Вы находитесь на странице: 1из 10

International Colloquium Transformer Research and Asset Management

Cavtat, Croatia, November 12 – 14, 2009

1 1 1 1
H. CAMPELO , C. M. FONTE , R.G. SOUSA* , J. C. B. LOPES
1
Fluidinova, Engenharia de Fluidos, S.A., Tecmaia
Rua Engenheiro Frederico Ulrich, 2650, 4470-605 Moreira da Maia, Portugal.
*renato.sousa@fluidinova.pt
2** 2 2
R. LOPES , J. RAMOS , D. COUTO
2
EFACEC – Energia, Máquinas e Equipamentos Eléctricos, S.A.,
P.O. Box 1018 – 4466-952 S. Mamede de Infesta, Portugal.
** ricardol@efacec.pt
3
M. M. DIAS***
3
LSRE/FEUP – Laboratory of Separation and Reaction Processes,
Faculdade de Engenharia da Universidade do Porto, Rua Dr. roberto Frias, 4200-465 Porto, Portugal.
***dias@fe.up.pt

DETAILED CFD ANALYSIS OF ODAF POWER TRANSFORMER

SUMMARY

In spite of their high efficiency, Power Transformers must dissipate non disregardable amounts
of energy due to the losses, which are mainly generated in the windings.. The heat generated in the coils
is removed by circulating a fluid, typically mineral oil, between them. The geometrical configuration of the
oil channels determines the oil flow distribution along the transformer, which affects the heat removal
efficiency and the position and magnitude of the hot-spots. The control of the temperature of the copper
isolation is essential to determine the life-time of the power transformers.
In the present case, directed oil forced air (ODAF) CORE type power transformer was simulated, where
windings of different configurations are present with multiple oil channels in each. Due to the different
possible paths for the oil, the variation of temperature and oil physical properties along the transformer,
Computational Fluid Dynamics (CFD) is the most suitable tool to analyze the flow inside a power
transformer, since experimental measurements are complex and intrusive and can only measure a limited
number of discrete points.

Key words: Cooling, power transformer, CFD, winding, ODAF

1. INTRODUCTION

In previous works [1,2] of Fluidinova and EFACEC, Computational Fluid Dynamics (CFD) was
used to analyse the oil flow and heat transfer in a SHELL type power transformer. Initially, experimental
measurements were performed with PIV (Particle Image Velocimetry) in order to compare the simulation
results with experimental data and validate CFD as a tool to analyze the performance of power
transformers. Once validated, CFD was used to optimize the geometrical configuration of the oil paths in
order to improve the heat removal and minimize hot-spots.
Another common type of power transformers is the CORE type, in which disc windings are typically
distributed circularly around the core as sketched in Figure 1. The windings are placed in horizontal
layers, separated by radial spacers that form the horizontal oil channels. The radial spacers are
supported in vertical guides that create the vertical oil channels. In Figure 1, the four windings that

1
constitute the power transformer are, from internal to external radius, stabilization (T), low voltage (BT),
high voltage (AT), regulation (Reg).
The windings are separated by vertical cylinders of insulating pressboard. From time to time, in the main
windings (AT and BT), there are some washers that obstruct the vertical channels either in the outer or in
the inner side, imposing a zig-zag oil flow along the radial channels.
Oil enters in the lower region of the windings and exits in the upper part.

Oil
outlets

Oil inlet

TOP VIEW 2D projection of vertical slice

Figure 1 - Sketch of the CORE type power transformer geometry used in this study.

CORE type power transformers models have already been the subject of several studies, with two main
approaches: network models and CFD models, as shown by the work of Zhang [3] and Hosseini [4]
respectively, among many other examples.
While network models allow for shorter simulation times, they are supported by several empirical
correlations that might not be valid for new unstudied power transformer configurations where dimensions
or local conditions are not in the validity range of the correlations. CFD studies, while more time
consuming, have the advantage of having more detailed information and are based on more general
physical equations that should be valid for a wider range of conditions.
Every model must however make some simplifications neglecting phenomena that should not be
dominant in order to shorten computational effort. CFD studies usually require a lot of computational effort
and the most common simplifications are considering a 2D axi-symmetric domain, assuming constant

2
heat flux on windings surface without taking into account solid thermal conduction or simulate solid
thermal conduction but with estimated thermal conductivities.
This paper describes the work made by Fluidinova and EFACEC on the analysis of CORE type power
transformers where 3D simulations were performed and both the fluid flow and the heat transfer (oil
convection and copper conduction) were taken into account. For the anisotropic thermal conductivity in
the windings, a detailed analysis was made on a portion of the power transformer, in order to obtain a
simplified model that could obtain similar results and could be used in the complete model without the
necessity of simulating all the insulating paper in detail. The four different windings that were present in
the studied power transformer were analyzed, although the attention was focused on the two central
windings which have higher power dissipations and consequently higher temperatures.

