Transformer Overloading

and Assessment of Loss-of-Life

for Liquid-Filled Transformerss

Final Project Report

Power Systems Engineering Research Center

Empowering Minds to Engineer

the Future Electric Energy System
 Transformer Overloading
and Assessment of Loss-of-Life
for Liquid-Filled Transformers

Final Project Report

Project Team

P.K. Sen, Project Leader

Sarunpong Pansuwan
K. Malmedal
Omar Martinoo
Marcelo G. Simoes
Colorado School of Mines

Karen Butler-Purry
Texas A & M University

PSERC Publication 11-02

February 2011
Information about this project

For information about this project contact:

P.K. Sen, Ph.D., P.E.

Colorado School of Mines
Division of Engineering
Golden, CO 80401
Phone: 303-384-2020
Fax: 303-273-3602
Email: psen@mines.edu

Power Systems Engineering Research Center

The Power Systems Engineering Research Center (PSERC) is a multi-university Center

conducting research on challenges facing the electric power industry and educating the
next generation of power engineers. More information about PSERC can be found at the
Center’s website: http://www.pserc.org.

For additional information, contact:

Power Systems Engineering Research Center

Arizona State University
577 Engineering Research Center
Tempe, Arizona 85287-5706
Phone: 480-965-1643
Fax: 480-965-0745

Notice Concerning Copyright Material

PSERC members are given permission to copy without fee all or part of this publication
for internal use if appropriate attribution is given to this document as the source material.
This report is available for downloading from the PSERC website.

 2011 Colorado School of Mines and Texas A & M University. All rights reserved..
 Acknowledgements

This is the final report for the Power Systems Engineering Research Center (PSERC) research
project titled “Transformer Overloading and Assessment of Loss-of-Life for Liquid-Filled Trans-
formers (Project T-25).” We express our appreciation for the support provided by PSERC’s in-
dustrial members and by the National Science Foundation under grant NSF EEC-0001880 re-
ceived under the Industry / University Cooperative Research Center program.

We wish to thank all PSERC members for their technical advice on this project. Special thanks
go to Messrs. Art Mander and Jeff Selman, (Tri-State Generation and Transmission), Simon
Chiang (Pacific Gas and Electric), Bob Wilson and Dan Hamai (Western Area Power Adminis-
tration) and Robert Saint (National Rural Electric Cooperative Association).

i
 Executive Summary

This is the final project report for research on Transformer Overloading and Assessment of Loss-
of-Life for Liquid-Filled Transformers in electric power systems. This subject has been ad-
dressed by many researchers and standards over the years, however, still with no clear consen-
sus. There are several areas where the addition of more advanced sensor and monitoring technol-
ogy can improve the remaining life expectancy estimations. This research developed an optimi-
zation methodology to minimize the cost and select the proper transformer size for new applica-
tions and to optimize the replacement of transformer for an existing system (retrofit applica-
tions). It is anticipated that the method described here will help utilities in making decisions to
minimize revenue requirements of the transformer over the long run to attain overall economic
efficiency.
Although the incentive to loading power transformers beyond their nameplate rating has always
been existed in the past, recently utilities show more inclination to fully utilize them to achieve
greater profit. One of the basic criteria which limit the transformer loading capabilities is the hot-
test-spot temperature of windings. According to the IEEE Std. C57.91-1995, for the thermally
upgraded paper, it is limited to 110C @ 30C ambient temperature for a 65C average winding
temperature rise. Higher winding hottest-spot temperature causes degradation (decrease in me-
chanical strength and increase brittleness) of the winding insulation and increases the potential of
transformer failure. Gas bubbles may also form at elevated operating temperature, which may
also cause the dielectric breakdown.
Under certain operating conditions, a transformer may be safely loaded beyond its nameplate rat-
ing. For every 1C ambient temperature reduction (from standard 30C) releases approximately
1% of overloading capability. The cold winter weather allows transformers for some overloading
or saving of the insulation life. While in the summer, transformers run at higher ambient temper-
atures. The insulation degrades rapidly under these high temperatures and transformer life could
be shortened substantially.
Utilities usually size and operate their transformers by matching the rating with the present de-
mand and taking into consideration the future growth. Industry standard suggests transformer life
expectancy to be approx. 30 years under “normal” operating conditions. In order to defer trans-
former replacement cost or cost of adding a second transformer under certain conditions, utilities
may overload the transformer beyond the nameplate rating and accept calculated reduced life.
This research addressed this very issue of economic decision based on the transformer remaining
life-expectancy model. The probability tree structure is utilized to describe the future load
growth pattern and uncertainty. Together with probability tree model, the transformer thermal
model has been employed to calculate service life of the transformer and determine when to re-
place an existing transformer.

Following the concepts of Per-Unit Life, Relative Aging factor, Equivalent Aging, and end-of-
insulation-life criteria, two simple equations have been developed to estimate the transformer
remaining life. A Windows based, object oriented program has been developed to calculate the
hottest-spot temperature, the top- and the bottom-oil temperature for each model. The program
also calculates the loss-of-insulation-life, the remaining life, and energy losses following the me-
thodology developed in this research.

ii
 Table of Contents

1.  Introduction ............................................................................................................................. 1 

2.  Transformer Design and Thermal Loading............................................................................. 3 
2.1  Introduction ..................................................................................................................... 3 
2.2  Basic Design Considerations .......................................................................................... 3 
2.2.1  Transformer Construction and Its Thermal Performance ........................................... 4 
2.2.2  Transformer Heating ................................................................................................... 8 
2.2.3  Transformer Failure Modes ........................................................................................ 9 
2.3  Thermal Aging Principles and Historical Perspectives .................................................. 9 
2.4  Insulation Deterioration Mechanisms ........................................................................... 11 
2.5  Transformer Life vs. Insulation Life ............................................................................. 11 
2.6  Overloading Limitation ................................................................................................. 14 
2.6.1  Hottest-Spot Limits ................................................................................................... 14 
2.6.2  Top-Oil Limits .......................................................................................................... 14 
2.6.3  Insulation Life ........................................................................................................... 15 
2.6.4  Ancillary Equipment ................................................................................................. 15 
2.6.5  Stray Flux Heating .................................................................................................... 15 
2.6.6  Bubble Generation .................................................................................................... 16 
2.7  Transformer Design Optimization ................................................................................ 16 
2.7.1  Design for Maximum Efficiency .............................................................................. 16 
2.7.2  Design for Minimum Cost of Material ..................................................................... 17 
2.7.3  Design for Minimum Total Owning Cost (TOC) ..................................................... 17 
2.8  Transformer Monitoring and Diagnostics ..................................................................... 17 
2.8.1  On-Line Monitoring .................................................................................................. 18 
2.8.2  Off-Line Diagnostics ................................................................................................ 19 
3.  Transformer Thermal Models, Overloading, and Loss-of-Insulation-Life ........................... 20 
3.1  IEEE “Classical Thermal Model” (Clause 7) ............................................................... 20 
3.2  IEEE Alternate Thermal Model (Annex G) .................................................................. 26 
3.2.1  Duct-oil temperature ................................................................................................. 28 
3.2.2  Hottest-spot temperature ........................................................................................... 29 
3.2.3  Bulk-oil temperature ................................................................................................. 30 
3.3  IEC 354 Thermal Model ............................................................................................... 32 
3.4  Loss-of-Life Calculation, IEEE Method ....................................................................... 33 
3.5  Comparison of the IEEE Thermal Models: Numerical Examples ................................ 34 
4.  Probabilistic Assessment of Transformers Loss-of-Life ...................................................... 38 
4.1  Introduction ................................................................................................................... 38 
4.2  Modeling ....................................................................................................................... 38 
4.2.1  Transformer Loading (K) .......................................................................................... 38 
4.2.2  Ambient Temperature (TA)........................................................................................ 38 
4.2.3  Correlation between loading and ambient temperature () ...................................... 39 
4.3  Monte Carlo simulation ................................................................................................ 39 
4.4  Numerical example ....................................................................................................... 41 
5.  Transformer Economic Evaluation ....................................................................................... 44 
5.1  Loss Evaluation: Conventional Method........................................................................ 44 
5.2  Revenue Requirement and Fixed Charge Rate ............................................................. 44 
iii
 5.2.1  Transformer Loss Evaluation .................................................................................... 45 
5.2.2  Discussion of Factors ................................................................................................ 47 
5.2.3  An Example .............................................................................................................. 49 
5.3  Engineering Economic Evaluation for Investor-Owned Utilities ................................. 50 
5.3.1  Characteristics of Investor-Owned Utilities (IOU’s) ................................................ 50 
5.3.2  Development of the Revenue Requirement Method for Transformer Economic
Evaluation ............................................................................................................................. 51 
5.3.3  Spreadsheet Illustration of the Revenue Requirement Method ................................ 53 
6.  Optimization Strategy for Transformer Procurement and Replacement .............................. 58 
6.1  Introduction ................................................................................................................... 58 
6.2  Transformer Cost and Losses Function ........................................................................ 58 
6.3  Random Failure of Transformer In-Service .................................................................. 61 
6.4  Expected Transformer Life ........................................................................................... 62 
6.5  Load Growth Uncertainty Modeling ............................................................................. 67 
6.6  Solution Methods .......................................................................................................... 72 
7.  Simulations and Case Studies ............................................................................................... 76 
7.1  Introduction ................................................................................................................... 76 
7.2  Solutions to the Problems ............................................................................................. 76 
7.2.1  Case #1, Sizing of a New Transformer with Moderate Load Growth ...................... 77 
7.2.2  Case #2, Transformer Replacement with Moderate Load Growth Rate ................... 81 
7.3  Summary ....................................................................................................................... 83 
8.  Conclusions ........................................................................................................................... 85 
8.1  Contributions of this Research: ..................................................................................... 86 
8.2  Future Work: ................................................................................................................. 87 
Bibliography ................................................................................................................................. 88 
Appendix A: Simplified Transformer Overloading Guidelines................................................... 92 
Appendix B: Delphi 4.0 (Pascal for Windows) Program's Screen Shots .................................... 95 
Appendix C: Comparison Between IEEE Loading Guide C57.91-1995 and IEC Loading Guide
IEC 354-1991 .............................................................................................................................. 103 
 
   

iv
 List of Figures

Figure 2.1 Transformer fluid flow .................................................................................................. 3 

Figure 2.2: Tensile strength reduction .......................................................................................... 12 
Figure 2.3: Degree of Polymerization Reduction ........................................................................ 12 
Figure 2.4 Remaining life by tensile strength method .................................................................. 13 
Figure 2.5 Remaining life by degree of polymerization method .................................................. 13 
Figure 3.1 Transformer’s temperature profile for IEEE classical thermal model ........................ 21 
Figure 3.2: Block diagram of the modified transient heating equations ....................................... 26 
Figure 3.3: Variation of duct oil temperature and top-oil temperature after a load step 0-
100% with an ambient temperature of (-10C) .......................................................... 27 
Figure 3.4: Temperature profiles of IEEE Annex G thermal model subjected to step load ......... 37 
Figure 3.5: Temperature profiles of IEEE Clause 7 thermal model subjected to step load .......... 37 
Figure 4.1: Monte Carlo simulation .............................................................................................. 40 
Figure 4.2: Relationship between ambient temperature and transformer loading with
correlation coefficient of 0.2...................................................................................... 41 
Figure 4.3: 3-D plot of ambient temperature and transformer loading ......................................... 41 
Figure 4.4: Distribution of hottest-spot temperature..................................................................... 42 
Figure 4.5: Distribution of relative aging rate (FAA) ..................................................................... 42 
Figure 4.6: Relative aging rate with different correlation coefficient on various ......................... 43 
Figure 5.1: Total owning cost of different manufactured transformers ........................................ 50 
Figure 5.2: Relationship of revenue requirements and costs for an IOU...................................... 51 
Figure 6.1: Transformer cost data ................................................................................................. 60 
Figure 6.2: Transformer losses data .............................................................................................. 61 
Figure 6.3: “Bathtub curve” .......................................................................................................... 61 
Figure 6.4: Load and ambient temperature profiles in the month of August ................................ 63 
Figure 6.5: Loss of transformer life vs. years in service ............................................................... 65 
Figure 6.6: Hottest-spot and top-oil temperature vs. years in service .......................................... 66 
Figure 6.7: Insulation’s remaining tensile strength vs. years in service ....................................... 66 
Figure 6.8: Insulation’s remain degree of polymerization vs. years in service ............................ 67 
Figure 6.9: Approximating load uncertainty using probability tree ............................................. 68 
Figure 6.10: Probability tree representation of load uncertainty .................................................. 69 
Figure 6.11: Illustration of load growth path ................................................................................ 70 
Figure 6.12: Cumulative probability of transformer load ............................................................. 71 
Figure 6.13: Approximating load uncertainty in each future year ................................................ 71 
Figure 6.14: Integrated structure of insulation degradation based transformer utilization
mode .......................................................................................................................... 75 
Figure 7.1: Hottest-spot and top-oil temperature of case #1’s 25 MVA transformer (Path
#41) ............................................................................................................................ 81 
Figure A.1: The thermal and electrical limits for various types of loading .................................. 93 
Figure B.1: Transformer Type and Cooling ................................................................................ 95 
Figure B.2: IEEE Annex G Input Data File ................................................................................. 95 
Figure B.3: Daily Load and Temperature Profile ........................................................................ 96 
Figure B.4: Monthly Load and Temperature Profile ................................................................... 96 
Figure B.5: Annual Load Growth Data........................................................................................ 97 
Figure B.6 (a) (b): IEEE Annex G Run Data and Plot................................................................. 98 
v
Figure B.7 (a) (b): IEEE Clause 7 Run Data and Profile ............................................................. 99 
Figure B.8: Life Cycle Analysis Calculated Data ...................................................................... 100 
Figure B.9: Life Cycle Study: Composite Chart ....................................................................... 100 
Figure B.10: Utility Financial Data and Energy Cost Input Data .............................................. 101 
Figure B.11: Optimum Transformer Sizing Output Data for New Procurement ...................... 101 
Figure B.12: Transformer Replacement Strategy Data.............................................................. 102 

vi
 List of Tables

Table 2.1: Range and average value of current density and flux density in various types of
oil-filled transformer with copper winding.................................................................. 6 
Table 2.2: Aging rate constant, B................................................................................................. 11 
Table 2.3: Insulation Life Definitions .......................................................................................... 14 
Table 3.1: Thermal characteristics of classical thermal model at rated load ................................ 22 
Table 3.2: Summary of exponents ................................................................................................ 31 
Table 3.3: Normal insulation life per IEEE C57.91-1995 ............................................................ 34 
Table 3.4: Required data for IEEE thermal models ...................................................................... 34 
Table 3.5: Transformer data for temp. rise comparison at steady-state, IEEE methods............... 35 
Table 3.6: Comparison of top-oil and hottest-spot temperature of IEEE thermal models
(constant load, 30C ambient temperature) ................................................................ 36 
Table 5.1: Loss Evaluation Calculation Example ......................................................................... 49 
Table 5.2: Spreadsheet of annual revenue requirement ................................................................ 56 
Table 5.3: Spreadsheet of annual revenue requirement with early replacement .......................... 57 
Table 6.1: Cost of typical forced-air cooled (OA/FA/FA) power transformer ............................. 59 
Table 6.2: The multiplying factor for MVA rating conversion from 55C winding rise to
65C winding rise rating ............................................................................................ 59 
Table 6.3: Transformer losses ....................................................................................................... 60 
Table 6.4: Illustrated life cycle study printout of transformer life ................................................ 64 
Table 7.1: Common transformer, economic, and financial data for all case studies .................... 77 
Table 7.2: Result of Case #1 Study .............................................................................................. 78 
Table 7.3: Result of a 25 MVA transformer on load growth path #41, case #1 study ................. 79 
Table 7.4: Life cycle study result of case #1’s 25 MVA transformer (path #41) ......................... 80 
Table 7.5: Result of case #2 study ................................................................................................ 82 
Table 7.6: Detailed result of delay replacement with 0.8 per-unit peak load criterion on load
growth path #41, case #2 study.................................................................................. 83 
Table 7.7: Life cycle study result case #2’s existing 18 MVA transformer (path #41) ................ 84 
Table A.1: Thermal and electrical limits for various types of loading ......................................... 93 

vii
 NOMENCLATURE

Symbols Related to Thermal Models and Loss- of- Life Calculation

C Transformer thermal capacity, watt - min./  C
CPW Specific heat of winding material, watt - min./lb.  C
TBO Bottom-oil temperature rise over ambient
TG Winding hottest-spot temperature rise over top-oil temperature
TGR Winding hottest-spot temperature rise over top-oil temperature at rated load
TG,i Initial winding hottest-spot temperature rise over top-oil temperature
TG,u Ultimate winding hottest-spot temperature rise over top-oil temperature
THS Incremental hottest-spot temperature difference between iteration
THS/WO Winding hottest-spot rise over adjacent duct-oil temp. at hottest-spot location.
TIMR Average oil rise (IEC)
TOAVG Incremental average bulk-oil temperature between iteration
TTDO/BO Duct-oil temperature rise over bottom-oil temperature

