Professional Documents
Culture Documents
Abstract—Sales of privately-owned plug-in electric vehi- replacement [1]–[3]. For instance, Level 2 charging of PEVs
cles (PEVs) are projected to increase dramatically in coming draws 7.2 kVA of power from the grid. For a 25 kVA ser-
years and their charging will impact residential service trans- vice transformer, this amounts to 29% of the rated load. The
former loads. Transformer life expectancy is strongly related to
the cumulative effects of internal winding temperatures, which average American household has two vehicles [4]. Given that
are a function of loading. Thermal models exist (e.g., IEEE a single service transformer may serve ten houses, PEV pen-
Standard C57.91) for predicting these internal temperatures, the etration rates of 25% could load a 25 kVA device to 144%
most sophisticated being the Annex G model. While this model of rated load, if the vehicles charge simultaneously. Because
has been validated with measurements from large power trans- charge times for PEVs are expected to overlap, even low pen-
formers, small residential service transformers have been given
less attention. Given increasing PEV loads, a better understand- etrations of PEVs can create harmful overloading for service
ing of service transformer aging could be useful in replacement transformers [5].
planning processes. Empirical data from this paper indicate Power transformers, such as those serving multiple neigh-
that the Annex G model over-estimates internal temperatures borhoods and rated upward of 500 kVA, are typically mon-
in small 25 kVA 65 ◦ C rise mineral-oil-immersed transformers. itored closely as their failure in the grid can be costly to
This paper presents an alternative model to Annex G by
using a genetic program. Empirical results using a thermally- utilities and can cause prolonged outages [6]. However, the
instrumented transformer suggest that this model is both simpler cost of monitoring each service transformer on the network
and more accurate at tracking empirical transformer data. We is generally too expensive, computationally or monetarily,
conclude that one can use a simple thermal model in combina- and approximations are used to determine device lifetimes.
tion with data from advanced metering infrastructure to more However, a utility that has advanced metering infrastruc-
accurately estimate service transformer lifetimes, and thus better
plan for transformer replacement. ture (AMI) has access to accurate loading history for its service
transformers and can estimate wear based upon these data. The
Index Terms—Asset management, electric vehicles, genetic expense of tracking these assets are then mostly computational.
programming, power transformers, smart grids.
Where loading is known for a transformer, artificial neural
networks (ANN) have been proposed to match the loading pro-
I. I NTRODUCTION file of a new transformer to the loading profiles of transformers
HIS PAPER focuses on methods used to estimate the in a recorded database to yield an estimate of remaining life-
T lifetime of service transformers in residential areas. The
term service transformer is used here to define the pole- or pad-
time of the new device [7]. Because this paper looks at the
damage caused by new PEV charging loads, representative
mounted transformers that directly serve residential loads. The databases are likely not available. Also, ANNs do not give
service transformers considered in this paper are 25 kVA 65 ◦ C the utility an intuitive understanding of how they predict their
rise mineral-oil-immersed devices. Though these particular outcomes. Another approach is to probabilistically estimate
assets are fairly inexpensive–around $700 for pole-mounted the number of transformers that will require replacement as a
and $1500 for pad-mounted plus installation costs–the entirety function of predicted PEV adoption rates as discussed in [2].
of the fleet will typically constitute a significant fraction of a While this analysis is helpful to utilities, it only helps for bud-
distribution utility’s physical assets. geting, leaving the real-time health of specific transformers in
Sales of plug-in electric vehicles (PEVs) are expected question. Finally, a thermal model can be used which can yield
to greatly increase. Inflating adoption rates are predicted the desired estimates on transformer life while preserving the
to stress the grid and require distribution transformer connection between how the model works and our intuition
about how aging occurs in transformers. Such models can be
Manuscript received February 15, 2014; accepted December 16, 2014. Date
of publication January 28, 2015; date of current version February 16, 2015. found in [8].