2. CFD MODELS

In order to simulate the oil flow along the power transformer with CFD, the geometrical model
should be as detailed as possible, the physical properties of the oil and solid parts must be correctly
specified and the appropriate boundary conditions should be set in order to obtain reliable results. In this
section the inputs of the model section are described before going into the description of the obtained
results.

2.1 Computational grid

The creation of the computational grid where the conservation equations will be solved is the first
part of the CFD simulation and influences the precision of the results if care is not taken. For the creation
of the computational grid, Gambit 2.4.6 software from Ansys Inc. was used. The simulations were then
carried out in the Fluent 6.3 software from the same company.
Although the power transformer has a cylindrical shape and is practically axi-symmetric, there are
however some details that are not taken into account if 2D axi-symmetric simulation is carried on. This
has mainly to do with the solid spacers that are positioned either radial or axially to separate the windings
creating the oil channels. These spacers reduce the heat transfer and oil flow area which would not be
taken into account in a 2D axi-symmetric simulation. In order to take these details into account a 3D
simulation model was created, representing the smallest repetitive section of the transformer (volume
limited by two vertical planes, one in the centre of a radial spacer and another in the middle of two radial
spacers).
The solid and liquid volumes were then divided in small finite elements where the equations of the mass,
energy and momentum conservation are solved.
A grid analysis was made in order to check the grid independence of the results. The final grid used for
6
the analysis had an average element size of 0.7 mm, Figure 2, with an overall number of about 3x10
elements.

Figure 2 - Detail of the computational grid created to simulate the oil flow and heat transfer in the power
transformer.

3
2.2 Oil Physical properties

The oil viscosity was measured in a cone-plate rheometer in the temperature range between
20ºC and 80ºC and the data was fitted to the following viscosity model:

B
K exp (1)
T
-1 -1
Where µ is the oil viscosity (kg.m .s ), T is the oil temperature (K) and K and B are fitting parameters
-7 -1 -1 3
with K = 1.433x10 kg.m .s and B =3.480x10 K.

Also the density was considered to vary with temperature by using the Boussinesq model:

= 0 (1- (T-T0)) (2)


-3 -3 -1
Where is the oil density (kg.m ) at temperature T(K), with 0 = 868 kg.m , = 0.00064 K and T0 =
293 K.

Thermal conductivity and heat capacity of oil were also taken into account, but with constant values since
-1 -1 -1 -1
its variation with temperature is negligible in the studied range, Koil = 0.126 W.m .K , Cp = 2016 J.kg .K .

2.3 Windings thermal conduction

The windings are composed by groups of copper wires, surrounded with insulating paper
(individually and grouped). The detailed description of the different materials in the full transformer
simulation would imply a very large number of finite elements in order to describe the thermal properties
changes along the windings and due to the insulation small thickness. According to the scheme of Figure
3, it is perceptible that the windings thermal conductivity is not isotropic.

Copper Insulation paper

Figure 3 - Scheme of windings anisotropic thermal conductivity

While in tangential direction the winding thermal conductivity is equal to the copper thermal conductivity,
in the axial and radial direction it depends on the relative amounts of copper and insulation material, with
thermal resistances in parallel and in series.
A detailed thermal analysis was performed with CFD in a disc of a winding, which was then simplified in
order to obtain an equivalent material with anisotropic thermal conductivities that would represent the
behavior of the copper and insulation paper portions that constitute a winding.
After some tests, it was found that the best way would be to consider the external paper insulation as an
individual resistance with paper thermal properties while the copper and the internal insulation paper were
substituted by a material with a thermal resistance equivalent to the parallel and series resistances of
copper and paper portions.
This was tested for the two main windings of the studied power transformer, which have several copper
wires in parallel with different configurations. The comparison between the detailed and the simplified
models for the windings in terms of thermal behavior are presented in Figures 4 and 5.

4
Temperature (ºC)

Figure 4 - Comparison of the thermal behavior of the high voltage winding with a detailed
geometrical description (up) and with a simplified model (down).

These tests show that the simplified model can predict the temperature variation in the windings. The
anisotropic thermal conductivities used in the CFD model are those presented in Table 1.

