Professional Documents
Culture Documents
net/publication/310517202
Dynamic top oil thermal model of oil-immersed power transformers with tap
changer
CITATION READS
1 196
4 authors, including:
Stefan Tenbohlen
Universität Stuttgart
378 PUBLICATIONS 3,690 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Stefan Tenbohlen on 23 November 2016.
Abstract
The paper presents the improved method for on-line calculation of top oil temperature. Basic of the model is well-
known lumped thermal circuit with nodes corresponding to each of phase windings and additional node for the core, oil
and the tank. Improvement in respect to the standards is consideration of the influence of temperature dependent
thermal conductivities (steady-state thermal characteristics) to the transient thermal characteristics. Change of the
cooling mode (AN - natural air flow / AF - forced air flow) and change of the number of fans in operation is taken into
account in the model. The finest content of the model, representing improvement in respect to the similar work
published in [1], is the manner of the calculation of the losses, including the dependence on the load, tap changer
position and average winding temperature.
oil over tank and cooler surfaces to outer air (W / K), ϑa - (PEddy) losses was considered as the losses in the tank and
ambient temperature varying in time (°C) constructive transformer parts (PConstr). In that way,
inaccuracy in eddy losses will reflect to winding to oil
2.2 Running simultaneous calculations temperature gradient, while in the calculation of oil
Naturally, the most critical temperatures should be calcu- temperature the error is minor.
lated, first of all the hottest winding temperature (hot- Total losses injected in the node with characteristic
spot); although the hottest oil in the winding is critical oil temperature ϑO are equal to the sum of no load (core)
temperature, typically it is not calculated in on-line losses (PFe) and PConstr, where PFe does not depend on the
applications and instead of it the top pocket oil is load and PConstr is proportional to the square of the load.
considered as characteristic (and critical) oil temperature.
The reasons are that as a rule oil temperatures are less 3 Determination of parameters of
critical than the temperature of solid winding insulation
and that the top oil can easily be measured, with standard
thermal network
temperature sensors. So, one of the calculations should be
done based on the thermal circuit with winding hot-spot 3.1 Possible approaches
temperatures and top oil temperature associated to the If detailed construction data are available, the steady state
nodes. thermal characteristics of the transformers can be
As stated, the input data for the calculations are the losses. determined via set of calculation using software for
The losses are dependent on the temperature, first of all thermal design based on detailed thermal-hydraulic
on winding temperature, specifically on the distribution of network model [4]. Such an approach was used in our
the temperature over the windings. Going into such previous paper [2]. Parameters of the model relevant for
deeper analyses (such as presented in our paper [3]) is not transient thermal characteristics cannot be easily
adequate approach to dynamic calculation of the determined. Namely, the thermal capacitance for
temperatures, especially when they should run online. For isothermal volume is equal to simple product of the
such applications the calculation method should be simple volume and specific heat capacity; nevertheless, the
and should not be demanding regarding computation temperature over the volumes (winding, for example, or
resources and time. Instead of it, the total losses in the oil) change and determining of equivalent heat capacity
windings are calculated at each of the discrete time steps, would require knowing the distribution of temperature
as the sum of two components, both proportional to the over each of the parts.
square of load and both being temperature dependent: Another approach, used here, relates to realistic situation
Joule losses, taken to be proportional to the average that no details about the construction are known, than
winding temperature, and eddy losses, taken to be there are the results of top oil temperature measurements
inversely proportional to the average winding during transformer operation, in conditions of variable
temperature. Since average winding temperature is ambient temperature and load current. In addition to these
needed for the calculation of the losses, separate on site measurements, it is supposed that the data of
calculation based on the thermal circuit is run in parallel, standard tests of the transformer are known.
where for characteristic temperatures associated to the
nodes for the windings average winding temperatures are 3.2 Data from standard transformer tests
taken (it also influences the parameters of the thermal
The nameplate transformer data are: ONAF/ONAN 40/24
circuit from Figure 1.).