This work was supported in part by the U.S. Department of Transportation, In general, loading transformers causes heating in the inter-
through the University of Vermont Transportation Research Center, nal windings and degrades the insulating material. The effect
under Grant TRC039, and in part by Green Mountain Power.
Paper no. TSG-00107-2014. of overloading is to cause accelerated aging compared to a unit
The authors are with the School of Engineering, University of Vermont, that is loaded to its rated power. If enough overloading accu-
Burlington, VT 05405 USA (e-mail: aseier@uvm.edu). mulates in a single transformer, its lifetime can be significantly
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. curtailed. Transformer aging theory and modeling methods are
Digital Object Identifier 10.1109/TSG.2015.2390624 discussed in the IEEE Guide to Loading Mineral-Oil Immersed
1949-3053 c 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on April 12,2022 at 09:40:32 UTC from IEEE Xplore. Restrictions apply.
1020 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 2, MARCH 2015
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on April 12,2022 at 09:40:32 UTC from IEEE Xplore. Restrictions apply.
SEIER et al.: DATA-DRIVEN THERMAL MODELING OF RESIDENTIAL SERVICE TRANSFORMERS 1021
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on April 12,2022 at 09:40:32 UTC from IEEE Xplore. Restrictions apply.
1022 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 2, MARCH 2015
TABLE I
IV. GP M ODELING S TRUCTURES A RISING F ROM GP AND F OUR S ELECTED S TRUCTURES
With regard to the complexity of the Annex G model, the T RAINED ON THE DATA IN F IG . 4
following section seeks to find a simpler thermal model that
tracks internal temperatures as well as the IEEE Annex G
model and we detail a broad approach by which a ther-
mal model can be created to accurately describe the internal
heating of service transformers based on user-defined input
variables.
To model the dynamics of the Annex G model, we start
with a HS temperature profile created by this model. To do
this, we input ambient temperature data and actual loading
data, which was increased to 150% of the measured values
to bring the internal temperatures closer to the transformer’s explanation and defense of this process is beyond the scope
limits and to expose the dynamics of the model. The Annex G of this paper, but is contained in [12].
HS temperature (THS ), the ambient temperature (TA ), and the As is common with GPs, it is up to the user to select a
loading (L) are used as inputs to the GP. It should be noted reasonable solution [12], [13], [15]. Solution (a) from Table I
that variables TO and TW , which are other outputs from the is not a good selection because it is constant that doesn’t
Annex G model, could have been used as additional inputs to depend on any of the input variables. Models (b) and (c) do
the GP. In this way, we would have to actually model multiple not depend on TD , which is connected to TA , a known input to
differential equations, one for each variable, and come up with the Annex G model. Therefore, we do not choose these either.
a set of equations that would need to be integrated forward Models (h) and (i) begin to create complicated polynomial fits
together to find the HS temperature, THS . Again, it is our from the variable L and are also not chosen.
purpose here to simplify, so we choose to relate the input We continue with solutions (d)–(g), shown in bold. It will
variables to one of the original outputs of the Annex G model, be shown that one of these models can be considered accept-
the HS temperature. able when compared to the Annex G model in the following
Heuristically, we would expect that the change in subsections. If this were not the case, the next logical step
˙ ) will depend on the difference between the HS
THS (i.e., THS would be to include more of the models from Table I to see if
temperature, THS , and the ambient temperature, TA . Hence, to any of these yield better results. If no models can ultimately
simplify the search space, we define TD as shown in (4), and be selected by the end of this process, the assumption is that
we seek a differential equation for the change in HS temper- the solution space is not sufficient and more input variables,
ature as shown in (5), which is another way of describing (2) beyond our chosen L and TD , need to be added and the GP
TD = THS − TA (4) must be rerun. For reference, the resulting models from the
˙ = f (L, TD ). training discussed in the following subsection are shown in
THS (5)
Table I.