TTDOR/BOR Duct-oil temperature rise over bottom-oil temperature at rated load

TTO Top-oil rise over the ambient temperature
TTOR Top-oil temperature rise at rated load
TTO,i Initial top-oil temperature rise
TTO,u Ultimate top-oil temperature rise
TTO/BO Top- and bottom-oil temperature difference
TW Incremental winding temperature difference between iteration
TWO Duct-oil temperature rise at winding hottest-spot location over bottom oil
TWR Average winding rise temperature at rated load
FAA Relative aging factor
FEQU Equivalent aging factor
H Hottest-spot factor
HHS Location of hottest temperature on the winding, per unit of winding height
k Radiation constant, watt/  C
K Transformer loading, per-unit

viii
L Insulation life, per unit
HS Oil viscosity at hottest-spot location
HSR Oil viscosity at hottest-spot location at rated load
W Oil viscosity in the duct
WR Oil viscosity in the duct at rated load
m Winding exponent
MW Winding mass, lb
n Oil exponent
PEHSR Eddy current loss at rated load and at hottest-spot temperature
PER Eddy current loss at rated load
PHSR Winding loss at rated load and at hottest-spot temperature
Ploss Total losses, watt
Ploss,i Initial total losses, watt
Ploss,u Ultimate total losses, watt
PR Total losses in watts at rated load
PT Transformer total losses, per-unit
PTR Total transformer losses at rated load
PW Winding loss, watt
PWHSR Winding loss at hottest-spot temperature
PWR Winding loss at rated load
QCore Heat generated from core loss
QHSABS Heat absorbed in winding at hottest-spot location
QHSGEN Heat generated in winding at hottest-spot location
QHSLOST Heat lost from winding hottest-spot location to oil
QOGEN Heat generated in oil
QOLOST Heat lost from oil to air
QStray Heat generated from stray loss
QWABS Heat absorbed in winding
QWGEN Heat generated in winding
QWLOST Heat lost from winding to oil
R Transformer loss ratio (load losses/no-load losses)

ix
MCP Summation of product of mass and specific heat of tank, core, and oil excluding winding
G Winding time constant

 GR Winding time constant at rated load

 TO Oil time constant

 TOR Oil time constant at rated load

TA Ambient temperature
TBO Bottom-oil temperature
TBOR Bottom-oil temperature at rated load
TDAO Average duct-oil temperature
TDAOR Average duct-oil temperature at rated load
THS Hottest-spot temperature
THSR Hottest-spot temperature at rated load
Tk Resistance correction factor
TKHS Winding resistance correction factor for hottest-spot temperature
TKW Winding resistance correction factor
TOAVG Average bulk-oil temperature
TOAVGR Average bulk-oil temperature at rated load
TTDO Top duct-oil temperature
TTDOR Top duct-oil temperature at rated load
TTO Top-oil temperature
TTOR Top-oil temperature at rated load
TW Average winding temperature
TWR Average winding temperature at rated load
x Exponent of duct-oil rise
y Exponent of average oil rise
z Exponent of bulk-oil temperature difference

x
Symbols Related to Economic Evaluation
 Load growth rate
AEC Annualized energy cost
ALCR Auxiliary losses cost rate, $/kW
ALL Auxiliary losses
CTR Cost of transformer
CC Carrying charge
CF Cost of random failure
CLL Cost of load losses
CNLL Cost of no-load losses
COL Cost of losses
CRF Capital recovery factor
CRR Capitalized revenue requirement
DB Depreciation
DT Tax depreciation for income tax purposes
DC Demand charge, $/kW.yr
e Energy cost escalation rate
e Average annual inflation rate
EA Annual equity return
EPR Equivalent peak ratio
ERR Equivalent revenue requirement
f(t) Probability distribution function
F(t) Cumulative distribution function
f(ST) Transformer cost function
fLL(ST) Transformer load losses function
fNLL(ST) Transformer no-load losses function
FCR Fixed charge rate
h(t) Hazard function
i Cost of capital rate or discount rate
ib Borrowed money rate

id Depreciation rate

xi
ie Equity return rate

it Tax rate
 Debt ratio, failure rate
LL Load losses
LLCR Load losses cost rate, $/kW
MV Market value of transformer
N Transformer life
NLCR No-load losses cost rate, $/kW
NLL No-load losses
PA Annual production expense
Pcore Core or no-load losses, kW
Pcu(t) Time-varying load losses, kW
Pcu,max Annual maximum load losses, kW
PA Probability that the auxiliary cooling will be on
PR Peak ratio or per-unit loading
PRF Peak responsibility factor
RA Annual revenue
RR Revenue requirement
ST Transformer maximum rating
SI System investment cost, $/kW
T Income tax
TLF Transformer loss factor
TOC Total owing cost
UI Unrecovered investment

xii
1. Introduction
Sizing of a large new power transformer or replacement of an existing transformer is
done traditionally from simplified analysis and technical considerations. The convention-
al loss evaluation technique, defined by the “Total Owning Cost (TOC)”, is still routinely
utilized by utilities to evaluate the values of transformer losses during procurement. Total
cost of losses during transformer operating life is comparable to its initial purchase price,
and the loss evaluation is always recommended during procurement. The TOC method,
however, has limitations. Regardless of the load cycles, ambient conditions, and future
load growth, transformer’s life (typically 30 yrs.) is assumed to be constant. Load cycles,
ambient conditions, and future load growth and possible overloading including uncertain-
ties are very important factors that affect transformer’s life, hence the total cost. Also, a
decision has to be made, whether to replace the transformer immediately or to delay its
replacement. Utility’s engineers should evaluate the remaining life of the existing trans-
former due to overloading, together with economic evaluation.

This research utilizes a simplified optimization strategy and an improved method for new
transformer sizing, cost evaluation and perhaps delay replacement analysis of existing
transformers. The method is based on the loss-of-life information from the hottest-spot
temperature calculated from transformer thermal model.

Most transformer failures are related to the deterioration of insulation with. For liquid-
filled transformers, the traditional insulation system is thermally upgraded oil-
impregnated paper. The concept of insulation integrity has led to the development of the
thermal insulation aging that has been known to be a function of both time and tempera-
ture (Arrhenius Reaction Rate Theory). However, transformer loss-of-life at various ele-
vated-operating temperatures cannot be accurately estimated and the corresponding re-
maining life expectancy is considered to be conservative.

This final report is comprised of 8 chapters followed by three Appendices. Chapter 2 dis-
cusses the design fundamentals of liquid-filled transformers. Various “On-line/Off-line”
monitoring techniques are also discussed.

In Chapter 3, two (2) thermal models from the IEEE loading guide, C57.91-1995 [1] are
discussed. A third model from the IEC loading guide, IEC 354-1991 [2] is also used to
compare. A comparison of the IEEE and IEC models is attached in the Appendix C. A
PC based computer program on Windows operating system is written to calculate all
temperature profiles and transformer loss-of-life including the Graphic User Interface
(GUI) helps user to easily access data and perform analysis.

Chapter 4 discusses the probabilistic modeling including the Monte Carlo simulation to
calculate the loss-of-life.

Chapter 5 has been dedicated to the development and discussion of economic evaluation,
Conventional loss evaluation, characterized by the “Total Owning Cost (TOC)”. A more
detailed technique called the Minimum Revenue Requirement method, which is applica-
ble to Investor-Owned Utilities (IOU’s) is also introduced.

1
Chapter 6 proposes the optimization scheme that will determine the size of a new trans-
former and provide a strategy for delaying replacement of an existing transformer based
on load cycle, present load, future growth, ambient operating conditions, and economic
consideration. Transformer cost and losses are derived by curve fitting data obtained
from various industry sources. Failure cost is also included in this scheme. The probabili-
ty tree structure is applied to future load growth that takes into account the uncertainty. A
basic computer program is written to estimate the transformer life, energy losses, and fi-
nancial results.

Chapter 7 includes the numerical calculations. Different case studies are discussed in-
cluding new transformer sizing and “delay replacement” strategy.

Chapter 8 includes conclusions and recommendations.

In summary, this research provided an integrated method for transformer sizing and pro-
vides an optimal solution for transformer replacement. Windows based computer pro-
gram is written to provide system planning engineers with fast, convenient, and practical
solutions.

2
2. Transformer Design and Thermal Loading
2.1 Introduction

Emergency and/or planned overloading of oil-filled power transformers beyond their na-
meplate rating depends on several factors including the design, operation, routine oil test-
ing and maintenance program, daily loading and load cycle, ambient conditions and ap-
plications. For most common applications, transformer overloading capabilities and the
life expectancy are determined by the winding “hottest-spot” temperature and its dura-
tion.

For all liquid-filled transformers, the insulation system is composed of thermally-

upgraded oil-impregnated paper. This cellulose paper insulation used today must main-
tain its mechanical strength and withstand the stresses that occur with surges and must be
able to withstand detrimental chemical transformations.

2.2 Basic Design Considerations

Over time, oil-impregnated paper insulation used in liquid-filled transformer winding

loses mechanical and electrical strength and becomes brittle when exposed to elevated
operating temperatures. These results are obtained from aging tests of transformers or are
obtained from laboratory tests on isolated pieces of insulation material.

A typical simplified transformer heating and cooling model used for the analysis is
illustrated in Figure 2.1. The winding I2R and eddy-current losses, the core losses, and the
stray losses in the tank and metal support structures are the principal sources of heat.
There exists a significant difference between the top- and bottom-oil temperature rises,
and this will vary depending on the type of cooling systems and winding construction.
The difference between the top and bottom oil rises with forced oil-cooling is in the order
of only a few degrees, whereas this difference is several times larger for forced air-cooled
transformers.

Figure 2.1 Transformer fluid flow

3
Four modes of cooling most commonly employed in liquid filled transformers are: (1)
Natural convection of oil and natural convection of cooling air over the radiator (OA); (2)
Natural convection of oil with forced convection of air over the radiator – either single
stage (OA/FA) or multiple (most common, two) stages (OA/FA/FA); (3) Directed forced
oil flow and forced airflow (DFOA); and (4) Non-directed forced oil flow; and forced
airflow (NDFOA).

OA cooled transformers are used in the distribution system and the ratings are typically
limited to less than 2.5 MVA. Transformers rated between 2.5-100 MVA are designed for
“Fan Cooling” (OA/FA/FA). A single-stage “OA/FA cooling” is used for the lower rat-
ings (2.5-10MVA range). Larger distribution transformers (10-100MVA) are normally
designed to have two stages of fan cooling (OA/FA/FA). Transformers rated above 100
MVA ranges are generally “FOA type” (or forced oil and forced air cooled design).

Another fact to be remembered is that the expected transformer life (common practice) is
estimated between 25-30 years, depending on the utility’s accounting preference. How-
ever, transformer operating for 40 years (or more) of service life is not uncommon in the
industry. This “normal life period” assumes some overloads but also includes long pe-
riods of light loads and lower ambient temperatures.

2.2.1 Transformer Construction and Its Thermal Performance

Heat can flow from the core to the oil. However from the winding, heat must go through
the insulation. Large transformers are designed so that at least one side of each insulated
coil can transfer heat directly to the oil. The heat transfer rate is proportional to the insu-
lation thermal conductivity and exposed surface area and inversely proportional to the
insulation thickness [3].

The core and windings define the basic dimensions of the transformer tank's length and
width, with the tank height determined by the level of oil necessary to cover the core (in-
cluding tap changer). Additional space for oil circulation is added on to the basic dimen-
sions. Tank design affects the ability of the transformer to dissipate heat to the surround-
ings. Vertical location of the core and winding within the tank also will influence the rate
of heat transfer to the oil. At elevated temperature, the transformer oil is oxidized and
forms sludge that reduces the heat flow from the winding, thereby, elevating its tempera-
ture. The use of inert gas to minimize sludge formation, combined with oil filtration, con-
trols the effects of oil oxidation.

Core - Magnetic cores for large power transformers are made of thin laminations of
grain-oriented-silicon-steel. Currently five main grades of silicon steel are widely used in
the transformer industry: M2, M3, M4, M5, and M6 in increasing order of gauge (from 7
to 14 mil thickness) and core losses. Besides the type of material and lamination thick-
ness, the operating flux density in the core determines core losses. Typically the design
value of the flux density in silicon steel core ranges from 1.6-1.8T (Tesla = Wb/m2). The
core flux density, core cross-sectional area, voltage, frequency and number of turns are
related by the basic equation:

4
 E  4 . 44 fNB m A ( 2.1 )

Where, E is the exciting [rms] voltage (V) (assumed sinusoidal voltage), f is the frequen-
cy (Hz), N is the number of turns, Bm is the maximum flux density in (T), and A is the
cross-sectional area of steel core (m2). When the frequency and the area of the core are
constant, the maximum flux density is proportional to ratio of exciting voltage and turns
(or volts per turn ratio, E/N).

Typically, under no-load condition and at rated voltage, transformers draw very little ex-
citing current (rms value of 2-4% of the full load current). Because of the nonlinear mag-
netization characteristic of the steel core, and since the flux-density in a typical design is
above the “knee point”, the exciting current contains harmonics. A typical exciting cur-
rent consists of (45%) 3rd, (15%) 5th and (3%) 7th and smaller percentages of higher fre-
quency harmonics [4]. In current design practices, the limit on the value of flux density is
imposed by the amount of distortion in the exciting current and the corresponding genera-
tion of audio noise. It is considered a good practice to keep the flux density at a lower
value of approximately 1.7T.

Coil and Winding - Transformer coils are designed to get the required number of
turns into a minimal space. At the same time, the conductor cross-section must be large
enough to be able to carry the current without overheating. These coils may be made of
copper or aluminum. Aluminum winding is generally cheaper and has lighter weight
(30% of copper’s weight). However, the aluminum winding has higher losses (62% high-
er resistance). At present, the typical payback period for Cu is estimated between 5 to 8
years.

The transformer size and the overall design are basically dictated by two main factors
called “Specific Magnetic Loading” (flux density “B” in the core) and “Specific Electric
Loading” (current density in the winding). Transformer size is related to load losses
(winding), maximum magnetic flux density and current density through an empirical eq-
uation of the form [5]

 A   % Pcu 
J  K f Bm     ( 2.2 )
 s   K eddy 


Where, J is the current density in conductor (A/mm2) , Bm is the maximum flux density
(T), A is the cross-section area of core (m2), s is the mean turn length (m), % Pcu is in the
core loss expressed as a percent of transformer rating, and Keddy = 1.05-1.2. The constant
K depends upon the unit and conductor material. Other constraints such as lowest unit
cost, minimum total owning cost, guaranteed load loss, etc., are also utilized to optimize
the design.

5
Typical current densities in a design may vary by a factor of 2:1. It depends primarily on
the winding material (copper vs. aluminum) and the cooling method. In “Class B” insu-
lated transformers (most common design with 65C avg. winding rise), the maximum
value of current density varies from about 3.5 A/mm2 smaller to 6.0 A/mm2 for large
transformers with forced cooling. Table 2.1 [5] shows the range and the average value of
current and flux densities.

Table 2.1: Range and average value of current density and flux density in various types
of oil-filled transformer with copper winding

Current density (A/mm2) Flux density

(T)
Class of transformer Range Average

Generator Step-up 2.6-4.0 2.9 1.70

Transmission 3.7-6.0 4.3 1.55

Distribution 2.0-2.5 2.3 1.55

Cooling Equipment - The radiator cooling now used have a fairly constant heat dis-
sipation rate per unit length. Cooling tubes that are farther spaced and the larger the tube
surface area, the greater is the cooling capability. Some manufacturers use a flat plate de-
sign for cooling tubes, allowing more surface area per tube and reducing manufacturing
costs. Pumps, when utilized, are used to increase the flow of oil, thereby increasing the
efficiency of the radiators and minimizing the temperature difference between the top-oil
and at the bottom-oil in the tank.

Cooling Mode - Natural draft (air) cooling (AA) is utilized for small transformers.
However, as the transformer size increases, the cooling surface area is insufficient. Addi-
tional cooling must be provided. Oil immersion increases the heat transfer rate and the
addition of external radiators attached to the tank increases the cooling surface area (OA).

Forcing air can substantially increase the rate of cooling above the self-cooled rating.
Larger MVA transformers may be designed for either one or two stages of forced air-
cooling (OA/FA or OA/FA/FA). Forced-Oil-Air cooling (FOA) employs pumps to draw
the oil out of the transformer tank to the external heat exchanger. Increasing of oil veloci-
ty also increases cooling efficiency and it reduces the top-oil temperature rise over bot-
tom oil. There are two types of forced oil cooling, non-directed flow (NDFOA) and di-
rected flow (DFOA). In non-directed flow transformers, the pumped oil from heat ex-
changers or radiators flows freely inside the tank. Directed flow transformers are de-
signed so that the principal part of the pumped oil from the heat exchangers or radiators is

6
forced to flow through the windings. Water-cooling can be used for large transformers
when economically justified.

There exist some relationships between the air-cooled rating (OA) and the corresponding
increase in rating due to the added cooling systems. For 3-phase transformers, assuming
the average ambient temperature of 30C, the multipliers are given below [6]. A 10 MVA
(OA) transformer will have a rating of 10 x 1.25 = 12.5 MVA (OA/FA) rating at 55C.

Transformer Size:

*  2.5 MVA, (FA) = 1.15 (OA), one stage @ 55 o C

* < 12 MVA, (FA) = 1.25 (OA), one stage @ 55 o C

*  12 MVA, (FA) = 1.33 (OA), one stage @ 55 o C

*  12 MVA, (FOA) = 1.67 (OA), one stage @ 55 o C
*  12 MVA, (FOA) = 1.87 (OA), one stage @ 65 o C

Oil Preservation Systems - During operation under heavy load, the transformer oil
level will rise above the initial fill level. All oil-filled power transformer designs have
some means of providing space for oil expansion. The most common designs employ the
conservator tank or provide gas space (filled with inert gas) in the main transformer tank
above the oil.