MVA, 110(±11× 1.5%)/10.5 kV, YNd5, 50 Hz. The
Table 1 contains the results of the measurements in
2.3 Further details about the losses different tap changer positions (Tap 1 - maximum voltage
The eddy losses depend on the distribution of stray flux. It - 128.15 kV, Tap 12 - zero position, rated voltage 110 kV,
differs at different tap position. If the detailed Tap 23 - minimum voltage - 91.85 kV). Table 2 presents
construction of the transformer is known, distribution of the results of open-circuit tests, at different voltages; the
stray flux and the values of the stray losses in each of the tap changer was in the zero position (Tap position 12).
conductors can be determined. Such an approach and the
results of its application were presented in our paper [2] Table 1 - The results of the factory short-circuit tests
on the case of three phase transformer with high voltage, RC HV IHVr PDC PTot PAdd uSC
Tap
low voltage, and tertiary windings, YNyn0 + d1, 132 / (m) (A) (kW) (kW) (kW) (%)
13.8 / 11 kV, rated power in ONAF mode 40 / 40 / 13.33 1 632 180 128.8 258.1 129.3 20.2
MVA, regulating voltage range -12 x 1.43 % to + 7 x 5 585
1.43%. Since the construction of the transformer treated 9 537
in this paper was not known, the value of eddy losses at 12 500 210 118.6 216.9 98.37 18.0
different tap positions were adopted according to 15 482
experience. The difference of total load losses measured 19 435
in short-circuit test (PSC) and the Joule (PJ) plus eddy 23 386 251 142.3 212.5 70.18 16.44
these tap positions short-circuit voltage is higher, meaning winding temperature as described in [6] or hot-spot
the stray flux is bigger. Following this quantitative measured using fiber-optics) elements of thermal circuit
phenomena, it was approximately adopted that the ratio related to winding (parameters of thermal conductance
eddy losses to DC losses (at 75 °C) drop linearly from winding to oil - λCu-O1, λCu-O2, and thermal capacitance of
25% in tap position 1 to 15% in tap position 23 (limits the windings CCu1, CCu2) can be determined.
15% and 25% are adopted arbitrary). The values of the For transformer in the case study from the paper neither
losses in the tank and constructive transformer parts for online data nor data about transient change of winding
tap positions 1, 12 and 23, with known measured total temperature during heat run test are available. So, from on
losses, are determined after subtracting DC and eddy line data during transformer operation values of thermal
winding losses from measured total losses (the high parameters Kto and nto - see (3), and thermal capacity CO
values 97.105 kW, 60.441 kW and 48.820 kW, decreasing were estimated. Separate set of parameters were
with tap position increase, are obtained). After that, the determined: one for AN cooling mode and one for AF
losses in the tank and constructive parts for rated current cooling mode (fans are switch on seldom: typically about
at other tap positions are determined using interpolation 10 hours during hottest summer days (in 2013. there were
(for tap positions 2 - 11 an interpolation between the such 18 days).
values for tap positions 1 and 12 is used, and for tap Using data set (for AN of one week duration and for AF
positions 13 - 22 an interpolation between the values for of ca. 10 hours of operation in this cooling mode), the pa-
tap positions 12 and 23 is used). The results are presented rameters (CO, Kto and nto) are determined as the values
in Table 3. when minimum deviation (average root mean square
deviation over duration of considered data set) of
Table 3 - Estimated values of losses (kW) at rated current calculated to measured top oil temperatures.
and temperature 75 °C
PDC PEddy PTotWind PTot PTankConstr 3.6.1 AN cooling mode (no fans in operation)
Tap
(kW) (kW) (kW) (kW) (kW) The characteristic data in one week of each of the months
1 128.8 32.19 160.9 258.05 97.11 for which estimation of the parameters is made are
5 130.9 30.345 161.2 83.77 presented in Table 4. The results presented in Table 5 are
9 132.8 28.38 161.23 70.44 obtained using Excel Solver tool [7], with constrain
12 134.1 26.82 160.9 216.9 60.44 nto <= 0.5 (typical coefficient, corresponding to the
15 138.5 25.80 164.3 57.27
natural air cooling, is 0.25 - it corresponds to the oil
19 140.7 23.66 164.4 53.05
exponent x = 1 / (1 + 0.25) = 0.8; for the forced air flow,
23 142.3 21.35 163.7 212.5 48.82
the coefficient x is higher, i.e. nto is lower). For each of
the months the average root mean square deviations of
PDC - The DC losses in the windings, PEddy - The eddy
calculated in respect to measured top oil temperature
losses in the windings, PTot Wind - The total losses in the
(ΔϑRMS) are given. The procedure did not converge for
windings, PTot - The total measured losses, PTankConstr - The
one week data record in March (N.A. in Table 5).
losses in the tank and constructive transformer parts
Losses in the windings were determined starting from the
Table 4 – Characteristic temperatures and average losses
values in Table 3, and the current value of the load,
in each of considered weeks
average winding temperature and tap position, as
Feb Mar Apr May Jun Jul
described in [2]; the losses in the tank and constructive
From 14. 6. 14. 1. 14. 8.
parts are taken to be equal to the product of their rated To 20. 12. 20. 7. 20. 14.
value for specific tap position (given in column PTankConstr
aAV 4.46 12.9 15.0 23.2 25.7 23.1
in Table 4) and square of relative current load. PTotAV 90.3 65.4 41.0 31.3 36.6 31.0
OAV 45.4 43.2 40.7 43.9 50.0 43.8
3.6 Estimation of the parameters from on Aug Sep Oct Nov Dec
site measurements From 1. 4. 5. 1. 11.
