Temperature rise prediction of a natural cooling

dry-type transformer
W. Wu* and J. A. Kern
ABB Inc., Bland, VA, USA
* wei.wu@us.abb.com

 solid insulation dry-type transformer are calculated by using a

Abstract—Dry-type transformer applications are growing in well-developed thermal network model and a CFD simulation
transformer market because the technology is non-flammable, model. Then the results are compared with heat-run test results
safer and environmental friendly. Since no oil is present in a dry- as modelling accuracy validation.
type transformer for dielectric insulation, solid insulation is often
used instead for compact dimensions which however constrains
air convection as the major cooling approach. Therefore, how to II. THERMAL NETWORK MODEL
accurately predict winding temperature rises when designing a Compared with CFD, thermal network model involves more
transformer is one of the significant tasks of a transformer approximations which will dramatically shorten the computing
manufacturer. In this context, the present paper uses a natural
speed and the accuracy shall be maintained in an acceptable
cooled, solid insulation dry-type transformer as an example to
illustrate the temperature rise prediction using a fast-calculating range.
thermal network model and a highly discretized computational The governing equations for describing heat transfer have
fluid dynamics (CFD) model. The model accuracies were then similar formats with the ones for electric circuits; also known
validated by comparing the prediction results with the heat-run as ‘thermo-circuit analogy’. For instance, Fourier’s law in heat
test results. transfer and Ohm’s law in electric circuits are analogous.
Thermos-circuit can then be solved with analogous Kirchhoffs’
Keywords—Dry-type transformer; temperature rise; thermal
model; network model; CFD; heat-run test
circuit laws. The analogous quantities have been summarized
into TABLE I; similar analogy applies upon fluid flow as well.
I. INTRODUCTION
TABLE I

T RANSFORMERS are essential components of electric

transmission and distribution networks. Among various
transformer technologies, dry-type technology avoids using oil
Analogy to Electric Circuit Principles
Electric Thermal Fluid flow
Through Current Heat transfer rate Mass flow rate
as dielectric or cooling media and is non-flammable, safer and variable I q Q
more environmental friendly. As such, dry-type transformers [Amps] [Watts] [kg/s]
have gained market share in special environments and Across Voltage Temperature Pressure
applications that require high standard of safety and reliability variable V θ P
such as urban areas, buildings and marine ships etc. [Volts] [oC] [Pascal]
Dissipation Electric Thermal Hydraulic
However, in order to obtain sufficient dielectric insulation
element resistance resistance impedance
and cooling, dry-type transformers mostly need bigger size Rel Rth Rp
and, consequently, more material cost. Therefore, cooling has [Ohms] [oC/Watt]
to be accurately predicted while one attempts to optimize a
transformer design. Numerical models have been developed Based on the analogy, network model encapsulates physics
for predicting winding temperature rises and they generally fall principles into a number of interconnected so-called ‘lumped
into two categories; either ‘thermal network models’ [1-3] or elements’, which then compose a thermal network and a fluid
approaches which incorporate a degree of computational fluid flow network. Both networks are coupled together and will be
dynamics (CFD) [3-5]. CFD is flexible and can reveal more numerically solved with iterations.
details of the occurring thermodynamic phenomena; however, The core and coil structure schematic of a vacuum cast dry-
it is with considerably higher level of discretization and thus type transformer is illustrated in Fig. 1 [6]. Surrounding the
consumes much more computational resources (processor and iron core, the coil comprises layer low voltage (LV) and disk
memory) and time. In comparison, thermal network models high voltage (HV) windings which are casted in epoxy.
run faster (less than a second for a typical dry-type design) and Because of the geometric symmetry, the structure is interpreted
are more widely used for transformer designing and into a 2D network model topology shown in Fig. 2. Each part
optimization. of the transformer structure has a corresponding node in the
In this study, winding temperature rises of a natural cooled, network topology, as labeled in the figure; so after the model
has been solved, the average temperature of each node can be
obtained.

978-1-5090-2246-5/16/$31.00 ©2016 IEEE

Fig. 1. Schematic view of a dry-type transformer core and coil [6].