Auxiliary Cooling Equipment - A thermally operated control device or a manually

operable switch can be used for control of auxiliary cooling equipment. The thermally
operated control device, measuring the top-liquid temperature, is used in an automatic
control system.

Auxiliary Devices - Transformers are also provided, in general, with the following
auxiliary devices (common for 10MVA and larger), when requested during the procure-
ment. Most of these devices are wired-up to the control cabinet for metering, monitoring,
protection, and/or SCADA system application.

 Load Tap Changer [LTC] ( 16 x 5 8 %)

 No-Load (or Off-load) Tap Changer [NLTC] ( 2 x 2½%)
 Bushing, CT’s (1,2,3 per bushing)
 Top-Oil Thermometer
 Hottest-Spot Thermometer
 Oil-Level Indicator
 Oil-Flow Indicator
 Pressure-Vacuum Gauges
 Pressure Relay and Pressure-Relief Devices

7
2.2.2 Transformer Heating

Transformer heating is caused primarily by the core losses (no-load losses), winding
(load) losses, and stray-load losses.

No-load Losses (Hysteresis and Eddy Current) - The no-load (or core) losses are
comprised of primarily two components: Hysteresis and Eddy Current.

 Hysteresis Loss: The hysteresis loss happens due to the hysteretic nature of ma-
terial. The hysteresis loss is given by the empirical formulae (Steinmetz),

Ph  k h fB mx ( Volume) ( 2.3 )

where: kh = Material constant

f = Frequency (Hz)
Bm = Maximum flux density (T =Wb/m2)
x = Exponent varies between 1.6 – 2.8 (or sometimes higher), and
(Volume) = Volume of the magnetic material

 Eddy Current Loss: The alternating magnetic field in a transformer core causes
current to flow in the core [eddy currents] and produce heat loss. This component of the
core loss can be reduced by laminating. The eddy-current loss is proportional to the (la-
mination thickness)2 and the loss can be expressed by the simplified equation:

Pe  k e f 2 t 2 B m2 (Volume) ( 2.4 )

where: ke = Material constant

f = Frequency (Hz)
Bm = Maximum flux density (T =Wb/m2)
t = Lamination thickness (mm), and
(Volume) = Volume of the magnetic material

If the transformer is over-excited, core loss increases because of the increased maximum
flux density (Bm) and the non-linearities of the magnetization characteristic. However, for
most common applications, the core loss is assumed to be constant.

Load (Winding) Losses - Load losses in transformers consist of two primary com-
ponents: (1) winding loss, due to the copper (or aluminum) winding resistance, and (2)
stray load loss due to the eddy currents induced in other structural parts of the transfor-
mer. The winding loss has two components: resistance loss, and winding eddy-current
loss. In all cases the predominant component of the losses is proportional to the (current)2
or (loading)2.

The guaranteed load loss from manufacturers is specified at 85C [7]. Unlike the
IEEE/ANSI loading guide that uses 30C average ambient temperature, all transformer

8
losses are tested at 20C ambient temperature. With 65C winding rise, this will yield the
85C test temperature. To evaluate this loss at any other temperature, the following cor-
rections are required:

 T  Tr 
PCU @ Tr  PCU @ 85C  k  ( 2.5 )
 Tk  85 

T  85 
Peddy@Tr  Peddy @ 85C  k  ( 2.6 )
 Tk  Tr 

T  85
Pstray@Tr  Pstray@85C  k  ( 2.7 )
 Tk  Tr 

Where, Tk is 234.5 for copper and 225 for aluminum. Tr is the new temperature.

2.2.3 Transformer Failure Modes

There are basically two major failure modes. They are:

Long-Term Failure - The dielectric strength of conductor insulation deteriorates

slowly under normal loading. The three major contributing factors are: moisture content,
amount of oxygen and heat cycle. Mechanical properties, such as, “Retained Tensile
Strength (RTS)” and/or retained “Degree of Polymerization (DP)” are the most common
criteria used to measure insulation integrity. Accelerated by heat, both the DP and RTS
reduce over time. However, the “exact” end-of-life is unknown.

Short-Term Failure - This is attributed to “bubble formation” in the oil. The dielec-
tric strength of the conductor insulation reduces drastically when bubble is generated.
Three mechanisms are currently recognized for bubble formation: super-saturation of the
oil, thermal decomposition of cellulose and vaporization of absorbed moisture. The ther-
mal decomposition of cellulose during sudden large increase in load has been identified
as the main cause.

2.3 Thermal Aging Principles and Historical Perspectives

Since the beginning of the use of power transformers, the conductor insulation has been
made of some form of paper or cloth. The main constituent of these materials is cellulose,
an organic compound molecule made up of a long chain of glucose rings or monomers,
typically ranging from 1400 to 1600 for new material. It has been established that, the
mechanical strength of the fiber is closely related to the length of the chains (Degree of
Polymerization, DP).

9
In 1930, Montsinger [7] did an insulation aging experiment by placing varnished cambric
tape insulation into a series of oil-filled test tubes. The test tubes were heated and the in-
sulation’s tensile strength was measured. It was reported that the rate of deterioration of
the tensile strength is doubled for each 5-10C (approximately) increases in [continuous]
operating temperature. The doubling factor was not a constant: about 6C in the tempera-
ture range from 100-110C and about 8C for temperature above 120C. However, a
doubling factor as a constant of 8C has been widely [and popularly] used in the industry.
The present IEC Loading Guide uses a constant value of 6C. Based on this research, the
RTS of 50% of its initial value was also introduced (and later adopted by the IEEE) as
the “end-of-life” criteria for insulation.

In 1948, Dakin [8] also made a significant advancement in defining insulation aging rate
following a modification of Arrhenius’ Chemical Reaction Rate Theory. According to
this theory, the rate of change of a measured property can be expressed in the form of a
reaction-rate constant R and can be expressed by:

R  A' e  B / T ( 2.8 )

Where, A’ and B are empirical constants

T is the temperature in Kelvin.

Dakin showed that all aging rate data (including Montsinger’s) could fit into the Arrhe-
nius Reaction Rate equation. This was later accepted widely by the technical community
and become the foundation for determining the loss-of-insulation-life. From the reaction
rate equation, the insulation life is now defined by:

L  A eB /T ( 2.9 )

Where, L is insulation life in either per-unit (or hours)

A is a constant, derived from the insulation life at 110C hottest-spot temp.
B is the same aging rate slope defined in equation [7]

Many investigators have measured cellulose aging rates under controlled conditions and
have presented their results. Some measured the RTS and others measured the DP or gas
evolution rates. Investigators found and confirmed agreements between changes in RTS
and DP. It was decided to select a single rate slope, the constant B, which is reasonably
accurate for all forms of cellulose. Table 2.2[1] summarizes the results of such published
literature. Placing the emphasis on the more modern data, a value of B of 15,000 is now
used in the transformer insulation life curve utilized in the recent IEEE loading guide
(C57.91-1995).

10
 Table 2.2: Aging rate constant, B

Source, Year Basis B

Dakin, 1947 20% tensile strength retention 18,000
Sumner, 1953 20% tensile strength retention 18,000
Lawson, 1977 10% tensile strength retention 15,500
Lawson, 1977 10% DP 11,350
Head, 1979 Mechanical/DP/gas evolution 15,250
Shroff, 1985 250 DP 14,580
Goto, 1990 Gas evolution 14,300
ANSI-C57.92-1981 50% tensile strength retention 16,054
ANSI-C57.91-1981 Distribution transformer life test 14,594

2.4 Insulation Deterioration Mechanisms

As mentioned on numerous occasions three mechanisms contribute to cellulose deteriora-

tion in operating transformers, namely hydrolysis, oxidation, and pyrolysis. The agents
responsible are water, oxygen, and heat, respectively.

Hydrolysis (Water) - The oxygen bridge between glucose rings is affected by wa-
ter, causing the rupture of the chains, and reduction of DP and weakening of fiber.

Oxidation (Oxygen) - Oxygen attacks the carbon atoms in the cellulose molecule to
form aldehydes and acids, releasing water, CO, and CO2. Since oxidation releases water,
it helps accelerate the hydrolysis mechanism and the insulation deterioration. Oxygen is
derived from either the atmosphere, or from the thermal degradation of cellulose. The
problem is worsened by the presence of catalysts and accelerators like moisture and cop-
per.

Pyrolysis (Heat) - Heat and the resulting high temperature will contribute to the
breakdown of individual monomers in the cellulose chain. Thermal degradation of the
cellulose also yields free water, as well as certain gases like carbon monoxide (CO) and
carbon dioxide (CO2). High temperature within a power transformer can cause the cellu-
lose insulation to shrink and become brittle. This leaves the solid insulation susceptible to
failure due to mechanical stress.

2.5 Transformer Life vs. Insulation Life

If an end-point of insulation life is to be defined, it must be done in terms of some mea-

surable physical characteristic properties. This could be mechanical (RTS), chemical
(DP) or electrical (dielectric strength) properties. Insulation dielectric strength is found to
deteriorate slowly if insulation is not mechanically disturbed and bubbles are not present.
Initially a mechanical property “RTS” was chosen. Later, the “DP” is also accepted as
another popular alternative. A number of end-of-life criteria have been suggested in the
literature, namely 50% [20% suggested by others] RTS, and 200 DP. The DP of 200
which is equivalent to 20% retained tensile strength seems to be the most preferable. The
direct measurement, when possible, of the RTS or DP on paper sample retrieved from
11
transformer is the accurate method. However, removal of paper insulation is expensive
and in many cases impractical.

Reference [10] published the results of RTS and the DP of thermally upgraded paper aged
in a sealed tube at 160C. These results are plotted in Figure 2.2 and Figure 2.3 and are
utilized in this research for estimating the remaining life expectancy.

Figure 2.2: Tensile strength reduction Figure 2.3: Degree of Polymerization Reduction

The 20% RTS and the DP of 200 is used as end-of-life criteria. The time (t) is in per-unit
life. The data were fit to the exponential curve by the least square method. The tensile
strength curve then can be written as [11]:

Retained Tensile Strength (RTS)  97.05e 1.58t ( 2.10 )

For degree of polymerization, the equation is given by:

Retained DP (DP)  622 e 1.135 t ( 2.11 )

The remaining life in per-unit (1-t) can be derived from the following equations:

Remaining Life  1  0.633 ln 

RTS 
 ( 2.12 )
 97.05 

12
 Remaining Life  1  0.881 ln 
DP 
 ( 2.13 )
 622 

Where, RTS is the remaining tensile strength, and

DP is the remaining degree of polymerization

Equation ( 2.12 ) and Error! Reference source not found. are plotted in Figure 2.4 and
Figure 2.5. Corresponding to % RTS of 20 or DP of 200, the remaining life is zero. If the
transformer is functional beyond these criteria, the transformer has exceeded its expected
life and this is denoted by the negative sign.

120 700

100 600

Degree of Polymerization
500
80
% Tensile Strength

400
60
300
40
200
20
100

0 0

Remaining Life (per unit) Remaining Life (per unit)

Figure 2.4 Remaining life by tensile Figure 2.5 Remaining life by degree of
strength method polymerization method

The ANSI/IEEE Std. C57.92[12] offers the end-of-insulation-life criteria of 65,020 hours
for 50% RTS of insulation at 110C hottest-spot temperature. This is based on low oxy-
gen and a moisture level of 0.5%. Table 2.3[12] shows the normal insulation life at vari-
ous levels of moisture and oxygen. The higher water content reduces (by a factor of 2 for
doubling the water content) the normal insulation life as:

Normal Life @[ 0.5% H 2 O ]

Normal Life  ( 2.14 )
2  [% H 2 O ]

13
 Table 2.3: Insulation Life Definitions

Water Oxygen Life

Basis Content Level Hours
50% Tensile 0.5% Low 65,020*
1.0% Low 32,510
2.0% Low 16,255
0.5% High 26,000
1.0% High 13,000
2.0% High 6,500
20% Tensile 0.5% Low 152,000
1.0% Low 76,000
2.0% Low 38,000
0.5% High 60,800
1.0% High 30,400
2.0% High 15,200
200 DP 0.5% Low 158,000
1.0% Low 79,000
2.0% Low 39,000
0.5% High 63,200
1.0% High 31,600
2.0% High 15,800

* ANSI/IEEE Std. C57.92’s end-of-life criteria

2.6 Overloading Limitation

Although transformers are overloaded, there are some limits.

2.6.1 Hottest-Spot Limits

The winding hottest-spot temperature at the top of the high or low voltage winding is the
most critical parameter. It determines the loss-of-life and indicates the potential risk of
releasing gas bubbles on a severe sudden overload condition.

If loss-of-life (of the solid insulation) is not tracked closely, the recent IEEE loading
guide [1] suggests a maximum continuous hottest-spot winding temperature limit of 140oC
(with some loss-of-life), which is the limiting temperature for long-term emergency load-
ing. During short-term emergency situations, hottest-spot temperature is allowed to ex-
ceed 140oC.

2.6.2 Top-Oil Limits

Due to convection and nature of cooling system design, the highest oil temperature in the
transformer tank will be at the top-oil region. When the top-oil temperature exceeds

14
105°C, it is possible for oil to expand beyond the tank capacity and causes the pressure
relief device to operate. Upon cooling, the reduced volume of oil may expose electrical
parts, including the bushing and the winding. Higher top-oil temperatures approaching
flash-point value of 145°C[13,14] pose a much greater danger of sudden ignition and explo-
sion. IEEE recommends that the top-oil temperature under any overloading should not
exceed 110C.

2.6.3 Insulation Life

Insulation loss-of-life of power transformers is closely related to a time function of tem-

perature, moisture, and oxygen content. From these parameters, the most significant de-
termining factor to insulation deterioration is the temperature reached by the hottest-spot
in the winding.

2.6.4 Ancillary Equipment

Overloading the transformer can have significant detrimental effects on associated

equipment. The bushings, tap-changers, bushing-type current transformers (BCT’s) and
leads may also be affected by the increased temperature.

Bushings are designed for a hottest-spot temperature of 105°C for a normal top-oil
temperature limit of 95°C. Operating the bushing above these limits can have damaging
effects such as internal pressure buildup, aging of gasket material, bubble formation when
the hottest-spot temperature exceeds 140°C. For bushings, the following guidelines[6] are
recommended:
 Transformer top-oil temp 110°C maximum
 Maximum [continuous] current 2 x rated bushing current
 Bushing insulation hottest-spot temp 150°C maximum

Tap changers, whether designed to change taps under load (LTC’s) or de-energized
conditions, are subjected to carbon build-up at elevated temperatures. Transformers are
normally designed so that the LTC rating is greater than the transformer rating. It has
been seen from practice that, more frequent maintenance is required on LTC’s, which are
subjected to operation at elevated temperatures compared to transformers running at low-
er temperature.

Bushing-type current transformers (BCT’s) have the transformer top-oil as their

ambient temperature. Overloading the transformer will result not only in higher top-oil
temperature, but higher BCT’s secondary current as well. The manufacturer should be
consulted regarding the BCT’s capability, if the transformer is loaded beyond its rating.

2.6.5 Stray Flux Heating

Stray flux produces localized heating in any metallic part. This heating results from in-
duced eddy-current losses, harmonics losses, and some hysteresis losses. Under extreme
conditions of transformer overvoltage, stray flux increases disproportionately due to core
saturation.

15
Various methods for stray flux control include the use of insulated (non-metallic) sup-
ports at the top and bottom of the coil windings, vertical core-clamp configurations, spe-
cial non-magnetic supports for LV bushings and BCT’s associated with high-current
leads, and tank wall shields. Stray flux can also be controlled in a magnetic circuit design.

2.6.6 Bubble Generation

Gas bubbles[13-23] within transformer oil are of a serious concern, since the dielectric
strength of the gases is significantly lower than the dielectric strength of the oil or the cel-
lulose insulation. Bubbles can form in the transformer from gas generated during faults or
from sudden overloading. The generated gas tends to re-dissolve after a long period of
time (approx. 20 hours[21].

Three mechanisms [23,24] are known by which gas can generate bubbles. IEEE recom-
mends the absolute upper limit of 180C winding hottest-spot temperature.

 Super-saturation of the oil with a blanket gas.

 Thermal decomposition of cellulose insulation.
 Vaporization of absorbed moisture in the cellulose.

2.7 Transformer Design Optimization

There are various optimization objectives one can identify, such as, minimize optimiza-
tion of unit cost, maximize efficiency, minimize the total life cycle owning cost, and meet
the guaranteed losses. The number of variables changes according to these design objec-
tives. Usually, computer aided design tools are necessary to efficiently solve this multi-
variable problem.

2.7.1 Design for Maximum Efficiency

Maximum efficiency occurs in a transformer at a pu loading (K), when the winding loss
(K2Pcu@FL) is equal to the core loss (constant). Mathematically,

K2 Pcu@FL = Pcore

1 Pcu @ FL
 = loss ratio (R) ( 2.15 )
K2 Pcore

Where, Pcu@FL is the winding loss at full load

Pcore is the core loss at rated voltage

16
The loss ratio indicates the loading point that yields maximum efficiency. The typical
value of loss ratio varies between 4-8. The loss ratio of 4 and 8 corresponds to 0.5 and
0.35 per-unit loading (K) respectively for maximum efficiency.

