switched on and all temperatures decrease fast.
These temperature variations can also be used to check the quality of the oil flow and the
dynamic thermal performance of the windings.
7.3
Temperature rise test on 200 MVA GSU for tolerance on sensor assembly and oil velocity
The additional losses due to the radial magnetic field are very low in the HV winding, so the Q
factor can be assumed to be 1.0 for the whole winding. By installing several sensors in the same way,
one can estimate the influence of the assembly on the temperature and in this way estimate an assembly
tolerance. Before start of the temperature rise test the offset is determined, which is later used to
correct the sensor readings ( see also paragraph 7.1 ).
HV side
Figure 21 Schematic view with cooler arrangement. Winding 1 is on the left, close to cooler 1 and 2.
Due to the site requirements, the coolers are not symmetrical oriented to the transformer, so by
switching on pumps, the total oil flow increases but the oil division between the phases changes
somewhat. ( see figure 21 )
Time
Condition
At start
At 5.1 hour
100% losses. Fan 2 and 3 ON. Pump 1,2 and 3 ON. So higher oil velocity
At 6.1 hour
125% losses. Fan 2 and 3 ON. Pump 1,2 and 3 ON. So higher oil velocity
At 7.0 hour
125% losses. Fan 2 and 3 ON. Pump 2 and 3 ON. So nominal oil velocity
At 7.5 hour
125% losses. Fan 3 ON. Only pump 3 ON. So low oil velocity and less cooling
At 7.7 hour
125% losses. Fan 2 and 3 ON. Pump 1,2 and 3 ON. So higher oil velocity.
22
30.0
25.0
20.0
Avg(1HV)-Topoil
Avg(3HV)-Topoil
15.0
Avg(HV123)-Topoil
Avg(2HV)-Topoil
topoil1 in HV2
topoil2 in HV2
10.0
5.0
0.0
4.5
5.5
6.5
and
7.5
The assembly method limited the variation of the measured temperatures of one winding
within +/- 1 K, so one may conclude that the assembly method has a temperature tolerance of +/- 1 K.
Due to this limited tolerance, one can average the temperatures for the separate phases, which
is presented in figure 22. The difference between phases is small, but cannot be neglected. It is likely
that this is due to the unequal oil flow between phases.
In the case of phase Nr 1, the temperature increase if one changes from 2 to 3 pumps ( at t =
The oil temperatures in the HV winding are positioned between the electrostatic shield rings
and should, in essence, give the same result. The test results show a difference and this complies with
previous results ( [4] + [9] ).
The average gradient between the hot spot in the HV2 winding and the top oil of the HV2
winding, is about the same as the average winding gradient. This complies with the fact that the hot
spot factor of the winding is about 1.0.
Average gradient
100% losses
----
Nominal current
9.1
125% losses
11 ( 3 pumps)
11.2 or 12 K ( 3 or 2 pumps )
23
8.
Conclusions
The temperature rise model as according Figure B.1 of IEC 60076-2 is a simplification of the
complex temperature behavior in a transformer. It should at least be extended with a second winding
and include the return flow through the cooler. Points that are measured or directly derived from
measurements, should be indicated. Also two similar figures should be added regarding OD and OF
cooling mode.
To calculate the hot-spot factor, it should be considered that a transformer is not a rotationally
symmetric construction, due to presence of core, clamping structure and tank.
The calculation of the Q- factor in a core type transformer requires only some 2D calculations on
properly selected cross sections, from which a sufficiently accurate value of Q can be derived.
The hot-spot factor H, as a product of Q and S, is an excellent tool for analysis, but is very difficult
to validate due to the tolerance build up. The hot spot temperature rise over ambient in steady state
should be used to validate the design of the transformer, in spite of the fact that this parameter is not so
sensitive.
The assembly method of the sensor, under the paper or in the spacer, has no influence on the
measured hot-spot temperature. The assembly tolerance of sensor in the axial spacer is +/- 1 K.
The hot-spot sensors can also be used for testing the dynamic temperature requirements by
performing step changes like switching on and off pumps and switching on and off relatively large
loads.
24
References
[1]
[2]
IEC 60076 7, 2005-12 Power Transformers Part 7 : Loading guide for oil-immersed
power transformers
[3]
IEEE C57.12.90 2006 Test Code for Liquid Immersed Distribution, Power and Regulating
Transformers
[4]
[5]
[6]
[7]
[8]
[9]
Statistical analysis of temperature rise test for increased accuracy on x- and y- exponents
C.J.G. Spoorenberg
[10]
To be published
[11]
[12]