The allowable operations for these input variables were:
1) addition; 2) subtraction; 3) multiplication; and 4) negation. B. Fitting Selected Structures to Training Data via Least
This search space has closure, meaning that solutions may be Mean Squares
composed of results from other solutions, and it is also rea- The next step is to take the structures obtained from running
sonable to assume that it is sufficient, meaning that the input the GP and find a best LMS fit to make the actual models.
variables and operations are enough to describe the dynamics. In making a model, we must train it on certain data and then
The justification of sufficiency comes with the acceptance of check its validity on new data. These models are trained on
a model at the end of this section. the inputs shown in Fig. 4, where the loading profile is an
hour long pulse and the temperature profile is a ramp. During
A. Running the GP and Selecting Solutions research, training the models with fictitious data like the load
To acquire our candidate thermal models, we ran the GP pulse and temperature ramp was more successful in finding
for over six hours with 16 cloud cores. To reiterate, the GP a best fit. The short load pulse and temperature ramp allows
started with a randomly selected subset of mathematical struc- us to separate out dynamics due to loading and temperature
tures from the infinite solution space created by the assumed differentials.
variables and mathematical operators of interest. These solu- As specified in [8], emergency overloads are intentional
tions were then mutated and recombined by the GP until the overloads that last for only a short duration. Loading guides
GP was manually terminated as discussed in Section III-B. in this document cite a maximum emergency overload limit
After this, what remains is a population of solutions repre- of 2 p.u., thus though the load pulse shown is fictitious, it is
senting the most fit individuals; the ten simplest out of the not an unreasonable estimate to a loading scenario. Loading
final population are considered here. They are shown, ordered goes from very underloaded, at 0.5 p.u. or 12.5 kVA, to
by complexity and labeled alphabetically (a)–(i), in Table I. overloaded, at 1.5 p.u. or 37.5 kVA. The temperature ramp
This table labels each solution equation structure with a let- changes from 0 ◦ C to 40 ◦ C during an interval in the mid-
ter which will be referenced throughout this paper. Detailed dle of the day. Such low temperatures help the model train
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on April 12,2022 at 09:40:32 UTC from IEEE Xplore. Restrictions apply.
SEIER et al.: DATA-DRIVEN THERMAL MODELING OF RESIDENTIAL SERVICE TRANSFORMERS 1023
Fig. 4. THS curve is produced to train structures (d)–(g). The loading pro- Fig. 5. Model validation via a comparison with independent THS predic-
file (top) and ambient temperature profile (middle) are used to create the tions for low loading. The loading profile (top) and ambient temperature
Annex G predicted THS temperature (bottom-black). Coefficients of struc- profile (middle) are used to create Annex G and model (d)–(f) THS pre-
tures (d)–(g) are set to minimize differential equation error and then integrated dictions (bottom) to validate the models for low loading scenarios.
forward and are shown compared to the Annex G prediction (bottom-colors).
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on April 12,2022 at 09:40:32 UTC from IEEE Xplore. Restrictions apply.
1024 IEEE TRANSACTIONS ON SMART GRID, VOL. 6, NO. 2, MARCH 2015
TABLE II
E RROR TABLE : M EAN S QUARED -E RROR (MSE) IN ◦ C/min2 AND E RROR losses associated with the modeled resistance of the core
PER P OINT (EPP) IN ◦ C F ROM THE T RAINING (T), THE L OW L OAD (L), and the resistance of the winding. These power losses mani-
AND THE H IGH L OAD (H). M ODELS R EFER TO T HOSE IN TABLE I
fest themselves as heating and are proportional to the square
of the current through the resistors. Because the voltage at
a transformer is relatively constant, the load (L) served by a
transformer is proportional to the current through the windings.