2.7.2 Design for Minimum Cost of Material

Variations on material costs between different designs are determined mainly by the pro-
portion of iron core and copper in the transformer. Depending on the relative costs of
these materials, an optimum mass ratio for minimum cost can be obtained. It has been
shown [5] that for the minimum cost of materials, the cost of the copper is equal to the
cost of core. In term of specific cost per-unit mass ($/lb), the mass ratio is equal to,

Core Mass ( lb ) Copper Cost per pound ($ / lb )

 ( 2.16 )
Copper Mass ( lb) Core Cost per pound ($ / lb )

The typical cost ratio of copper and silicon-steel core is 3 to 1. It has been found that for
large transformer designs, a good correlation exists between the actual and optimum val-
ue of cost ratio. However, in secondary transmission, distribution units, the actual cost
ratios are higher and the mass ratios are lower than optimum, indicating that the amount
of copper relative to iron is greater than that required for minimum cost of material.

2.7.3 Design for Minimum Total Owning Cost (TOC)

The engineers traditionally determine the cost effectiveness by calculating the values of
the no-load and load losses. These are often referred to as the “A” and “B” factors. They
are multiplied by no-load and load losses respectively and applied to the total owning
cost (TOC),

TOC = NLL * A + LL * B + C ( 2.17 )

Where, NLL is the no-load loss in kW,

A is the capitalized cost per kW of NLL (A factor),
LL is the load loss in kW at the transformer's rated load,
B is the capitalized cost per kW of LL (B factor),
C is the initial cost of the transformer including transportation, sales taxes, and other
costs to prepare it for service.

The average values of A and B used in the US is $3,430 and $1,090 per kW, respectively.
The transformer that meets the transformer purchaser’s technical specification with the
lowest total owning cost becomes the most cost-effective transformer.

2.8 Transformer Monitoring and Diagnostics

It is of utmost importance for all electric utilities to minimize the overall cost of running
the electric power systems while maintaining an overall reliable and robust electrical sys-
17
tem. The power transformers are one of the most expensive and critical elements in a
power system. Cost saving can be realized through a delay in the replacement of trans-
formers and a reduction in maintenance effort. Well designed monitoring systems for
power transformers can help achieve these aims. A catastrophic failure, sudden and un-
planned outage of a transformer is associated with considerable costs that include loss in
produced energy, process down-time, and penalties. The repair costs normally are very
expensive. With modern technology, it is possible to monitor a large number of parame-
ters. The economic constraints make it useful to differentiate between “Monitoring” and
“Diagnostics”. Monitoring is here defined as on-line collection of data, sensor develop-
ment and development of methods for condition measurement of power transformers. Di-
agnostics contains interpretation of data, but also all off-line measurements.

2.8.1 On-Line Monitoring

There are a number of on-line monitoring devices available in the market:

 Dissolved Gas-in-Oil Analysis (DGA): The gas-in-oil analysis is the most impor-
tant feature of an online monitoring system for condition assessment. In cases of over-
heating, partial discharge or local breakdown inside the transformer, several gases are
produced and dissolved in the oil. If a certain generation rate of gas is exceeded, gas bub-
bles arise. The most important transformer fault gases are H2, CO, CO2, CH4, C2H2, C2H4
and C2H6. In every case of an internal fault, H2 will be produced. The content of individ-
ual gases depends on the kind of fault. IEEE C57.104-1991[25] specifies the interpretation
of gases generated in oil. There are two types of gas sensors commercially available. (i)
The semiconductor sensor [26] that detects only H2 or a composition of H2, and (ii) More
complex sensors can detect several or all of these gases. These are based on infrared
technology or gas chromatography [27,28]

 Moisture Content: Increases in the moisture levels of the oil indicate that there is
a possible rupture or faulty seal in the tank. Excessive moisture increases the electrical
conductivity of the transformer oil and aging rate of insulation.

 Partial Discharge: There are a couple of methods to detect partial discharge. One
is through a high frequency current transformer connected to the transformer neutral and
a capacitive voltage coupler. These sensors detect an impulse due to partial discharge and
discriminate between internal and external partial discharge. Another method is by an
acoustic signal. The ultrasonic transducers are used as acoustic detection device located
outside the main tank.

 Load Tap Changer (LTC) Monitoring: Transformers with LTC have generally
higher failure rate. The mechanical and electrical failures (springs, bearings, shafts, drive
mechanisms, transition resistors, insulation and contacts) of LTC ranks high. Three sim-
ple measurements can be used to monitor faults in LTC. These are measurement of motor
current measurement, temperature different between the LTC compartment and main
tank, and acoustic technique.

18
  Temperature: The hottest-spot of the winding is the main limiting factor. Conven-
tional temperature measurements using thermometers or thermocouples are not a direct
hottest-spot measurement. However top-oil temperature is an approximate indicator for
hottest-spot temperature. A hottest-spot thermometer is an optional device. A current
proportional to the winding current is supplied to the heater from current transformer in
the winding in which the hottest-spot is to be measured. Fiber optic sensors can be in-
stalled in the winding when the transformer is manufactured. Two types of sensors have
been used, fibers which measure the temp. in one point, and distributed fibers which
measure the temperature along the length of the fiber.

 Oil Conductivity: This is accomplished by using a porous ceramic sensor and

electrodes and analyzing the sensor’s leakage current.

2.8.2 Off-Line Diagnostics

A number of “Off-Line Diagnostics” schemes for transformer health monitoring have

been in existence for a long period of time including]29].

1) Gas-in-Oil or Dissolved Gas Analysis (DGA)

2) Partial Discharge (PD)
3) Degree of Polymerization (DP)
4) Furanic Compounds Analysis
5) Thermography
6) Frequency Response Analysis
7) Leakage Inductance
8) Dielectric Response Oil Analysis

19
3. Transformer Thermal Models, Overloading, and Loss-of-
Insulation-Life
Transformer’s overall life expectancy and overloading capabilities depend on several fac-
tors. However, it is determined primarily by the winding hottest-spot temperature. The
overloading guideline and the corresponding loss-of-life calculation as presented in the
ANSI/IEEE C57.91-1995 [1] is discussed here. Reference to the IEC Guide [2] is available
in the Appendix.

3.1 IEEE “Classical Thermal Model” (Clause 7)

ANSI/IEEE C57.91, is based on average characteristics of a wide range of transformer

ratings and designs. The guide uses the top-oil rise and the hottest-spot conductor rise
over top-oil to calculate the hottest-spot temp..and is defined by:

T HS  T A   T TO   T G ( 3.1 )

Where, TA is the ambient temperature

TTO is the top-oil rise over the ambient temperature, and
TG is the winding hottest-spot temperature rise over top-oil temperature.

Figure 3.1 shows corresponding thermal diagram. The assumptions are::

 The oil temperature inside the windings increases linearly from bottom to top of
the winding regardless of the cooling type,
 The temperature rise of conductor at any position up the winding increases linear-
ly and parallel to the oil temperature rise,
 The hottest-spot temperature rise is higher than the temperature rise of the con-
ductor at the top of the winding, because of the increase in stray losses. To find
hottest-spot temperature, the hottest-spot allowance (TG) is added to the top-oil
temperature.

20
 Figure 3.1 Transformer’s temperature profile for IEEE classical thermal model

The top-oil temperature rise under steady-state condition is proportional to the total trans-
former loss and given by:

n n
P   K 2R  1
TTO  TTOR  ( PT )  TTOR   loss   TTOR  
n
 ( 3.2 )
 PR   R 1 

I Actual load (MVA)

K    Per - Unit Loading
IR Rated load (MVA)

where: TTOR is the top-oil temperature rise at rated load

PT is the per-unit transformer losses
Ploss are the total losses in watts
PR is the total losses at rated load
K is per-unit loading
R is the loss ratio and n is an exponent (oil).

The rated condition refers to maximum (highest) rating (rating at 65C average winding
temperature rise for transformers with thermally upgraded insulation)

The winding hottest-spot temperature rise over top-oil temperature is given by:

21
  TG   TGR  ( K 2 ) m ( 3.3 )

where: TGR is the winding hottest-spot temp. rise over top-oil temp. at rated load
K 2 is the per-unit winding loss, and
m is an exponent (winding).

The IEEE standard recommends the use of imbedded detector for measuring TGR, the
winding hottest-spot temperature rise over top-oil temperature at rated load. An alternate
approach is to add a fixed value over the average winding rise given by:

TGR  average winding rise over top oil  15  C

TGR  ( TWR  TTOR )  15  C ( 3.4 )

Where, TWR is average winding temperature rise at rated load.

TGR is calculated from average winding rise over top-oil temperature rise plus 15C cor-
responding to 65C winding rise respectively.

Figure 3.1 illustrates the relationships. Table 3.1 also provides the necessary limits of the
design values for oil-cooled transformer

Table 3.1: Thermal characteristics of classical thermal model at rated load

(indicated by suffix R), IEEE Std. C57.91-1995

Cooling System
OA FA FA NDFOA DFOA
<133% >133%
Hottest spot rise THSR 80 80 80 80 80
Top oil rise TTOR 55 50 45 45 45
Oil time constant TOR (hr) 3.0 2.0 1.25 1.25 1.25
Winding time constant GR (min) 5.0 5.0 5.0 5.0 5.0
Loss ratio R 3.2 4.5 6.5 6.5 6.5
Oil exponent n 0.8 0.9 0.9 1.0 1.0
Winding exponent m 0.8 0.8 0.8 0.8 1.0

Equations 3.2 and 3.3 are steady-state solutions. In order to find transient solution for the
top-oil temperature, the energy balance equation is used

Energy Generated = Energy Radiated + Energy Absorbed

The differential equation for the top-oil temperature rise may be written as [31]

22
 Ploss dt  k  TTO dt  Cd  TTO ( 3.5 )

where: Ploss is the total losses in the transformer in watts

k is the radiation constant in W/  C
C is the transf. thermal capacity in Watt - min./  C , and can be calculated from
the empirical relationship: C = 0.06 * (weight of core and coil assembly in
pounds) + 0.04 * (weight of tank and fitting in pounds) + 1.33 * (gallons of oil);

Equation ( 3.5 ) can be rewritten as

C dTTO P
  TTO  loss
k dt k

dTTO P
 TO   TTO  loss ( 3.6 )
dt k

Where,  TO is the oil time constant and defined by the ratio of C and k.

From energy balance equation ( 3.5 ), dTTO is equal to zero at steady-state. The sub-
script u in TTO denotes the ultimate temperature. This gives

n
P  K 2 R 1
Ploss  kTTO ,u or loss  TTO,u  TTOR   
k  R 1 

Equation (3.6) then becomes, (the subscript i in TTO denotes the initial temperature).

dTTO
 TO   TTO  TTO ,u ( 3.7 )
dt

TTO (0)  TTO ,i

The solution of the differential equation ( 3.7 ) is

TTO  ( TTO ,u  TTO ,i )(1  e  t /  TO )  TTO ,i ( 3.8 )

23
The oil time constant ( TO ) in equation ( 3.6 ) to ( 3.8 ) can be written as

C ( TTO ,u  TTO ,i )
 TO  C
k Ploss

The subscript u and i denote for ultimate and initial values, respectively. The value of the
oil time constant ( TO ) varies with the top-oil temperature rise and power losses. The
manufacturer usually provides the oil time constant at rated load, and is given by:

 TTOR
 TOR  C
PR

(Ploss,u  Ploss,i )
k
(TTO,u  TTO,i )

The oil time constant at any load ( TO ) can then be expressed in terms of the correspond-
ing value at rated load by the following equation 3.9):

 TTO ,u   TTO ,i 
  
 TOR TTO ,u  TTO ,i  TTOR   TTOR 
 TO  *  ( TOR )
TTOR (Ploss ,u  Ploss ,i ) / PR 1 1
( 3.9 )
 TTO ,u  n  TTO ,i n
    
 TTOR   TTOR 

With the hottest-spot winding temperature rise over top-oil temperature being proportion-
al to the transformer winding loss, the corresponding initial and ultimate temperature rise
are given by

TG ,i  TGR  ( K i2 ) m

TG ,u  TGR  ( K u2 ) m ( 3.10 )

Where, m is winding exponent. Similar to oil time constant derivation, winding time con-
stant ( GR ) can be derived as:
24
 TGR
 GR  M W C PW
PWR

 TG ,u   TG ,i 
  
 TGR   TGR 
 G  ( GR ) 1 1 ( 3.11 )
 TG ,u  m  TG ,i m
    
 TGR   TGR 

Where: MW is winding mass, CPW is specific heat of winding material, and PWR is the
winding loss at rated load. The winding hottest-spot rise over top-oil differential equation
is written as:

dTG
G   TG  TG ,u ( 3.12 )
dt

The solution of the differential equation ( 3.12 ) is

TG  (TG ,u  TG ,i )(1  e  t /  G )  TG ,i ( 3.13 )

As the hottest-spot time constant is very short (between 5-10 minutes) compared to the
top-oil value (1-3 hours), the hottest-spot conductor rise can be approximated as in equa-
tion ( 3.10 ) by

 T G   T GR  ( K 2 ) m

IEEE “Clause 7” thermal model has some limitations. The model does not include the
behavior of oil in cooling duct and results in higher hottest-spot temperature. The model
also ignores the time delay due to the ambient temperature change (TA ) . The modified
top-oil temperature equation by adding ambient temperature variation 31] into the top-oil
temperature equation is:

dTTO
 TO  TTO  TTO ,u  T A ( 3.14 )
dt

25
The above equation and zero initial conditions and rearrangement yields

1 1
 TTO ( s )    TTO ,u ( s )   T A ( s ) ( 3.15 )
1   TO s 1   TO s

The solution can be obtained by numerical method from the block diagram shown below
(Figure 3.2). The unmodified model gave large error (5C error for 10C daily ambient
temperature variation).

Figure 3.2: Block diagram of the modified transient heating equations

3.2 IEEE Alternate Thermal Model (Annex G)

The IEEE Classical model uses the top-oil temperature rise over ambient temperature to
calculate the winding hottest-spot temp. Recent investigations [33,34] have shown that dur-
ing overloads the temperature of the oil in the winding cooling ducts rises rapidly at a
time constant equal to that of the winding (contrary to the oil). During this transient con-
dition (as shown in Figure 3.3), the oil temperature adjacent to the hottest-spot location is
higher than the top-oil temperature in the tank. The calculations in Annex G are based on
Pierce[34] and account for the type of fluid, cooling mode, winding duct oil temperature
rise, resistance and viscosity changes, stray losses, eddy current losses, hottest-spot loca-
tion, ambient temperature, and load changes.

26
 Figure 3.3: Variation of duct oil temperature and top-oil temperature after a load step 0-
100% with an ambient temperature of (-10C)

The principle of the model is governed by two basic heat transfer equations:

Heat balance equation: QGEN = QABS + QLOST

Where, QGEN is heat generated by heat sources, QABS is heat absorbed in heat sources, and
QLOST is the heat lost to cooling medium.

And heat absorption equation: Q ABS  M  C P   T

Where, M is the mass of absorbed material and CP is the specific heat of material, and T
is the temperature difference.

The hottest-spot temp. of the model is made up of the following components:

T HS  T A   T BO   TWO   T HS / WO ( 3.16 )

Where,  T BO is bottom-oil rise over ambient,

 TWO is duct-oil rise at winding hottest-spot location over bottom-oil,
THS/WO is winding hottest-spot rise over adjacent duct-oil temperature at hottest-spot lo-
cation.

27
The calculation consists of a number of iterations of small time interval, t. All variables
are updated for every iteration. All equations use actual temperature, and not the tempera-
ture rise.

3.2.1 Duct-oil temperature

Within the winding, the heat generated, absorbed and lost by the winding is defined by:

QWGEN  K 2 (T KW PWR  PER / T KW )  t

5/ 4 1/ 4
 T  TDAO   WR 
QWLOST   W    ( PWR  PER )t For OA,FA,NDFOA
 TWR  TDAOR   W 
 T  TDAO 
QWLOST   W (PWR  PER )t ; for DFOA
 TWR  TDAOR 
QWGEN  QWLOST  QWABS  M W C PW  TW ( 3.17 )

where: QWGEN , QWABS , and QWLOST is the heat generated, absorbed and lost in time interval
(t).

K is the transformer loading in per-unit.

TKW is the winding resistance correction factor.
PWR and PER are the winding and eddy current losses at rated load.
TW is the average winding temperature.
TWR is the average winding temperature at rated load.
TDAO is the average duct-oil temperature.
TDAOR is the average duct-oil temperature at rated load.
W is oil viscosity in the duct.
WR is oil viscosity in the duct at rated load.
MW is mass of the winding.
CPW is the specific heat of the winding.

The new value of the average winding temp. is calculated from TW from ( 3.17 ).

TW,new = TW,old + TW ( 3.18 )

The duct-oil temperature rise is given by

x
 QWLOST 
TTDO / BO  (TTDOR  TBOR )  ( 3.19 )
 ( PWR  PER )t 

28
Where, TTDO/BO is the duct-oil temperature rise over bottom-oil temperature, TTDOR and
TBOR is the top-duct oil and bottom-oil temperature, and x is exponent of duct-oil rise (0.5
for OA, FA, NDFOA and 1.0 for DFOA)

The duct-oil temperature rise gives the update to TTDO, TDAO, and TWO as

TTDO = TBO + TTDO/BO

TDAO = (TTDO+TBO)/2

TWO  T BO  HHS   TTDO / BO

Where, HHS is the location of hottest temperature of the winding. It is equal to 1.0 when
the hottest-spot is at the top of winding and equal to 0.0 when at the bottom.