Fig. 3. CFD simulation results of temperature distribution.

TABLE II
Comparison of thermal network model, CFD and test
temperature rise results, K
Thermal CFD Test
network model results
model
LV winding rise 105.6 99.7 102
HV winding rise 80.5 61.0 88

V. CONCLUSIONS AND FUTURE WORK

The study has demonstrated the applicability of the thermal
network and CFD models to a natural cooling, solid insulation
Fig. 2. Thermal and fluid flow network model topology; this is a 2D model dry-type transformer. Validation shows that the LV winding
and the left side dash line is the symmetric axis.
has a simpler construction and therefore has got better
modelling accuracy. Future studies will include correcting the
III. CFD MODEL over-estimation of the HV cooling. The details of the HV turns
CFD utilizes a set of general purpose numerical techniques and turn insulation will need to be taken into account, resulting
to solve Navier-Stokes equations and analyze applications that in a lower effective thermal conductivity, in order for the
involve fluid flows; transformer temperature evaluation is one expected accuracy for the HV winding to be improved.
of these applications. Because of the complexity of CFD
modelling, one usually chooses mature commercial or open- VI. REFERENCES
source CFD software packages in practice. For example the [1] W. Wu, Z. Wang, A. Revell, H. Iacovides, and P. Jarman,
commercial package SOLIDWORKS Flow Simulation [7] was "Computational fluid dynamics calibration for network modelling of
used for this study. One of the advantages is that the 3D CAD transformer cooling oil flows-part I heat transfer in oil ducts," IET
geometry in SOLIDWORKS can be used without Electric Power Applications, vol. 6, issue 1, pp. 19-27, Jan. 2012.
modification; mesh is generated natively in the software. The [2] W. Wu, Z. Wang, A. Revell, and P. Jarman, "Computational fluid
dynamics calibration for network modelling of transformer cooling
solver calculates both fluid flow and solid heat conduction flows–Part II: pressure loss at junction nodes," IET Electric Power
regions together. In particular, k-ε model is used for Applications, vol. 6, issue 1, pp. 28-34, Jan. 2012.
considering turbulence flow; radiation is taken into account as [3] A. Weinläder, W. Wu, S. Tenbohlen, and Z. Wang, "Prediction of the
well. oil flow distribution in oil-immersed transformer windings by network
Although CFD simulation takes longer time than network modelling and computational fluid dynamics," IET Electric Power
model (hours versus seconds), it is able to reveal the details of Applications, vol. 6, issue 2, pp. 82-90, Feb. 2012.
[4] A. Skillen, A. Revell, H. Iacovides, and W. Wu, "Numerical prediction
the temperature distribution; for instance in Fig. 3. of local hot-spot phenomena in transformer windings," Applied Thermal
Engineering, vol. 36, pp. 96-105, Apr. 2012.
IV. MODEL VALIDATION AND DISCUSSION [5] W. Wu, A. Nogues, and J. Kern, "OpenFOAM CFD modelling on
forced air cooling dry-type transformers," presented at ESI OpenFOAM
The modelling results are compared with heat-run results in User Conference 2015, Stuttgart, Germany, 2015.
TABLE II. The LV layer winding rise matches the test value [6] E. Rahimpour, and D. Azizian, "Analysis of temperature distribution in
better than the HV disk winding. This is because HV has more cast-resin dry-type transformers," Electrical Engineering, vol. 89, issue
turns and thus a more complex construction which needs a 4, pp. 301-309, Apr. 2006.
degree of simplification. Particularly the CFD model is still [7] A. Sobachkin, and G. Dumnov, "Numerial Basis of CAD-Embedded
CFD," SOLIDWORKS White Paper, Feb. 2014. [Online]. Available:
preliminary, it neglected the turn insulation and modelled the https://www.solidworks.com/sw/docs/Flow_Basis_of_CAD_Embedded
winding blocks as conductor material. This over-estimated the _CFD_Whitepaper.pdf
cooling and temperature rises were therefore calculated lower.