The power loss in the winding resistance is then proportional
to the square of this instantaneous current. Hence, the L2 term
in (6) supports this rationalization. Furthermore, Newton’s
Law of Cooling tells us that a body cools in proportion to the
units of ◦ C/min2 because it is relating to the change in the HS
difference between its temperature and the ambient tempera-
temperature. Also tabulated are the EPP which is a measure of
ture, i.e., the term TD from (6) [16]. Finally, the constant in
how symmetric the error is. This number is found by summing
this equation may be an adjusting factor since we are directly
the errors and dividing by the number of data points and has
relating an internal temperature to the ambient temperature.
units of ◦ C. Together the MSE and EPP yield a more com-
Such physical insight would not be obtained using alternative
plete description of the model’s ability to track the Annex G
empirically-based modeling methods such as ANNs.
predictions. From Table II, model (f) has the most consistently
low EPP and has low MSE.
V. C ONNECTION TO A SSET M ANAGEMENT
D. Results From Modeling Approach Thus far, we have explained how the Annex G model may
This process has shown that many models came organically not be appropriate for describing the heating measured exper-
out of a GP. From these models, four were selected for further imentally in our instrumented service transformers. We have
study that seemed to be both well related to the physics of the also proposed an approach by which a simple thermal model
actual problem and reasonably simple. The actual coefficients can be found for service transformers. This section seeks to
for these models were thrown away so that the structures were explain how a utility would be able to leverage such a model
kept. They were then refitted with new coefficients using fic- for better asset management of its service transformer fleet.
tional training data and the least mean squares method. To The large fleet of service transformers owned by utilities is
understand whether these fits worked for the specific data or an important subset of assets that must be appropriately man-
generally for other data sets, the models were validated by aged. Given that the utility in question has already installed
integrating them forward with new data. These results show AMI, access to loading data for these transformers is relatively
that, qualitatively, all of the models tracked both the changes in inexpensive. This includes getting local ambient temperatures
temperature of the HS, as predicted by the Annex G method, from nearby national weather stations (e.g., the KBTV radio
as well as the overall accumulated HS temperature. We showed station for South Burlington). Initial results show that the
that the best model for the Annex G method was (f) as restated Annex G thermal model for loading these transformers is
in (6). The achievement here is finding a differential equation overly conservative in that it predicts internal temperatures to
that use only two measurable inputs, three terms, and still be higher than what are experimentally recorded. In addition,
tracks the Annex G output reasonably well for very different the cumbersome nature of the model along with the large num-
loading data sets ber of device-specific constants required, presents a barrier to
˙ ≈ f (L, TD ) = 0.178 + 0.000939L2 − 0.0149TD . utilities. The approach to modeling hot-spot temperatures can
THS (6)
be used by utilities to find a concise, intuitive model which
Recall that the Annex G model is made up of a large set can leverage loading data.
of differential equations. Thus, the model presented in (6) In this way, utilities see the effect of increased loading
manages to follow the dynamics of the Annex G model’s pre- from PEV charging as it appears. This requires neither a
dictions of the HS temperature [Fig. 5 (bottom)], while being history of transformer failure as a comparison nor a cum-
much simpler. Model (f) is not meant to be a replacement for bersome model which is unintuitive and computationally
the Annex G model for all transformers. However, we have intense. The method only depends on evidence that insula-
conjectured that a 25 kVA service transformer may not require tion pyrolysis, the destruction of the material via heating, is
the full complexity of the Annex G model and have shown that the determining factor in transformer aging. With up-to-date
the structure of model (f), using only one differential equation, information on service transformer aging, utilities can pre-
has been validated to predict HS temperatures in accordance emptively install additional capacity before certain devices
with the IEEE standard model. It should be noted that this breakdown and appropriately budget for future device purchase
model has been trained and validated against the Annex G and installation.
model for a single set of transformer parameters. This means In addition, the approach explained in this paper is flex-
that the model coefficients may need to be tuned for use with ible to the input data used. The Annex G model uses only
slightly different transformers. loading and ambient temperature as inputs, though other fac-
It is important to note that the model in (6) makes physi- tors (e.g., wind speed, solar intensity, etc.) may be important in
cal sense, given thermodynamics and circuit theory. If we properly modeling transformer heating. By measuring this data
recall the equivalent circuit for a transformer, there are power and using it as an input to the approach shown in Section IV,
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on April 12,2022 at 09:40:32 UTC from IEEE Xplore. Restrictions apply.