3.2.2 Hottest-spot temperature

In finding the hottest-spot temperature, it is assumed that the entire winding is at the hot-
test-spot temperature. The winding loss at hottest-spot temperature is

PWHSR  T K PWR

234 .5  THSR
TK  for Cu winding and
234 .5  TWR

225  THSR
TK  for Al winding
225  TWR

where PWHSR is the winding loss at hottest-spot temperature, TK is the temperature correc-
tion factor, THSR is the hottest-spot temperature at rated load.

Similar to the duct-oil temperature, within the winding at hottest-spot location, the heat
generated, absorbed and lost by the winding is defined as

Q HSGEN  K 2 (T KHS PWHSR  PEHSR / T KHS )  t

29
 5/ 4 1/ 4
 T  TWO    HSR 
QHSLOST   HS    ( PWHSR  PEHSR )t ; for OA,FA,NDFOA
T  T
 HSR WOR  
 HS 

 T T 
QHSLOST   HS WO ( PWR  PER )t ; for DFOA
 THSR  TWOR 

Q HSGEN  Q HSLOST  Q HSABS  M W C PW  T HS ( 3.20 )

Where: Q HSGEN , Q HSABS , and Q HSLOST is the heat generated, absorbed and lost in time (t).
TKHS is the winding resistance correction factor.
PHSR and PEHSR are the winding and eddy current losses at rated load at hot
test-spot temp.
THS is the hottest-spot temperature.
THSR is the hottest-spot temperature at rated load.
HS is the oil viscosity at hottest-spot location.
HSR is the oil viscosity at hottest-spot location at rated load.

The new value of hottest-spot temperature is calculated from THS of equation ( 3.20 )

THS,new = THS,old + THS ( 3.21 )

3.2.3 Bulk-oil temperature

The model considers the heat that is generated, absorbed, and lost (to the air)

QOGEN  QWLOST  Q Stray  QCore

1/ y
 T  TA 
QOLOST   OAVG  PTR t
 TOAVGR  TAR 

QWLOST  QStray  QCore  QOLOST  MC P  TOAVG ( 3.22 )

where: QOGEN , QWLOST , QStray , QCore , and QOLOST is the heat generated in oil, heat lost form
winding to oil, stray loss heat, core loss heat, and the heat lost.
TOAVG is the average bulk-oil temperature, (TTO+TBO) /2.
TOAVGR is the average bulk-oil temperature at rated load.
y is the exponent of average oil rise (See Table 3.2)
30
 PTR are the total transformer losses at rated load.
MCP is the summation of the product of mass and specific heat of tank, core,
and oil excluding winding.

Table 3.2: Summary of exponents

Exponent OA FA NDFOA DFOA

x Duct oil rise 0.5 0.5 0.5 1.0
y Average oil rise 0.8 0.9 0.9 1.0
z Top to bottom oil rise 0.5 0.5 1.0 1.0

The new value of the average bulk oil temp. is calculated from TOAVG Equation ( 3.22 )

TOAVG,new = TOAVG,old + TOAVG ( 3.23 )

The top- and bottom-oil temperature difference is given by:

z
 Q 
TTO / BO  (TTOR  TBOR ) OLOST  ( 3.24 )
 PTR  t 

Where, TTO/BO is top- and bottom-oil temperature difference,

TTOR and TBOR is top-oil and bottom-oil temperature, and
z is exponent of top-oil to bottom-oil rise in radiator (0.5 for OA, FA and 1.0 for
NDFOA, DFOA).

The top- and bottom-oil temp. difference gives the update to TTO and TBO as

TTO = TOAVG + TTO/BO /2

TBO = TOAVG - TTO/BO /2

When the calculation completes the new value of all temperatures, the calculation loops
back to starting point. It reiterates by time step of (t) until it reaches the end of the load
cycle. The steady-state temperatures rise can be calculated from the following equations

 T  TBOR   T  TBOR 
TTO   TOR   ( PT ) y   TOR   ( PT ) z ( 3.25 )
 2   2 

31
  T  TBOR   T  TBOR 
TBO   TOR   ( PT ) y   TOR   ( PT )
z
( 3.26 )
 2   2 

F or OA, FA, and DFOA Cooling,  TTDOR / BOR   TTDOR   T BOR

For NDFOA Cooling,  TTDOR / BOR   TWR   T BOR

For all cooling type  TTDO / BO   TTDOR / BOR  K 2 x ( 3.27 )

For OA, FA, and NDFOA Cooling,

  TTDOR / BOR   TTDO / BO

 TW    TWR   TBOR    K 1.6  TBO  ( 3.28 )
 2  2

 T HS   T HSR   T BOR   TTDOR / BOR   K 1.6   T BO   TTDO / BO ( 3.29 )

For DFOA Cooling,

 T  TTDO / BO
 TW   TWR  TBOR  TDOR / BOR   K 2.0  TBO  ( 3.30 )
 2  2

 T HS   T HSR   T BOR   TTDOR / BOR   K 2.0   T BO   TTDO / BO ( 3.31 )

This model requires more information on the transformer than the classical model. How-
ever, it provides more informative results. For this research, this model has been used. It
yields more accurate winding losses and hottest-spot calculations.

3.3 IEC 354 Thermal Model

IEC 354 thermal model[2] for transformer hottest-spot calculations is very similar to the
IEEE classical model (Clause 7). IEC 354 uses 20C average ambient temperature and
hottest-spot temperature of 98C for 65C average winding rise transformer (compared to
30C average ambient temperature and 110C hot-spot temperature for ANSI/IEEE). IEC

32
354 uses different temperature equations (by size and cooling type) to calculate hottest-
spot temperature. A direct comparison of the IEEE/ANSI guide and the IEC 354 can be
found in reference [11].

3.4 Loss-of-Life Calculation, IEEE Method

In the IEEE loading guide (C57.91-1995), aging equations have been changed to accom-
modate the recent results. There is no longer the absolute life value. Instead, “the relative
aging rate” and “per-unit life” have been introduced. The per-unit life (L) for 65C aver-
age winding temperature rise transformer is defined by (3.32). For, THS = 110C , the per-
unit life = 1.00.

 15, 000 
 
18  THS  273  ( 3.32 )
L = 9.80 x 10 e

The equation for the Relative Aging Factor (FAA) can be derived from ( 3.32 ):

 15 , 000 
 39.164   ( 3.33 )
T HS  273 
FAA = e 

The value of FAA is greater than 1, when the hottest-spot temperature is greater than
110C, suggesting loss-of-life (from normal aging) and less than 1 when hottest-spot
temperature is less than 110C, meaning life extension. Equation ( 3.34 ) may be used to
calculate the equivalent aging of the transformer with respect to the reference temperature
(110C) which will be consumed in a given time period (T):

1T
FEQA   FAA dt ( 3.34 )
T0

Equation ( 3.34 ) is a dimensionless quantity. The actual loss-of-life in hrs. can be calcu-
lated by multiplying FEQA and T(hrs). The % loss of life can then be calculated as:

FEQA x T
% Loss of Life   100 ( 3.35 )
Normal Insulation Life

As discussed earlier, the normal insulation life of an oil-filled transformer is not uniquely
defined. IEEE provides some acceptable guidelines for the normal insulation life values
and the corresponding criteria. This is summarized in Table 3.3.

33
 Table 3.3: Normal insulation life per IEEE C57.91-1995

Normal Insulation Life

Basis
Hours Years

50% Retained Tensile Strength(Former C57.92) 65,000 7.42

20% Retained Tensile Strength of Insulation and/or 150,000 17.12
200 Retained Degree of Polymerization (DP)
Distribution Transformer’s Functional Life 180,000 20.55

Note: Data in Table 3.4 is applicable to well-dried, oxygen-free, 65C avg.

winding rise, hottest-spot temperature of 110C transformer

3.5 Comparison of the IEEE Thermal Models: Numerical Examples

As mentioned earlier in this Chapter, the IEEE classical model (Clause 7) is simpler and
requires less information. The model, as discussed in IEEE Standard Annex G, on the
other hand is complex and requires more input data to calculate the transient responses of
transformer. The comparison of the input data requirement is tabulated in Table 3.4.

Table 3.4: Required data for IEEE thermal models

IEEE Classical Model (Clause 7) IEEE Detailed (Annex G) Model

Top-oil temperature rise at rated load Top-oil temperature rise at rated load
Hot-spot temp. rise over top-oil @ rated load Hottest-spot temperature rise at rated load
Loss ratio at rated load Average winding temp. rise at rated load
Winding time constant Bottom oil temperature rise at rated load
Oil time constant or Losses data from test report
Weight of core & coil kVA base of test data
Weight of tank & fittings Winding temperature rise at tested rating
Gallons of fluid Winding losses
Type of cooling system Winding eddy current losses
Stray losses
Core losses
Weight of core & coil
Weight of tank & fittings
Gallons of fluid
Type of cooling system
Type of cooling fluid
Type of winding material
Winding time constant
Location of hottest spot
pu eddy current losses at hot-spot location

Several simulations were performed to compare the two methods. Although the two
models use different approaches and heat transfer equations, the steady-state solutions for
the top-oil and hottest-spot temperatures are surprisingly close. Table 3.5 contains partial
transformer data (design values) used in the comparison for steady-state temperature rise.
The transformer loading is increased by an increment of 0.2pu to a continuous loading of
34
 1.4pu. The comparison is tabulated in Table 3.6. The temperature difference is within 2-
3C for most cases except for DFOA and NDFOA cooling systems and at very high load-
ing of 1.2 and 1.4pu. The Annex G model yields much higher top oil and hottest-spot
temperature at loading beyond 1.2pu. This is due to the higher duct oil rise exponent, x of
DFOA cooling system (see Table 3.2).

Table 3.5: Transformer data for temp. rise comparison at steady-state, IEEE methods

Clause 7 input data THSR = 110C

(Design Values) Loss ratio (R) = 4.1
TTOR = 55C for OA
TTOR = 45C for FA, NDFOA, and DFOA

Annex G input data THSR = 110C

(Design Values) TTOR = 55C for OA
TTOR = 45C for FA,NDFOA, and DFOA
TBOR = 25C
PWR = 138,257 watts
PSR = 42,085 watts
PC = 43,986 watts
Mass of core & coil = 61,050 lb.
Mass of tank & fittings = 26,050 lb.
Oil volume = 4,110 gallons

Note: The suffix letter “R” on all the suffixes is referred to the rated (or full-load)
condition (design limit)

IEEE Annex G thermal model, however, has improved transient response, since it takes
heat transfer in duct oil into considerations. A computer program is written. The program
is capable of calculating the temperature profiles from the IEEE thermal models. The
program also computes the transformer loss-of-life based on insulation deterioration.

35
 Table 3.6: Comparison of top-oil and hottest-spot temperature of IEEE thermal models
(constant load, 30C ambient temperature)

Cooling Load Top oil temperature Hottest-spot temperature

type (pu) Clause 7 Annex G Diff. Clause 7 Annex G Diff.
OA 0 44.9 47.5 +2.6 44.9 47.5 +2.6
0.2 46.9 49.3 +2.4 48.8 51.3 +2.5
0.4 52.4 54.3 +1.9 58.1 60.5 +2.4
0.6 60.9 62.2 +1.3 71.9 73.7 +1.8
0.8 71.8 72.4 +0.6 89.3 90.3 +1.0
1.0 85.0 85.0 0 110.0 110.0 0
1.2 100.1 99.9 -0.2 133.5 134.2 +0.7
1.4 116.9 117.2 +0.3 159.7 161.5 +1.8
FA 0 40.4 42.5 +2.1 40.4 42.5 +2.1
0.2 41.9 43.9 +2.0 44.6 46.8 +2.2
0.4 46.4 48.0 +1.6 54.4 56.5 +1.1
0.6 53.5 54.6 +1.1 68.9 70.7 +1.8
0.8 63.1 63.6 +0.5 87.6 88.6 +1.0
1.0 75.0 75.0 0 110.0 110.0 0
1.2 89.1 89.1 0 136.0 136.0 0
1.4 105.3 106.1 +0.8 165.3 165.8 +0.5
NDFOA 0 38.8 40 +1.2 38.8 40.0 +1.2
0.2 40.3 41.4 +1.1 42.9 46.9 +4.0
0.4 44.6 45.6 +1.0 52.7 57.9 +5.2
0.6 51.8 52.5 +0.7 67.3 72.9 +5.6
0.8 62.0 62.3 +0.3 86.5 90.3 +3.8
1.0 75.0 75.0 0 110.0 110.0 0
1.2 90.9 91.0 +0.1 137.8 132.1 -5.7
1.4 109.7 110.6 +0.9 169.7 156.8 -12.9
DFOA 0 38.8 38.8 0 38.8 38.8 0
0.2 40.3 40.1 -0.2 41.7 41.3 -0.4
0.4 44.6 44.2 -0.4 50.2 48.8 -1.4
0.6 51.8 51.0 -0.2 64.4 61.9 -2.5
0.8 62.0 61.1 -0.9 84.4 81.7 -2.7
1.0 75.0 75.0 0 110.0 110.0 0
1.2 90.9 93.9 +3.0 141.3 154.1 +12.8
1.4 109.7 120.1 +10.4 178.3 219.6 +41.3

The program features easy data input, graphical user interface, save, print and copy to
clipboard. The program calculates hottest-spot temperature, top-oil temperature and other
temperature profiles using daily load and ambient temperature profiles based on IEEE
thermal model Annex G, Clause 7, and IEC 354. It also plots temperature profiles and
text results. The program calculates the loss-of-transformer functional life based on ther-
mal model (IEEE only). Program utilizes monthly load, temperature profiles, and load
growth rate from table. Estimated remaining tensile strength and degree of polymeriza-
tion are also reported.

The graphical outputs of a test case for step load changes are plotted in Figure 3.4 and
Figure 3.5. The step load increases from 0 to 1.0 pu and remains constant for 4 hours.

36
The forced-air (FA) cooled transformer data from Table 3.45 is utilized. The ambient
temperature is held constant at 30C.

When compared, it is seen that, the hottest-spot temperature in Annex G model rises
more sharply when the load steps up. With the inclusion of heat transfer in duct oil, for
any typical load profile, the IEEE Annex G thermal model yields higher hottest-spot tem-
perature than corresponding IEEE Clause 7 thermal model , so does the loss of insulation
life.

Temperature Prof ile Plots (IEEE A nnex G Thermal Model)

120 5
Hot s pot
4.5 Top oil
Loading
100
4

3.5
80

Loading (Per unit)

Temperature (C)

60 2.5

40
1.5

1
20

0.5

0 0
0 2 4 6 8 10 12
Time (hr)

Figure 3.4: Temperature profiles of IEEE Annex G thermal model subjected to step load

Temperature Prof ile Plots (IEEE Claus e 7 Thermal Model)

120 5
Hot s pot
Top oil
4.5
Loading
100
4

3.5
80
Loading (Per unit)
Temperature (C)

60 2.5

40
1.5

1
20

0.5

0 0
0 2 4 6 8 10 12
Time (hr)

Figure 3.5: Temperature profiles of IEEE Clause 7 thermal model subjected to step load

37
4. Probabilistic Assessment of Transformers Loss-of-Life
4.1 Introduction

Utilities generally select the size and overload the transformers, when necessary, follow-
ing the IEEE loading guide or a guide of their own. This generally provides the determi-
nistic calculation of transformer loss-of-life at given elevated hottest-spot temperatures.
In this method, the exact mean ambient temperature and the transformer hottest-spot
temperature values are assumed known. However, in a real-life practical application, be-
cause of a number of variables, the exact amount of overloading is difficult to determine.
Use of some statistical approaches to evaluate these parameters which may be obtained
from utility’s monitoring system and weather station data, may provide more flexibility
and better result. In the probabilistic analysis, discussed in this chapter, ambient tempera-
ture (TA) and the transformer loading (K) are taken as uncertainties. Monte Carlo tech-
nique is utilized to calculate the probabilistic distribution of the winding hottest-spot
temperature and calculate the expected value of transformer aging rate and the corres-
ponding percent loss-of-life. This also can help the utilities to make decisions regarding
the transformer overloading and the sizing of transformer under uncertainties.

4.2 Modeling

4.2.1 Transformer Loading (K)

The daily loading data of a transformer can be obtained from some real-time monitoring
systems (via SCADA). However, the future forcasting depends on a number of variables.
In a simplified model, the transformer loading is assumed to be charaterized by a
Gaussian distribution (normal distribution) with its mean value of (k) and between 5-20
percent [36] of the mean as its standard deviation (σk). Mathematically,

K = Gauss (  k ,  k2 ) ( 4.1 )

4.2.2 Ambient Temperature (TA)

Similarly, the ambient temperature at any given day and time can also be described by a
Gaussian distribution from the available historical or forecasted data. The typical value of
ambient temperature standard deviation is between 10-30 percent [36] of its mean. Mathe-
matically,

TA = Gauss (  a ,  a2 ) ( 4.2 )

38
4.2.3 Correlation between loading and ambient temperature ()

In a real power system, the transformer loading is related with the ambient temperature.
The correlation coefficient () ranges between (–1) to (+1). The correlation is generally
positive in the summer months and turns negative in the winter.