As stated in section 3.1., it is possible to determine pa- To 7. 10. 11. 7. 17.
rameters of thermal circuit from available on line data aAV 27.6 19.3 14.9 16.0 0.42
recorded during normal transformer operation. Typically, PTotAV 34.6 42.1 61.0 41.1 96.0
load, ambient temperature, tap position and top oil tem- OAV 50.5 45.0 44.9 38.4 43.7
perature are available. From these data parameters of
thermal conductance describing heat transfer from the oil Table 5 –The results of estimation of all 3 parameters
to ambient (λO-a) and the thermal capacitance of oil, core (limitation: nto <= 0.5)
and tank with radiators (CO) can be determined. If data Feb Mar Apr May Jun Jul
about winding temperature are known (average winding Kto 0.81 N.A. 0.31 0.34 0.30 0.33
temperature using continuous measurements of average nto 0.27 N.A. 0.5 0.5 0.5 0.5
CO 66.6 N.A. 44.8 53.7 41.6 42.7 occurred, yielded with constant value of the thermal
RMS 0.88 N.A. 1.91 2.21 1.66 1.71 conductivity (nto = 0) of Kto = 1.749kW/K. Comparable
value, at rated losses in AN cooling mode, of thermal
Aug Sep Oct Nov Dec conductivity in AN cooling mode amounts 1.0605 kW/K,
Kto 0.30 0.32 0.37 0.38 0.33 i.e. thermal conductivity in the AF cooling mode is
nto 0.5 0.5 0.5 0.5 0.5 1.749 / 1.0605 = 1.649 times higher than in the AN
CO 45.2 51.4 58.1 64.2 67.4 cooling mode. This number corresponds and is near equal
RMS 1.62 2.13 1.85 2.05 1.38 to the ratio of transformer rated powers in AF (40 MVA)
and AN (24 MVA) cooling modes: 40 / 24 = 1.667.
The graph for the week with the characteristic pattern and
high average root mean square deviation of top oil
temperature, ΔϑRMS = 2.13 K, in September, is presented
4 Application of the model
in Figure 2. This section presents few graphics of application of the
method and parameters specified in Table 6 for AN mode
and in section 3.6.2 for AF cooling mode.
7 Literature
[1] Weigen, C.: Chong, P.: Yuxin, Y. :Power
transformer top-oil temperature model based on
thermal–electric analogy theory, Euro. Trans. Electr.
Power. Vol. 19, 2009, pp. 341-354
[2] Radakovic Z.: Tenbohlen S.: Thermal model of oil
power transformers with tap changer, Turk J Elec
Eng & Comp Sci. Vol. 24, 2016, pp. 3293-3308
[3] Radakovic, Z.: Radoman, U.: Kostić, P.:
Decomposition of the Hot-Spot Factor, IEEE Trans.
on Power Delivery. Vol. 30, No. 1, 2015, pp. 403-
411
[4] Radakovic, Z.: Sorgic, M.: Basics of Detailed
Thermal-Hydraulic Model for Thermal Design of Oil
Power Transformers, IEEE Trans. on Power
Delivery. Vol. 25, No. 2, 2010, pp. 790-802
[5] IEC Publication 60354, Loading Guide for Oil-
Immersed Power Transformers. Geneva,
Switzerland: IEC, 1991
Figure 5 Results for the 14 hour period with fans in [6] Radakovic, Z.: Numerical determination of
operation (from 21 Jun, 12 h to 22 Jun 2h), for characteristic temperatures in directly loaded power
Kto = 1.749 kW / K, nto = 0 oil transformer, European Transaction on Electrical
Power (ETEP). Vol. 13, No. 1, 2003, pp. 47-54
[7] The Microsoft Office 2010, Microsoft Excel, Solver
5 Conclusions Add-in.
In previous paper [2] we published the method which
includes key factors influencing top oil temperature:
ISBN 978-3-8007-4310-0
View publication stats
521 © 2016 VDE VERLAG GMBH ∙ Berlin ∙ Offenbach