SEIER et al.: DATA-DRIVEN THERMAL MODELING OF RESIDENTIAL SERVICE TRANSFORMERS 1025
such dependencies can be included. For example, work with [6] M. Pradhan and T. S. Ramu, “On the estimation of elapsed life of oil-
ANNs has shown that current harmonics in devices may have immersed power transformers,” IEEE Trans. Power Del., vol. 20, no. 3,
pp. 1962–1969, Jul. 2005.
an effect on heating which is not captured by the Annex G [7] J. Jardini, H. Schmidt, C. M. V. Tahan, C. C. B. De Oliveira, and
model [17]. Another factor that may be of import is hydrolysis, S. U. Ahn, “Distribution transformer loss of life evaluation: A novel
which has been shown to effect insulation aging for transform- approach based on daily load profiles,” IEEE Trans. Power Del., vol. 15,
no. 1, pp. 361–366, Jan. 2000.
ers loaded far below their rated limits [18]. Finally, utilities [8] IEEE Guide for Loading Mineral-Oil-Immersed Transformers, IEEE
would also have the option to include load management, Standard C57.91-1995, 2012, pp. 1–112.
and control loading based on calculated internal temperatures. [9] G. Swift, T. Molinski, and W. Lehn, “A fundamental approach to trans-
former thermal modeling. I. Theory and equivalent circuit,” IEEE Trans.
Monitoring their transformer fleet in this way would allow for Power Del., vol. 16, no. 2, pp. 171–175, Apr. 2001.
more accurate and flexible management. [10] M. Rosillo, C. Herrera, and G. Jaramillo, “Advanced thermal modeling
and experimental performance of oil distribution transformers,” IEEE
Trans. Power Del., vol. 27, no. 4, pp. 1710–1717, Oct. 2012.
VI. C ONCLUSION [11] D. Susa, M. Lehtonen, and H. Nordman, “Dynamic thermal mod-
elling of power transformers,” IEEE Trans. Power Del., vol. 20, no. 1,
Initial data collected for this paper indicate that the Annex G pp. 197–204, Jan. 2005.
model is overly conservative in that it overestimates internal [12] J. Koza, Genetic Programming: On the Programming of Computers by
temperatures for the instrumented 25 kVA service transformers Means of Natural Selection. Cambridge, MA, USA: MIT Press, 1992.
[13] J. Bongard and H. Lipson, “Automated reverse engineering of nonlin-
used in this paper (see Fig. 2). Because a precise estimate of ear dynamical systems,” Proc. Nat. Acad. Sci. USA, vol. 104, no. 24,
a transformer’s HS temperature is the standard indicator for pp. 9943–9948, Jun. 2007.
insulation aging, an approach to finding a more appropriate [14] M. Schmidt and H. Lipson, “Distilling free-form natural
laws from experimental data,” Science, vol. 324, no. 5923,
model was detailed. The approach used a GP to remodel out- pp. 81–85, 2009. [Online]. Available: http://www.sciencemag.org/
put from the Annex G model to indicate the effectiveness of content/324/5923/81.abstract
the method. The result was a first order differential equation, [15] H. Kamal and M. Eassa, “Solving curve fitting problems using genetic
programming,” in Proc. 11th Mediterr. Electrotech. Conf. (MELECON),
which was far simpler than the Annex G model, and tracked Cairo, Egypt, 2002, pp. 316–321.
the more complicated model well. Our existing measurement [16] L. C. Burmeister, Convective Heat Transfer, 2nd ed. Hoboken, NJ, USA:
system did not provide data for HS temperature; collection Wiley, 1993.