4.3 Monte Carlo simulation

Monte Carlo simulation technique is utilized to randomly create a set of data points cor-
responding to a transformer loading (K) at an ambient temperature of (TA) with correla-
tion coefficient of (). A hottest-spot temperature profile is then produced from the ther-
mal model discussed earlier. Then relative aging rates (FAA) are calculated from the hot-
test-spot temperatures using equation ( 3.33 ). The Expected Value (mean) of the relative
aging rate is calculated as defined by equation ( 4.3 )and substituted in equation (4.4) for
the calculation of percent loss-of-life. Mathematically,

E ( F AA )  F

AA Pr ( F AA ) dF AA ( 4.3 )

The percent Loss of Life becomes

 E(F AA
) dt
( 4.4 )
% Loss of Life  0
 100
Normal Insulation Life

The Monte Carlo simulations generate bivariate normal distributions with correlation
coefficient (between the ambient temperature and transformer loading) in a study region.
In order to generate the bivariate normal distributions with a correlation coefficient, a pair
of independent standard normal distributions, U,V:N(0,1), is generated from the computer
program MATLAB or Box-Muller method[37]. By substituting U and V in equation ( 4.3 )
and ( 4.4 ), yields bivariate normal distributions X and Y that represent ambient tempera-
ture (TA) and transformer loading (K), respectively as:

X   X   X (U ) ( 4.5 )

Y   Y   XY  Y (U )   Y 1   XY
2
(V ) ( 4.6 )

39
Where,  X , Y and  X ,  Y are the means and standard deviations of variables X and Y,
and  XY is the correlation coefficient between (–1) to (+1). The computation technique
using the Monte Carlo Simulation is displayed in a flowchart (Figure 4.1).

Figure 4.1: Monte Carlo simulation

40
4.4 Numerical example

To better understand the modeling and the Monte Carlo Simulation technique discussed
earlier, an actual simulation is performed. In this example, a constant loading (K) of 1.0
per-unit (  k ) with 5% standard deviation ( k ) , and the ambient temperature’s mean of
30C (  a ) and a 10% standard deviation ( a ) is used. The correlation coefficient () is
assumed randomly to be 0.2. The Monte Carlo simulation generates 500 random numbers
for each random ambient temperature and transformer loading. The random relationship
between the ambient temperature and the transformer loading is plotted in Figure 4.2.
Figure 4.3 depicts the probability of the TA and K relationship generated randomly by the
program.

Figure 4.2: Relationship between ambient temperature and transformer loading with cor-
relation coefficient of 0.2.

The transformer in this example is assumed to be a naturally cooled unit (OA). Figure 4.4
shows the Monte Carlo simulation results of the distribution of transformer hottest-spot
temperatures. It shows that the mean hottest-spot temperature is 109.97C (ideal value of
110C), and the calculated standard deviation is 7.1C.

Figure 4.3: 3-D plot of ambient temperature and transformer loading

41
 Figure 4.4: Distribution of hottest-spot temperature

The probability distribution of the relative aging rate (FAA) as calculated from equation
( 4.3 ) for those randomly selected points is shown in Figure 4.5. The calculated value of
the relative aging rate in this example is 1.28. This is an important number. In a determi-
nistic situation as applied in the thermal model, transformer loading of 1.0 pu at an am-
bient temperature of 30C produces a relative aging rate of 1.0 and a corresponding trans-
former life of 1.0 pu. However, with the probabilistic analysis discussed in this section,
when the uncertainties are taken into account for a naturally cooled transformer at 30C
ambient temperature, the relative aging rate at a mean 1.0 per-unit loading is 28% higher
than the corresponding deterministic (Classical) value. This is because, the relative aging
at a higher ambient temperature (TA) and higher pu loading (K) yields more loss-of-life
compared to the corresponding savings at lower values of K and TA.

Figure 4.5: Distribution of relative aging rate (FAA)

42
The same technique have been extended to other types of cooling and variable correlation
coefficient (). The results are shown in Figure 4.6. The relative aging rate increases ra-
pidly as correlation factor increases from (–1) to (+1) and the value is always greater than
1.0.

Figure 4.6: Relative aging rate with different correlation coefficient on various
types of transformer cooling

In summary, the probabilistic approach always produces a higher value of aging rate and
the corresponding loss-of-life. This is because of much higher insulation deterioration
rate at elevated temperature as compared to low temperature. It is also clear, that the rela-
tive aging rate varies widely with the correlation between ambient temperature and trans-
former loading and the type of transformer cooling.

43
5. Transformer Economic Evaluation
In order to accommodate increased demand, overloading of transformers and accepting
the reduced life expectancy as a result of higher hottest-spot temperature rise must be
evaluated carefully against the delay in transformer replacement or addition of a second
transformer. The utility engineers usually determine the size of transformers from the
current loading, expected future load growth, and other appropriate engineering judg-
ments. Utilities evaluate the economic impact of losses in a transformer prior to purchas-
ing along with the initial price. The conventional method called “Total Owning Cost”. A
detailed analysis called “Revenue Requirement Method” is more appropriate for the In-
vestor-Owned Utilities (IOU’s).

5.1 Loss Evaluation: Conventional Method

Procurement of transformers by the Loss Evaluation is the most widely used and it is
based on the lowest total owning cost. The method evaluates the cost of transformer’s
losses and the purchasing price. The service life of transformer is assumed to be constant
regardless of size. Generally it is chosen to be 30 years.

5.2 Revenue Requirement and Fixed Charge Rate

In the Revenue Requirement method, the annual operating cost is written as:

EA = RA – (PA + DB+ T) ( 5.1 )

where: EA is the annual equity return (net income),

RA is the annual revenue,
PA is the annual production expense,
DB is the depreciation, and
T is the income tax.

Equation ( 5.1 ) can be rearranged:

RA = PA + (EA + DB + T) ( 5.2 )

Except for the annual production expense (PA), the remaining items within the parenthesis
on the right hand side of this equation can be expressed as a function of the capital in-
vestment. Hence, the equation may be re-written as:

RA = PA + FCR * (Capital Investment) ( 5.3 )

44
 RA = PA + ( i + id + it) * (Capital Investment) ( 5.4 )

where: FCR is the fixed charge rate in per unit,

i is the cost of capital (in decimal) or discount rate,
id is the depreciation rate, and
it is the income tax rate.

A very simple, yet realistic, example of fixed charge rate (FCR) is as follow,

 Cost of capital 10.0%

 Depreciation rate 0.6%
 Income tax rate 8.6%
 Fixed charge rate 19.2% or (FCR = 0.192)

Rearranging, the first cost revenue requirement (R) is given by:

PA ( 5.5 )
R= + Capital Investment
FCR

This equation ( 5.5 ) is used in the Loss Evaluation Method and the objective function is
defined as:

PA
min R = + Capital Investment ( 5.6 )
FCR

5.2.1 Transformer Loss Evaluation

The Capital Investment is the purchasing price. PA is the annualized cost of core and
winding losses including cost of demand charge for these losses. When annualized cost of
losses is divided by the fixed charge rate (FCR), it yields the total (net) revenue require-
ment for the cost of losses.

For each of no-load (core), load (winding), and auxiliary losses, there will be a demand
(kW) component based on the capital cost of generation and transmission/ distribution
equipment. In addition, there will be a cost due to the energy (kWh) component. As a re-
sult, the total cost of losses is:

Cost of Losses ($/kW) = Demand Cost + Energy Cost

8760  Annualized Energy Cost ( 5.7 )

= System Investment ($/kW) +
Fixed Charge Rate

45
 = Annualized Demand Charge ($/kW - yr) + 8760  Annualized Energy Cost
Fixed Charge Rate

Equation ( 5.7 ) is conveniently applied by utilities with generation and without genera-
tion. The formulas given below yield the total evaluated cost for each category of losses
separately in $/kW. These costs of losses are then multiplied by transformer’s corres-
ponding kW losses and added to the purchase price, so that losses can be properly taken
into account.

8760(AEC)
NLCR ($/kW) = SI + ( 5.8 )
FCR

8760(AEC)(TLF)(EPR)2 ( 5.9 )
LLCR ($/kW) = SI (PRF) 2(EPR) 2 +
FCR

8760(AEC)(PA) ( 5.10 )
ALCR ($/kW) = SI (PRF) 2 +
FCR
where: NLCR = No-load (core) losses cost rate ($/kW)
LLCR = Load (winding) losses cost rate ($/kW)
ALCR = Auxiliary losses cost rate ($/kW)
SI = System investment cost for additional generation, transmission and
Distribution needs ($/kW)
8,760 = Number of hours in a Year
AEC = Annualized energy cost ($/kWh)
FCR = Fixed charger rate (in decimal)
PRF = Peak responsibility factor
EPR = Equivalent peak ratio
TLF = Transformer loss factor
PA = Probability that the auxiliary cooling will be “ON”

In Total Owning Cost method (TOC), equations ( 5.8 ) and ( 5.9 ) are usually referred to
as “A” and “B” factors, respectively.

Total Owning Cost (TOC) =

( 5.11 )
Purchase Price + NLCR * NLL + LLCR * LL + ALCR * ALL

where: NLL = No-load (core) loss in kW.

LL = Load (winding) loss in kW.
ALL = Auxiliary load loss in kW.

Most commonly, if auxiliary loss is neglected, equation ( 5.11 ) is rewritten as

46
 TOC = Bid Price + A * NLL + B * LL ( 5.12 )

5.2.2 Discussion of Factors

 System Investment (SI) represents the investment in generation, transmission, and

distribution facilities necessary to supply the additional demand resulting from the trans-
former losses at the system peak. There are basically two methods for evaluating the SI
value: (i) the actual construction cost of a recent generating station and the required
transmission/distribution facilities, and (ii) if the utility is purchasing power rather than
self-generating, the SI value can be obtained by dividing the demand charge ($/kW-yr) by
the fixed charge rate (FCR).
 Annualized Energy Cost (AEC): Since the energy price changes with inflation, it
is recommended that AEC be used instead of present energy cost. The AEC is calculated
by listing the projected cost of energy and discount these annual inflated cost by appro-
priate present worth factor. Add each of the present worth value of the energy cost and
multiply by the capital recovery factor (CRF). Mathematically, AEC = sum of present
worth value  CRF

 1-X N     (1+i ) N  
= present energy cost  X     i   
 ( 5.13 )
    (1+i )  1
N
  1-X 

where: N = Transformer book life

1 e
X=
1+i
e = Energy escalation rate (in decimal)
i = Cost of capital or discount rate (in decimal)
CRF = Capital recovery factor

 Fixed Charge Rate (FCR) represents the annual cost necessary to support a capital
investment. The rate includes the cost of money, depreciation, income tax, insurance and
maintenance expenses independent of energy (kWh) sold.
 Peak Responsibility Factor (PRF) is intended to compensate for the fact that the
transformer peak load losses does not necessarily occur at the same time as the system
peak. This means that only a fraction of the peak transformer losses will contribute to the
system peak demand. The value of PRF is determined by ( 5.14 ):

Transforme r load at the time o f system p eak

PRF= ( 5.14 )
Transforme r peak loa d

47
The typical values of peak responsibility factor can range from 1.0 down to 0.35. Since
PRF is a ratio of load (kVA), the losses are the function of (PRF)2. The followings are
typical values[42] of PRF.

Transformer Type PRF

Generator step-up 1.0
Transmission substation 0.9
Distribution substation 0.8
Distribution 0.35

 Equivalent Peak Ratio (EPR): The peak ratio (PR) is to relate the losses to the
rated transformer load. The peak ratio (peak pu) loading is defined by:

Peak annua l transfor mer load

PR  ( 5.15 )
Full rated transform er load

If the load grows by a given percent every year, the equivalent peak ratio (EPR) should
be used instead. EPR can be calculated from:

(1+g ) 2 t  1
Equivalent Peak Ratio ( EPR)  PR ( 5.16 )
ln(1  g ) 2 t

where: PR = the present peak ratio

g = the load growth rate (in decimal)
t = time in years

 Transformer Loss Factor (TLF) is the ratio of the average load loss to the peak
load losses over a given period of time (typically one year period).

Average lo ad loss
TLF  ( 5.17 )
Peak load loss

If the annual load data (load cycle) is available, the load losses can be calculated from
this information. However, the utility engineers try to relate the transformer loss factor
with a more readily available information called Load Factor (LF). LF is defined as the
ratio of the average load over the peak load. TLF cannot be determined uniquely from the
load factor. The utility engineers have been using an empirical formula defined by equa-
tion ( 5.18 ) for TLF in terms of LF as

Transformer Loss Factor (TLF )  a( LF ) 2 + b( LF ) ( 5.18 )

48
where: LF = Load factor

“a” and “b” are constants, such that a+b = 1.0

The constant “a” can vary from 0.8 - 0.89, whereas, “b” varies between 0.2 - 0.11. The
recommended typical values[42] of “a” and “b” are 0.84 and 0.16, respectively.

 Probability that the auxiliary cooling will be “ON” (PA) for the transformer life-
time depends on many factors such as transformer size, load profile, load growth, and
ambient temperature. The choice is a matter of engineering judgment.

5.2.3 An Example

Assume the following data to perform the loss evaluation, and to purchase a 2,000kVA,
(OA) liquid filled, 65C average winding temperature rise transformer.

Book life 30 years

System investment (SI) $1,400/kW
Discount rate (i) 10%
Energy escalation rate (e) 2%
Fixed charge rate (FCR) 0.192
Present Energy cost (EC) $0.035/kWh (3.5 cents/kWh)
Peak responsibility factor (PRF) 0.6
Peak ratio (PR) 0.7
Load growth rate (g) 2%
Load factor (LF) 0.6

Table 5.1: Loss Evaluation Calculation Example

Losses (kW) Cost of Losses Total Owning Cost

Manufacturer No-load Load Bid Price A*NLL+B*LL Equ. (4.13)
(NLL) (LL)
X 5.8 23.2 $28,000 $49,449 $77,449
Y 4.0 18 $31,000 $36,622 $67,622
Z 4.0 14 $34,000 $31,584 $65,584

Based on the above data, the following values are calculated.

AEC = $0.046 (or 4.6 cents/kWh)

EPR = 0.97
TLF = 0.4
NLCR = 3,487 $/kW, and

Table 5.1 is self-explanatory. It compares the actual bid price and the corresponding losses
with the total owning cost values for three different designs (manufacturer X, Y, and Z).

49
Design X is the standard design. Design Y and Z is the low loss design (better core ma-
terial and larger conductor size).

After the loss evaluation is taken into account, the transformer manufactured by Z is the
most cost effective over the normal operating life, even though the bid price (initial pur-
chase price) is the highest.

Purchase Price Cost of Losses

Manufacturer

Z $34,000 $31,584

Y $31,000 $36,622

X $28,000 $49,449

$0 $20,000 $40,000 $60,000 $80,000 $100,000

Figure 5.1: Total owning cost of different manufactured transformers

This example shows that the cost of losses over the transformer life is worth evaluating
and is comparable to the initial purchased price.

5.3 Engineering Economic Evaluation for Investor-Owned Utilities

Utilities are monopolistic in the sense, that they are granted a franchise to provide quality
and reliable power supply within an area. Utilities, in turn, charge a unit price allowing
only for a fair profit beyond the unit cost and have the capability to attract enough capital
to finance new projects. Amongst other responsibilities of PUC, its major duty includes
the setting of rates, so that excessive profits are eliminated, and the establishment and
maintenance of standards of service.

5.3.1 Characteristics of Investor-Owned Utilities (IOU’s)

The characteristics of investor-owned utilities as appropriate for this discussion are listed
below:

 IOU’s are capital intense whose ratio of fixed costs to variable costs is very high.

 The rates charged to customers for a utility’s services are based on the total costs,
including a fair return for the stockholders after income tax.

 A basic concept of “rates” setting is that they must be able to earn enough profit
to pay dividends and to attract the capital necessary for rendering the service.

50
  The earnings of a utility are limited by the rate base. The upper limit of profit is
usually set not to exceed about 12% - 16% on equity capital.

 Because of the stable nature of their business and earnings, utilities commonly
finance their capital expenditures with a higher percentage (50-70% range) of bor-
rowed capital. Utilities must rely on a larger proportion of new capital for expan-
sion than do other companies.

 Utilities are much less limited in term of the availability of capital than are non-
utility companies, due to their greater stability of revenues and earnings.

5.3.2 Development of the Revenue Requirement Method for Transformer Eco-

nomic Evaluation

The economic strategy widely used by regulated utilities is the “Minimum Revenue Re-
quirement Method.” [44] It calculates the revenues that a given project must collect just to
meet all the costs associated with it, including a fair return to investors. The relationship
between revenue requirements and various components of its costs is shown in Figure
5.2. At the end, the revenue requirements are minimized.

Return for
Carrying Charges (CCk)

Owners
Revenue Requirement (RRk)

Income
Tax
Total Cost

Interest on
Debt
Total Cost

Charge on

Annual Expenses

Figure 5.2: Relationship of revenue requirements and costs for an IOU

The minimum revenue requirement consists of carrying charges resulting from capital
investments that must be recovered, plus all associated expenses. It can be shown that,
the annual carrying charges in year k ( CC k ) is given by

51
 CCk  [(1   )ie  ib ]  UI k  DBk  Tk ( 5.19 )

where: CC k is annual carrying charge in year k ($)

 is debt ratio (fraction of borrowed money in total capital)
ie is equity return rate in decimal
ib is cost of borrowed money in decimal
UI k is unrecovered investment (cost of transformer) at year k
UI k = I (initial investment), k =1
UI k = UI k 1  DBk 1 , 2  k  N
N is transformer book life
D Bk is book depreciation in year k, 1  k  N
Tk is income taxes paid in year k

The depreciation calculated in this research from the “straight line method (SL)”:

I  MV
D Bk  ( 5.20 )
N

Where, MV is market value (Salvage Value) of transformer at the end of year N.