Table I - Anisotropic conductivities used in the low and high voltage windings.
-1 -1
Thermal conductivity (W.m .K ) Low voltage High voltage
Tangential 388.5 388.5
Axial 3.9 286.9
Radial 0.9 0.6

Temperature (ºC)

Figure 5 - Comparison of the thermal behavior of the low voltage winding with a detailed geometrical
description (up) and with a simplified model (down).

2.4 Boundary conditions

In order to have reliable results coming from the CFD simulations, apart from the materials
physical properties, the boundary conditions imposed in the limits of the studied domain must be as close
as possible to reality.
In the domain inlets and outlets, pressure conditions were considered. Since the absolute pressure is not
important for the current case, a null relative static pressure was considered in the outlet and the inlet

5
static pressure was set in order to give the experimental oil flow rate. The inlet oil temperature was also
specified according to the measured value.
The external pressboard walls (outer radius, top and bottom) are considered adiabatic. In the windings,
power dissipation per volume is specified as a function of the height. The boundary conditions used in the
simulation are specified in Table 2. In the vertical faces limiting the repetitive section in the centre of a
radial spacer and in the middle of two spacers, a symmetry boundary condition was used.

Table II - Boundary conditions used in the CFD simulations.

Pstatic = 310 Pa (to adjust for the experimental oil flow rate of
3
Inlets 37.4 m /h)
Toil_inlet = 80ºC
Solid surfaces
No slip condition
(momentum)
3
AT 343 708 W/m (48 4446 W total)
3
Windings power BT 312 509 W/m (31 341 W total)
dissipation T 3
4 333 W/m (90 W total)
3
Reg 413 100 W/m (7 931 W total)
Outlets Pstatic = 0 Pa

3. RESULTS

With the models and boundary conditions described above, CFD simulations were run where in
each element of the domain the temperature, heat fluxes, pressure and oil velocity were obtained.
In Figure 6 is shown the overall temperature contours in the full studied geometry that consists of a
repetitive section of the full power transformer, and also a zoom of the upper region.

Figure 6 - Temperature contours in the full studied geometry.

Figure 6 shows that as expected, the higher temperatures are in the windings, while the oil slowly
increases its temperature along the way up, being possible to observe a higher temperature in the oil
outlets in the top of the geometry.

6
In the windings, it is also interesting to see that the average temperature of each section between two
washers increases along the height of the transformer, due to the higher oil temperature. Despite of that,
in each group of discs, it is not the highest disc that has the higher temperature. This is explained by a
higher oil velocity in the upper horizontal channels as will be seen in a following section. Another fact that
is observed in Figure 6 is that in each section, the higher temperatures are not in the centre of the discs
but are displaced to the inner or outer radius according to the oil flow direction. This is an important fact
when making experimental measurements since in the same disc the temperature may vary by more than
10ºC in the radial direction.
The average and hot-spot winding temperatures are presented in Table 3 and compared to the results
obtained in a 2D axi-symmetric simulation and experimental values.

Table III - Comparison of average and hot-spot temperatures in AT and BT windings for a 2D axi-
symmetric and a 3D simulation and estimated experimental values.
2D temperatures (ºC) 3D temperatures Experimental
(ºC) estimation (ºC)
AT (average temperature) 99.6 102.9 106.7
BT (average temperature) 98.6 106.6 117.9
AT (hot-spot) 112.3 115.5 -
BT (hot-spot) 108.7 118.4 -

As Table 3 shows, the 2D axi-symmetric simulations underestimate the winding temperatures, since it
neglects the effect of the oil-winding heat transfer surface reduction due to the presence of the radial and
vertical spacers. This effect is more significant in the BT winding, since due to the shorter winding radius,
the percentage of area reduction is higher. It is also observed that the relative temperature difference is
changed from 2D to 3D simulation, and from the experimental estimated values, according to the IEC
60076-2 standard, it is seen that the 3D simulation temperatures are closer to those expected. The
differences between the 3D simulation and the experimental temperatures might be caused by:

1. uncertainty of the oil flow rate: it was estimated by the overall power transformer oil flow rate (for
3 phases) assuming a certain percentage of bypass flow rate and that the remaining would be
equally distributed for the three phases.
2. eventual geometry differences: CFD simulation assumes perfect construction of the power
transformer according to the project design, while during real construction there are small
misalignments when assembling and when the power transformer is vertically compressed.
3. uncertainty of the experimentally estimated temperature, that results from extrapolation of a
cooling ramp, for t=0s.