[17] J. Pylvanainen, K. Nousiainen, and P. Verho, “Studies to utilize loading
and modeling of HS data remains for future work. guides and ANN for oil-immersed distribution transformer condition
Given that utilities have increasing access to residential load monitoring,” IEEE Trans. Power Del., vol. 22, no. 1, pp. 201–207,
time-series data and that weather data are readily available, Jan. 2007.
[18] A. Kachler and I. Hohlein, “Aging of cellulose at transformer service
real-time monitoring of service transformers is a realistic goal. temperatures. Part 1: Influence of type of oil and air on the degree of
If these assets are monitored, utilities can gain information on polymerization of pressboard, dissolved gases, and furanic compounds in
how much additional loading capacity exists at the residential oil,” IEEE Elect. Insul. Mag., vol. 21, no. 2, pp. 15–21, Mar./Apr. 2005.
level to serve predicted PEV charging requirements. They can
also closely monitor the aging rates of these devices and make
Andrew Seier (S’11) received the B.S. and M.S. degrees from the University
decisions on how to schedule future replacements and upgrade of Vermont, Burlington, VT, USA, in 2011 and 2013, both in electrical
these devices proactively instead of reactively. Therefore, by engineering.
leveraging data collected by AMI, utilities can more closely He is currently a Developer in Python and Javascript for Plotly, Montreal,
QC, Canada, which handles data visualization for engineers and scientists.
monitor aging in assets that were previously difficult to track.
R EFERENCES Paul D. H. Hines (S’96–M’07–SM’14) received the B.S. and M.S. degrees
[1] L. Fernández, T. Román, R. Cossent, C. Domingo, and P. Frías, in electrical engineering from the University of Washington, Seattle, WA,
“Assessment of the impact of plug-in electric vehicles on distribu- USA, and Seattle Pacific University, Seattle, in 1997 and 2001, respectively,
tion networks,” IEEE Trans. Power Syst., vol. 26, no. 1, pp. 206–213, and the Ph.D. degree in engineering and public policy from Carnegie Mellon
Feb. 2011. University, Pittsburgh, PA, USA, in 2007.
[2] J. Sexauer, K. McBee, and K. Bloch, “Applications of probability He is currently an Assistant Professor with the School of Engineering,
model to analyze the effects of electric vehicle chargers on distribution Department of Computer Science, University of Vermont, Burlington,
transformers,” IEEE Trans. Power Syst., vol. 28, no. 2, pp. 847–854, VT, USA.
May 2013.
[3] Q. Gong, S. Midlam-Mohler, V. Marano, and G. Rizzoni, “Study of PEV
charging on residential distribution transformer life,” IEEE Trans. Smart Jeff Frolik (S’85, M’95, SM’11) received the B.S. degree in electrical engi-
Grid, vol. 3, no. 1, pp. 404–412, Mar. 2012. neering from the University of South Alabama, Mobile, AL, USA; the M.S.
[4] United States Department of Transportation. (Jan. 2015). Number degree in electrical engineering from the University of Southern California,
of Households by Household Driver Count. [Online]. Available: Los Angeles, CA, USA; and the Ph.D. degree in electrical engineering sys-
http://nhts.ornl.gov/tables09/FatCat.aspx tems from the University of Michigan, Ann Arbor, MI, USA, in 1986, 1988,
[5] A. Hilshey, P. Hines, and J. Dowds, “Estimating the acceleration of and 1995, respectively.
transformer aging due to electric vehicle charging,” in Proc. IEEE Power He is an Associate Professor with the School of Engineering, University
Energy Soc. Gen. Meeting, San Diego, CA, USA, 2011, pp. 1–9. of Vermont, Burlington, VT, USA.
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on April 12,2022 at 09:40:32 UTC from IEEE Xplore. Restrictions apply.