Since depreciation claimed for income tax purposes and interest paid on debt are tax de-
ductible, the income tax in any given year (k) is determined by

Tk  it (CCk  ib  UI k  DTk ) ( 5.21 )

Where, DTk is the tax depreciation in year k and it is the effective income tax rate.

it
Tk   [(1   )i e  UI k  D Bk  DTk ] ( 5.22 )
1  it

The revenue requirement in year k (RRk) (see Figure 5.2) is then:

RRk  CC k  COLk ( 5.23 )

Where, COLk is the cost of transformer losses in year k.

52
To compare between the alternatives, the “levelized” revenue requirement ( RR ) is intro-
duced. It can be found from discounting all annual revenue requirements for all years of
study period to the beginning year and multiplying by the capital recovery factor (CRF).
Mathematically,

N
 1 
RR    RRk    CRF (i) ( 5.24 )
k 1  (1  i) k 

 (1+i) N 
when CRF (i )  i  ( 5.25 )
 (1+i)  1 
N

Where, RR is levelized revenue requirement in dollars,

i is the discount rate or real (inflation-free) after-tax cost of capital.

The real (inflation-free) after tax of capital, i can be found from:

 (1  it )ib  (1   )ie  it e

i ( 5.26 )
1 e

where: i is the real (inflation-free) after-tax cost of capital, and

e is average annual inflation rate.

Alternate way to compare is the “Capitalized Revenue Requirement (CRR)”:

RR
CRR  ( 5.27 )
i

Capitalized Revenue Requirement is the same measure as the Total Owning Cost (TOC)
in loss evaluation method discussed earlier in equation ( 5.12 ).

5.3.3 Spreadsheet Illustration of the Revenue Requirement Method

The following example is discussed to demonstrate the concept:

In this example, the cost of a transformer throughout its service life is calculated. The
transformer is subjected to a constant 1.0 pu load. The transformer and utilities financial
data used in this assessment are listed below:

Transformer size = 20 MVA (Highest rating in case of a FA design)

No-load (core) loss = 30 kW
Load (winding) loss = 80 kW @ 20MVA output

53
 Transformer Service life (book life) = 30 years
Transformer and installation cost = $400,000
Market (Salvage) value, MV = $40,000 (at the end-of-life)
Demand charge = $140/kW-yr
Energy cost (present) = $0.035/kWh
Demand charge escalation rate = 0%
Energy cost escalation rate = 0%
Peak responsibility factor PRF = 0.8
Cost of no-load losses = $13,400 per year (=30*140+30*8760*0.035)
Cost of load losses = $33,500 per year (=80*0.8*140+80*8760*0.035)
Real (inflation-free) cost of borrowed money, ib = 5% per year
Real (inflation-free) return on equity, ie = 16% per year
Debt ratio,  = 0.3
Effective income tax rate, it = 50%
Book depreciation method = straight line
Average annual inflation rate, e = 0%

$400,000  $40,000
D Bk   $12,000 from equation ( 5.20 )
30

Generally, DTk  DBk  $12,000

COLk  $13,400  $33,500  $46,900

i  [0.3(1  0.5)0.05  (1  0.3)0.16  (0.3)( 0.5)( 0)] /(1  0)  0.12 Eq.( 5.26 )

 (1  0.12) 30 
CRF  0.12  =0.12414 from equation ( 5.25 )
 (1  0.12)  1
30

The result of the example is illustrated in Table 5.2. In case the transformer in the exam-
ple has to be replaced before 30 years, the unrecoverable investment less book deprecia-
tion plus market value of that year is the unpaid investment cost. It has to be paid by ad-
ditional revenue at the end of that year.

Let’s assume that the transformer from the previous example is overloaded to the end of
its life in 5 years. As a result, it has to be replaced at the end of the 5th year. Unrecovera-
ble investment (UIk) at the end of 5th year due to early replacement is $340,000
(=$352,000-$12,000). The transformer is sold at the market value (salvage) of $40,000.
Hence, the additional revenue required to cover the cost is $300,000 (=$340,000-
$40,000). So the 5th year revenue requirement has to be $443,028. Table 5.3 illustrates
this early replacement cost.

54
In summary, the revenue requirement method using the spreadsheet format is easy to fol-
low and can be applied to transformer replacement problem.

55
 Table 5.2: Spreadsheet of annual revenue requirement

Year, k UI k D Bk Tk CC k COLk RRk

1 $400,000 $12,000 $44,800 $107,600 $46,900 $154,500

2 $388,000 $12,000 $43,456 $104,732 $46,900 $151,632
3 $376,000 $12,000 $42,112 $101,864 $46,900 $148,764
4 $364,000 $12,000 $40,768 $98,996 $46,900 $145,896
5 $352,000 $12,000 $39,424 $96,128 $46,900 $143,028
6 $340,000 $12,000 $38,080 $93,260 $46,900 $140,160
7 $328,000 $12,000 $36,736 $90,392 $46,900 $137,292
8 $316,000 $12,000 $35,392 $87,524 $46,900 $134,424
9 $304,000 $12,000 $34,048 $84,656 $46,900 $131,556
10 $292,000 $12,000 $32,704 $81,788 $46,900 $128,688
11 $280,000 $12,000 $31,360 $78,920 $46,900 $125,820
12 $268,000 $12,000 $30,016 $76,052 $46,900 $122,952
13 $256,000 $12,000 $28,672 $73,184 $46,900 $120,084
14 $244,000 $12,000 $27,328 $70,316 $46,900 $117,216
15 $232,000 $12,000 $25,984 $67,448 $46,900 $114,348
16 $220,000 $12,000 $24,640 $64,580 $46,900 $111,480
17 $208,000 $12,000 $23,296 $61,712 $46,900 $108,612
18 $196,000 $12,000 $21,952 $58,844 $46,900 $105,744
19 $184,000 $12,000 $20,608 $55,976 $46,900 $102,876
20 $172,000 $12,000 $19,264 $53,108 $46,900 $100,008
21 $160,000 $12,000 $17,920 $50,240 $46,900 $97,140
22 $148,000 $12,000 $16,576 $47,372 $46,900 $94,272
23 $136,000 $12,000 $15,232 $44,504 $46,900 $91,404
24 $124,000 $12,000 $13,888 $41,636 $46,900 $88,536
25 $112,000 $12,000 $12,544 $38,768 $46,900 $85,668
26 $100,000 $12,000 $11,200 $35,900 $46,900 $82,800
27 $88,000 $12,000 $9,856 $33,032 $46,900 $79,932
28 $64,000 $12,000 $8,512 $30,164 $46,900 $77,064
29 $52,000 $12,000 $7,168 $27,296 $46,900 $74,196
30 $40,000 $12,000 $5,824 $24,428 $46,900 $71,328

RR  $133,567
CCR = $1,113,058

56
 Table 5.3: Spreadsheet of annual revenue requirement with early replacement
Year UI k D Bk Tk CC k COLk RRk

1 $400,000 $12,000 $44,800 $107,600 $46,900 $154,500

2 $388,000 $12,000 $43,456 $104,732 $46,900 $151,632
3 $376,000 $12,000 $42,112 $101,864 $46,900 $148,764
4 $364,000 $12,000 $40,768 $98,996 $46,900 $145,896
5 $352,000 $12,000 $39,424 $96,128 $46,900 $143,028
+$352,000
-$12,000
-$40,000
=$443,028
6 New transformer’s data is used for calculation beginning from year 6th
7
8
9
10

57
6. Optimization Strategy for Transformer Procurement and Re-
placement
6.1 Introduction

Sizing of power transformers or replacement of existing transformers is done intuitively

from simplified technical and/or economical decisions. Combining both technical and
economic considerations, however, is more desirable. In this research, a new method is
proposed for utilities to be able to maximize the optimal use of their existing transfor-
mers, to select the appropriate transformer size and to determine the replacement strategy.
A number of situations in real world deal with transformer procurement and replacement,
such as, (i) select a new transformer for a new project (new design), (ii) delay transformer
replacement until its life ends (delay replacement option), and (iii) replace the existing
transformer immediately with a larger unit (replace now option). The life of the transfor-
mer for all problems depends on its size, loading and load growth rate. The delay re-
placement option defers new investment, however, with the sacrifice of transformer’s life
and higher cost of losses.

In this chapter, a new optimization scheme is proposed, that utilizes the probability tree
load-growth structure to take care of overloading risks and replacement timing. Only the
IEEE Clause 7 thermal (simplified) model from Chapter 3 is applied here due to limita-
tion of available transformer data. However, Annex G model can easily be used provided
there are available transformer data. The cost of random failure of a transformer in-
service is also included in the scheme. It is assumed that transformer cost and losses can
be formulated into some continuous function depending upon transformer sizes. Howev-
er, the actual cost and loss data, if available, can also be used with some modification.
The cost and loss data are derived from forced-air cooled (FA) transformers. For other
cooling types, similar data could be added and left for future work. The thermal model
and hottest-spot temperature (Chapter 3) determine the remaining life of the transformer.

6.2 Transformer Cost and Losses Function

For oil-cooled (OA/FA/FA) transformers utilized in large distribution substations ranging

between, say 10-100 MVA (the primary voltage of 69 kV, 115 kV or 230 kV, and the
secondary voltage rated at 12.47 kV to 34.5 kV), the cost depends on a number of factors.
The key factors are: the HV side voltage and the corresponding Basic Insulation Level
(BIL) and the losses. The data given in Table 6.1 is for typical (OA/FA/FA) transformers.
The installation cost of $595/MVA [45] including labor and equipment is also added to the
transformer cost. A curve-fitting program utilizes a quadratic function ( f ( S T ) ) to esti-
mate the cost of transformers. These data points are plotted in Figure 6.1 and the trans-
former cost is approximated as:

f ( S T )  K 2 S T2  K 1 S T  K 0 ($/MVA) ( 6.1 )

58
  8.453ST2  862.9ST  33164

Where, K0, K1, and K2 are constants and ST is the transformer size (highest rating) in
MVA (65C average winding temperature rise and 110C hottest-spot temperature).

The 65C ratings in Table 6.1 are calculated using common multiplying factors. The typi-
cal multiplying factor to convert MVA rating at 55C winding rise to MVA rating at 65C
winding rise are given in
Table 6.2. It is clear from

Figure 5.1 that the transformer cost function ($/MVA) decreases as transformer rating
increases and reaches approximately a constant value at around 50MVA and higher rat-
ing.
Table 6.1: Cost of typical forced-air cooled (OA/FA/FA) power transformer
ST, MVA Transformer Cost
 
55 C 65 C Transformer, $ Installation, $ $/MVA*
10 11 277,000 6,500 24,652
20 23 395,000 13,700 17,770
25 28.75 429,000 17,100 15,516
40 46 486,000 27,400 11,160
50 57.5 605,000 34,200 11,116
* $/MVA is based on 65C winding rise rating.

Table 6.2: The multiplying factor for MVA rating conversion from 55C winding rise to
65C winding rise rating
Cooling Multiplying
Type Factor
OA 1.165
FA 1.150
NDFOA 1.140
FOA 1.125

Note: Tables 6.1 and 6.2 are calculated based on typical top-oil and hottest-spot tempera-
ture rise and a loss ratio of approximately 3.0.

59
 30000

25000

20000

Cost, $/MVA
15000

10000

5000

0
0 10 20 30 40 50 60 70
Transformer Size, MVA

Figure 6.1: Transformer cost data

No-load and load losses also vary by transformer sizes and designs. Since these values
are extremely important quantities, it is recommended that actual values from the manu-
facturers be used. Typical losses in kW/MVA are tabulated in

Table 6.3. The quadratic equations are developed to fit the curve and the no-load and load
losses in kW/MVA can be found from equation ( 6.2 ) and ( 6.3 ), respectively.

f NLL ( S T )  (6.216  10 5 ) S T2  0.01219 S T  1.523 ( 6.2 )

f LL ( S T )  0.0002657ST2  0.0524S T  4.969 ( 6.3 )

Table 6.3: Transformer losses

ST, MVA No-load losses Load losses Loss Ratio

55 C 
65 C kW **
kW/MVA *
kW **
kW/MVA * (R)
10 11.5 17 1.48 53 4.61 3.12
20 23 29 1.26 86 3.74 2.97
30 34.5 40 1.15 115 3.33 2.88
40 46 50 1.09 141 3.06 2.82
50 57.5 60 1.04 165 2.87 2.75
60 69 70 1.01 188 2.72 2.69
80 92 87 0.95 232 2.52 2.67
100 115 106 0.92 272 2.36 2.57

* kW/MVA is based on 65 C average winding rise rating.
** kW is based on 65C average winding rise rating.

60
 5

4
Losses, kW/MVA

3
Load Losses

No-load Losses
1

0
0 20 40 60 80 100 120
Transformer Size, MVA

Figure 6.2: Transformer losses data

6.3 Random Failure of Transformer In-Service

Transformers may fail at random during its normal life, and it costs utilities to purchase a
new unit. A typical hazard function, h(t), describing this random failure, also known as
“bathtub curve” [46], is plotted in Figure 6.3.

Figure 6.3: “Bathtub curve”

It is believed that the distribution of failure, f(t), during the normal operating period is an
exponential decay and the hazard function is constant, as derived by[46].

f (t )
h (t ) 
1  F (t )

61
 f (t )  e (  t )

t
F (t )   f (t )dt  1  e ( t )
0

e (  t )
h(t )  
1  (1  e (  t ) )

where: f(t) is exponential distribution function

F(t) is exponential cumulative distribution function
 is the failure rate, and
t is the time.

When transformers are subjected to a random failure rate of  , utilities need additional
revenue to cover the cost of transformer failure. It can be formulated as:

CFk    C TR  ( N exp  k ) / N exp ; for 1  t  N exp ( 6.4 )

where: CFk is the cost of failure in year k

C TR is the cost of transformer in dollars
N exp is expected transformer life in year, and
k is the year that transformer fails

The failure rate of the transformer is reported[47] to be about 0.2-0.5%.

6.4 Expected Transformer Life

It has been discussed earlier that, in this research the normal transformer life (1 pu) of
150,000 hours (17.12 years) is used. This corresponds to the insulation’s tensile strength
reduction to 20% of its original value or the insulation’s degree of polymerization reduc-
tion to 200. Based on these criteria, the remaining transformer life expectancy for the
cyclic load is calculated. A computer program is written to calculate the loss-of-
insulation life of the transformer subjected to cyclic load. Typical monthly load profiles
and ambient temperature profiles as shown in Appendix B are used. In a real power sys-
tem, the actual load profiles may be obtained from SCADA. The actual ambient tempera-
ture profile can be obtained from the weather station data at the transformer location or
installed on-line monitoring devices. Table 6.4. is from the test run for a 52.26 MVA
forced-air cooled (OA/FA/FA) transformer with typical load and ambient temperature
profiles. The transformer data used in this test case is obtained from reference [1]. The
annual peak load in the month of August is assumed to be 1.1 per-unit. The assumed Au-
gust load and ambient temperature profile is repeated again in Figure 6.4. The load is as-
62
sumed to grow at a constant rate of 2.5% compounded annually. The thermal model used
corresponds to the IEEE Annex G. (or the Detailed Model).

August Load and Ambient Temperature Profile

1.1 40 Loading
Ambient
1

Ambient Temperature (C)

30
Loading (Per Unit)

0.9 20

0.8
10

0.7
0

0.6
-10
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hr)

Figure 6.4: Load and ambient temperature profiles in the month of August

63
 Table 6.4: Illustrated life cycle study printout of transformer life

Transformer Thermal Loading Program

Life Cycle Study of Transformer using IEEE Annex G Thermal Model

Year 2001
% Loss Remain Remain Max. Max. Acc. Used
of Life Tensile DP Hot Spot Top Oil %LOF Life(yrs)
January 0.000 97.05 1400 59.9 42.8 0.000 0.00
February 0.001 97.05 1400 68.7 50.2 0.001 0.00
March 0.002 97.04 1400 74.5 53.9 0.004 0.00
April 0.002 97.04 1400 75.4 54.8 0.006 0.00
May 0.007 97.03 1399 84.5 61.7 0.013 0.00
June 0.035 96.98 1397 99.0 74.5 0.048 0.01
July 0.239 96.61 1382 118.2 89.0 0.287 0.05
August 0.292 96.17 1365 120.2 91.1 0.579 0.10
Sept. 0.033 96.12 1363 98.5 74.0 0.612 0.10
October 0.010 96.10 1362 87.5 64.8 0.621 0.11
Nov. 0.005 96.09 1362 82.3 59.5 0.627 0.11
Dec. 0.002 96.09 1362 72.3 51.6 0.628 0.11

Maximum Loading= 1.100 Used Life= 0.11 years(942 hrs)

----------------------------------------------------------------------
Year 2012
% Loss Remain Remain Max. Max. Acc. Used
of Life Tensile DP Hot Spot Top Oil %LOF Life(yrs)
January 0.007 28.94 261 85.0 59.1 76.589 13.11
February 0.025 28.93 261 97.3 68.7 76.614 13.12
March 0.074 28.89 261 107.4 75.1 76.688 13.13
April 0.080 28.86 260 108.3 76.1 76.767 13.15
May 0.293 28.72 259 121.7 85.9 77.060 13.20
June 1.602 28.00 255 140.1 101.8 78.663 13.47
July 16.032 21.74 212 168.5 123.3 94.695 16.21
August 18.911 16.12 171 170.6 125.5 113.605 19.45
Sept. 1.519 15.74 168 139.6 101.2 115.124 19.71
October 0.384 15.65 168 124.7 89.1 115.508 19.78
Nov. 0.223 15.59 167 119.4 83.6 115.731 19.82
Dec. 0.058 15.58 167 105.1 72.8 115.789 19.83

Maximum Loading= 1.443 Used Life= 19.83 years(173683 hrs)

----------------------------------------------------------------------

The transformer used, in the first year of service, is just 0.11 years of its normal life
(17.12 years). The transformer lost most of its life in July and August due to higher load
64
level and higher ambient temperature. The load shape is assumed to be unchanged in our
study for simplification. It is then multiplied by the load growth factor (1+load growth
rate in decimal) for each consecutive year. Any daily and seasonal variations can be used
to calculate the loss-of-life. The load growth increases the peak load and the loss-of-life
sharply especially at later years before transformer life ends. The transformer life is con-
sidered to reach the end when remaining tensile strength reaches 20% of its original value
or when the degree of polymerization reaches 200. The initial degree of polymerization
of the new insulation is assumed to be 1400. The equations for calculation of remaining
tensile strength and degree of polymerization are given in Chapter 2. Figure 6.5-Figure
6.8 are the plots from the program.