Concerning the oil velocity, which apart from the oil temperature is one of the variables that has impact on
the surface heat flux and consequently on the winding temperature, Figure 7 shows the velocity
magnitude in the upper and lower regions of the power transformer. It is possible to see that the higher
velocities in the main windings are on the vertical channels when there is a washer, since all the oil flow
of the winding must pass in this small channel.

7
a) b)

Figure 7 - Velocity magnitude in the upper, a), and lower, b), regions of the power transformer.

It is also possible to see that in the horizontal channels the velocity is higher in the channels just before
the washers. This becomes clearer in Figure 8 where the radial oil velocity component is plotted against
the vertical position for one of the main windings.

Figure 8 - Radial oil velocity in the high voltage winding as a function of the height.

8
This figure explains why the highest discs in each section are not the hottest ones. Although the oil
temperature increases with height, decreasing the thermal driving force for the winding cooling, the oil
velocity in the last channels increases substantially, removing more heat from the neighbor windings. The
channels in the middle of each section have the lower oil velocities, which are responsible for the higher
disc temperatures in these regions, as shown in Figure 6.

Other information that can be taken from the CFD simulations is the oil flow patterns. Figure 9 shows
some streamlines close to the entrance, in the middle of the winding and close to the oil outlets.

a) b)

c)
Figure 9 - Streamlines in the lower region close to the inlets, a), in the middle of the AT winding, b), and in
the upper region close to the outlets, c).

This figure shows that due to the 90º angles, the sharp oil direction changes cause the formation of
several vortices along the power transformer and when the oil has several flow possibilities, it is not easy
to guess in advance the way it will take. In the most critical region, the horizontal channels between

9
windings, vortices with different sizes are observed, depending on the oil velocity and channels size,
which shows that while the flow is basically laminar (max. Re= 280), care should be taken when trying to
predict the oil velocities and heat transfer coefficients with a network modeling approach using standard
correlations for fully developed flow.
Due to the complex geometry and oil flow patterns in power transformers, CFD is an essential tool to
analyze different configurations with a higher level of precision than standard correlations obtained from
simpler phenomena.

4. CONCLUSIONS

Computational Fluid Dynamic simulations were performed in CORE type power transformer, showing the
detail of the oil flow patterns and the consequent oil and windings temperature. It was shown that 2D axi-
symmetric simulations under predict the windings temperature since they do not take into account the
winding-oil heat transfer area reduction due to the presence of the radial and vertical spacers. Three
dimensional simulations are more precise and the results are closer to those estimated experimentally.
The temperature variation in the windings is predicted and it is shown that the maximum temperature is in
the upper section of the winding, due to the increasing oil temperature. The maximum temperature is not
however in the highest disc but in the intermediate ones, since there the oil velocity is lower, decreasing
the heat flux removal from the windings.
For future developments, optical fiber temperature measurements will be made in several points of the
power transformer in order to compare in more detail the CFD model and the experimental values.

5. BIBLIOGRAPHY

[1] Paulo J. Gomes, Renato G. Sousa, Madalena M. Dias, José C.B. Lopes, Manuel Silvestre,
Duarte Couto, Paulo Mesquita, Large Power Transformer Cooling – Flow Simulation and PIV
analysis in an Experimental Prototype, ARWtr 2007 – Advanced Research Workshop on
Transformers, 29-31 October, Baiona 2007

[2] Paulo J. Gomes, Renato G. Sousa, Joana I. Cardoso, Madalena M. Dias, José C.B. Lopes,
Manuel Silvestre, Duarte Couto, Jácomo Ramos, Paulo Mesquita, Mário Maia, Studies in a
Large Power Transformer – Heat Transfer and Flow Optimization using CFD, ARWtr 2007 –
Advanced Research Workshop on Transformers, 29-31 October, Baiona 2007

[3] J. Zhang et al., Experiments and modelling of heat transfer in oil transformer winding with
zigzag cooling ducts, Applied Thermal Engineering, Vol. 28, pg. 36-48, 2008.

[4] R. Hosseini et al., Determination of OD cooling system parameters based on thermal


modelling of power transformer winding, Simulation Modelling Practice and Theory, Vol. 16, pg.
585-596, 2008.

[5] D. Lussier et al., Theoretical determination of anisotropic effective thermal conductivity in


transformer windings, Int. Comm. Heat Mass Transfer, Vol. 30, No. 3 pp. 313-322, 2003.

10

Вам также может понравиться