Lo s s o f Tra ns fo rm er Life (in yea rs )

20 A nnual LOF
A c c . LOF
18

16 Life (in years)

4.37
14
Los s of Lif e (y rs )

12

10

8
6.71

4.37
4
2.88
1.92
2
1.29
0.88
0.42 0.6
0.11 0.15 0.21 0.29
0
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Y ear

Figure 6.5: Loss of transformer life vs. years in service

65
 H ot Spot and Top Oil Tem perature

170 Hot Spot

Top Oil
160

150

140

130
Temperature (Degree C)

120

110

100

90

80

70

60

50

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Y ear

Figure 6.6: Hottest-spot and top-oil temperature vs. years in service

R em aining Tens ile Strength

Remain TS
90 18 A c c . LOF

16
80

14
70
Remaining Tens ile Strength

A c c . Los s of Lif e (y rs )

12

60
10

50
8

40 6

4
30

2
20

0
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Y ear

Figure 6.7: Insulation’s remaining tensile strength vs. years in service

66
 R em aining D egree of Polym erization (D P)
1,400
Remain DP
1,300 18 A c c . LOF

1,200
16
1,100
14
1,000

A c c . Los s of Lif e (y rs )
900 12
Remaing DP

800 10

700
8
600
6
500

4
400

300 2

200
0
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Y ear

Figure 6.8: Insulation’s remain degree of polymerization vs. years in service

It is clear from Table 6.4 that in this example, the expected transformer life is approx-
imately 11 years and 7 months. This shorter service life results from the 2.5% annual load
growth and high initial annual peak load in the month of August. The top-oil and hottest-
spot temperature exceeds limits of 110C and 140C respectively.

6.5 Load Growth Uncertainty Modeling

Transformer life expectancy is closely related to its present load and the future load
growth rate. In the previous example, a fixed load growth of 2.5% (compounded annual-
ly) is used. As shown in Table 6.4, in the first year of service, the transformer just used
0.11 years from its normal life expectancy. However, it consumed 6.71 years of life in the
last year of operation. Actual transformer loading and the prediction of the future load
growth rate has a lot of uncertainties. In this research, the probability tree method is uti-
lized to take into account these uncertainties. An alternate approach is to utilize the tech-
nique of Monte Carlo simulations. Probability tree method is simple to follow, however,
computationally expensive.

67
 Figure 6.9: Approximating load uncertainty using probability tree

Figure 6.9 shows a construction of a probability tree from the probability curve. The
smooth probability curve, in this example, has been discretized into three pieces. The ex-
pected load (20 MVA) is assumed to have a probability of 0.6. Both the higher (21 MVA)
and the lower (19.6 MVA) loads have a probability of 0.2. The number in the circle in
Figure 6.9(b) is the probability of the load that follows the value. All loadings are se-
lected randomly.

Figure 6.10 illustrates an example of the probability tree beginning. The tree grows by a
power of 3 for every year. As such when the time progresses by 4 years, the load can
grow into 81 (=34) different paths. The assessment of load growth rate beyond 81 differ-
ent paths can be tedious and repetitive. In order to simplify the problem, the probability
tree is limited to 4 years of study. After the 4th year the load growth rate is assumed to be
constant.

68
 Figure 6.10: Probability tree representation of load uncertainty

Figure 6.11 shows load growth paths. Path 1 is the highest possible growth and the prob-
ability of path 1 is the product of all the probabilities along its path from year 1st to 4th ,
which is

Pr,1 = (0.2)(0.2)(0.2)(0.2) = 0.0016

In the same way, the probability of path 5 is

Pr,5 = (0.2)(0.2)(0.6)(0.6) = 0.0144

The summation of all probabilities of all paths is equal to 1.0, i.e.,

81
Mathematically, P
i 1
r, i  1 .0

69
 Figure 6.11: Illustration of load growth path

Beginning 5th year, the load growth rate is assumed to be a constant value of . Mathe-
matically, load at year k (Lk) is

Lk  L4 (1   ) k  4 ( 6.5 )

Where, k is year number ( k  5 ), L4 is the load at the end of probability tree table (4th
year), and  is the load growth rate in decimal.

Figure 6.12 illustrates the fact that the degree of uncertainty widens as the year
progresses. The curve on year 2001 is more vertical because of better prediction of uncer-
tainty (less number of paths). The high and low predicted load is close to the expected
value. The compounded growth on high and low branches (Figure 6.9(b)) of probability
tree spread out the high and low boundary of the projected load. Figure 6.13 shows the
projection of load uncertainty for a long period. Figure 6.13 displays probability distribu-
tion of load, whereas, Figure 6.12 shows cumulative distribution of load.

70
 Figure 6.12: Cumulative probability of transformer load

Figure 6.13: Approximating load uncertainty in each future year

With 81 different load growth paths, the levelized revenue requirements (discussed in
Chapter 5) are calculated for each path. Then, the equivalent levelized revenue require-

71
ment (ERR) is calculated by weighting levelized revenue requirement with path probabil-
ity as

81
ERR   Pr , p  RR p ( 6.6 )
p 1

Where, p is path number, Pr,p is probability of path p, and RR p is levelized revenue re-
quirement of path p.

Along with each load growth path, a new transformer is required when existing transfor-
mer has expended its useful life to the thermal loading. This will add additional invest-
ment for the utilities at some future year as predicted by the thermal loading described in
Chapter 3.

6.6 Solution Methods

A program is written to solve the optimization problem, where the objective function is
the minimization of the equivalent levelized revenue requirement (ERR):

81
min ERR  min  Pr , p  RR p
p 1

The block diagram of the complete solution is shown in Figure 6.14. The diagram in-
cludes the combination of transformer model and utilities financial model. The program
uses the IEEE classical model (Clause 7). By using information from transformer manu-
facturers and utilities, the transformer cost and losses are calculated from equations ( 6.1
), ( 6.2 ), and ( 6.3 ). The program assumes that the transformer size can vary continuous-
ly. One MVA incremental step is used. In order to simplify the computation, the histori-
cal load and ambient temperature profiles have been applied on monthly basis. Twelve
load profiles and twelve ambient temperature profiles (one for each month) are stored in
the database. The annual peak load has been generated from the probability tree structure
that derived from short-term (4 years) load forecast.

The transformer thermal model calculates remaining transformer life by annual cyclic
loading. If the transformer reaches the end of its life (150,000 hrs.), the new transformer
has to be installed. Criterion for the new transformer size is based on the study method.
For a new design and replace now option, the new transformer size is set to have the
same initial annual per unit peak load of the existing transformer when it was installed.
For the delay replacement option, the existing transformer has been used until the end of
its life and then replaced with different size of a new transformer ranging from 0.5 - 1.0
per unit peak loading of its first year of operation. In the case of delay replacement study,
the remaining life of the existing transformer has to be pre-calculated by backward ther-
mal loading available in the program or direct measurement from actual insulation re-
72
maining strength. The transformer thermal model also calculates the annual cost of losses
of the transformer as:

COLk  CNLLk  CLLk

8760
CNLLk  DCk  Pcore  ECk  P
t 0
core dt

8760
CLLk  DCk  PRF  Pcu,max  ECk P
t 0
cu (t )dt

where: COLk , CNLLk , and CLLk are the cost of losses, cost of no-load losses,
and cost of load losses in year, k respectively
DCk is demand charge, $/kW
EC k is cost of energy, $/kWh
Pcore is core or no-load losses, kW
Pcu (t ) are copper or load losses, kW
Pcu , max are the annual peak load losses
PRF is peak responsibility factor (See Chapter 4)

The cost of transformer random failure is calculated from equation ( 6.4 ). The utilities
financial model uses the following financial data as an input to calculate carrying charge,
tax and depreciation (Chapter 5).

Equity return rate ( ie )

Borrowed money rate ( ib )

Debt ratio ()

Tax rate (it)

Depreciation method (only “Straight Line” method is used in this report.)

Accounting book life of transformer is assumed (30 years).

The revenue requirement for year k can then be written as:

RRk  CCk  COLk  CFk

73
The program calculates revenue requirement of each load growth path. Each load growth
path is 30 years in length. The levelized revenue requirement can be found by using equa-
tion (5.24). The final equivalent levelized revenue requirement is then calculated from the
probability weighting on levelized revenue requirement for 81 different paths.

The various transformer sizes are generated by computer program. The program calcu-
lates each equivalent levelized revenue requirement for each transformer. The minimum
equivalent levelized revenue requirement determines the best alternative.

74
 Probability Tree Structure Path Levelized Probability

1 $134,000 0.0016
Probability
Probability

81 $101,000 0.0016

Equivalent Levelized
Energy and Demand Charge
Escalation Rate
Economic Data

Ambient Temperature and Cost of Losses

Historical Estimated Remaining Life Transformer Thermal Utility’s
Model Cost of Transformer Financial
Model
Cost of Failure
Load
End-of-Life Debt Ratio

New Transformer
Equity Return

Random Failure Model

Thermal Data, Electrical Losses Borrowed Money Rate
Transformer Data
Tax Rate

Figure 6.14: Integrated structure of insulation degradation based transformer utilization mode
75
7. Simulations and Case Studies
7.1 Introduction

The program incorporated thermal model, loss-of-life calculation, failure cost, replacement crite-
rion, and economic evaluation. Two case studies are discussed and one possible scenario: the
medium load growth rate of 1.75% /year with +0.25% and –0.75% deviation. After the 4th year, a
constant load growth rate of 2%/year has been assumed.

 Case #1, a transformer has to be purchased for a new project. The existing load at the be-
ginning is estimated at 20MVA. The growth rate is medium at 1.75% per year. A study period of
30 years has been used in the simulations.

 Case #2, the existing transformer size is assumed to be 18 MVA (highest rating). It is as-
sumed to be in service for 25 yrs. with an estimated remaining life of 25% (or 0.25pu.). The load
growth rate is assumed at 1.75%. There are two choices: (i) Continue overload the transformer,
until the life is completely utilized and then replace appropriately and optimally sized to accom-
modate the load growth. The end-of-life criteria used are 20% RTS or remaining DP of 200. The
program evaluates the size. It is assumed that the load on the new transformer when it is com-
missioned is 0.5, 0.6, 0.7, 0.8, 0.9, and 1.0 per-unit load. As an example, if the existing transfor-
mer’s life ends at the end of year 5th and the next year (6th) load is 22.46 MVA, with the sizing
criteria of 0.5, the size of the new transformer is selected as 44.92 MVA (= 22.46/0.5).

7.2 Solutions to the Problems

Results of the two cases are presented here, utilizing similar transformer design (except for the
MVA output), economic model, and financial data are tabulated in Table 7.1. The MVA rating
corresponds to 65C average winding temperature rise.

76
 Table 7.1: Common transformer, economic, and financial data for all case studies

Top-oil rise over ambient at rated load = 45.0C

Hottest-spot rise over top oil at rated load = 35.0C
Oil time constant = 1.25 hours
Winding time constant = 5.0 minutes
Cooling mode is FA

Energy cost = 0.035 $/kWh

Energy escalation rate = 2.0%
Demand charge = 120 $/kW-yr
Demand charge escalation rate = 2.0%
Peak Responsibility Factor (PRF) = 0.8

Random failure rate = 0.5%

Return on equity rate = 16.0%
Borrowed money rate = 5.0%
Debt ratio = 0.5
Tax rate = 50%
Market (Salvage) value = 10% of investment cost
Inflation rate = 0%
Transformer book life = 30 years

Existing transformer size = 18MVA

Year in service = 25 years
Remaining life = 0.25 per unit
Existing transformer price = $340,000
Market value of existing transformer price = $34,000

7.2.1 Case #1, Sizing of a New Transformer with Moderate Load Growth

In this case, a new transformer is to be installed for a project. The load growth rate is assumed
moderate. The first four years of load growth is constructed using the probability tree with the
high, expected, and low load growth rate of 2.0, 1.75, and 1.0 %, respectively. Based on the load-
ing information, a transformer size between 20 to 40 MVA is considered. Table 7.2 lists Equiva-
lent Revenue Requirement (ERR) for each size. The ERR reaches minimum of $114,625 for the
transformer size of 25 MVA. Accordingly, 25 MVA is the most economical size for this project.
The first year’s annual peak load of this new transformer is 0.81 per unit (= 20.35/25).

Table 7.3 shows the detailed result and the economic evaluation of a 25 MVA transformer on the
most probable load growth path #41 from the life cycle study performed on this transformer.
With this load growth pattern of path #41, the 25MVA transformer can last approximately 30
years. The simulation result is shown in Table 7.4. Figure 7.1 shows the temperature profiles.
77
The maximum hottest-spot and top-oil temperature at 30th year is 167.8C, 105.7C respectively.
The hottest-spot temperature is too high and may cause bubble generation and dielectric strength
reduction during the overload period.

Table 7.2: Result of Case #1 Study

Sizing Program on new application only

Top-oil rise over ambient at rated load = 45.0 C

Hottest-spot rise over top oil at rated load = 35.0 C
Oil time constant = 1.25 hours
Winding time constant = 5.0 minutes
Cooling mode is FA
---------------------------------
Equivalent Revenue Requirement(ERR)

Size ERR
20.00 $120,426
21.00 $118,901
22.00 $117,486
23.00 $116,327
24.00 $115,375
25.00 $114,625
26.00 $114,827
27.00 $115,050
28.00 $115,294
29.00 $115,558
30.00 $115,842
31.00 $116,148
32.00 $116,480
33.00 $116,840
34.00 $117,234
35.00 $117,666
36.00 $118,140
37.00 $118,663
38.00 $119,241
39.00 $119,879
40.00 $120,584

78
 Table 7.3: Result of a 25 MVA transformer on load growth path #41, case #1 study

25.00MVA Path No.41

Year MVA UIk Tk CCk COLk CFk RRk

1 25.00 421866 33749 90701 28089 0 118791
2 25.00 409210 32737 88360 29180 1969 119508
3 25.00 396554 31724 86018 30321 1898 118238
4 25.00 383898 30712 83677 31518 1828 117023
5 25.00 371242 29699 81336 32861 1758 115954
6 25.00 358586 28687 78994 34274 1687 114956
7 25.00 345930 27674 76653 35762 1617 114032
8 25.00 333274 26662 74312 37329 1547 113187
9 25.00 320618 25649 71970 38979 1477 112426
10 25.00 307962 24637 69629 40717 1406 111752
11 25.00 295306 23624 67288 42549 1336 111173
12 25.00 282650 22612 64946 44480 1266 110692
13 25.00 269994 21600 62605 46516 1195 110316
14 25.00 257338 20587 60264 48662 1125 110051
15 25.00 244682 19575 57922 50926 1055 109903
16 25.00 232026 18562 55581 53314 984 109879
17 25.00 219370 17550 53239 55834 914 109987
18 25.00 206714 16537 50898 58493 844 110235
19 25.00 194058 15525 48557 61299 773 110630
20 25.00 181402 14512 46215 64262 703 111181
21 25.00 168746 13500 43874 67391 633 111898
22 25.00 156090 12487 41533 70695 562 112790
23 25.00 143434 11475 39191 74185 492 113868
24 25.00 130778 10462 36850 77871 422 115143
25 25.00 118122 9450 34509 81766 352 116626
26 25.00 105466 8437 32167 85882 281 118331
27 25.00 92810 7425 29826 90233 211 120270
28 25.00 80154 6412 27485 94831 141 122456
29 25.00 67499 5400 25143 99693 70 124906
30 25.00 54843 4387 22802 104833 0 127635

Levelized Revenue Requirement = $114,817

79
 Table 7.4: Life cycle study result of case #1’s 25 MVA transformer (path #41)

Transformer Thermal Loading Program

Life Cycle Study of Transformer using IEEE Clause 7 Thermal model

Year 2001

% Loss Remain Remain Max. Max. Acc. Used

of Life Tensile DP Hot Spot Top Oil %LOF Life(yrs)
January 0.000 97.05 1400 45.3 31.1 0.000 0.00
February 0.000 97.05 1400 52.4 36.7 0.000 0.00
March 0.000 97.05 1400 55.8 38.2 0.000 0.00
April 0.000 97.05 1400 56.6 39.0 0.001 0.00
May 0.001 97.05 1400 63.0 43.5 0.001 0.00
June 0.002 97.04 1400 75.0 53.8 0.003 0.00
July 0.011 97.03 1399 88.3 63.1 0.014 0.00
August 0.013 97.01 1398 90.3 65.1 0.027 0.00
September 0.002 97.01 1398 74.6 53.3 0.029 0.01
October 0.001 97.00 1398 66.4 46.9 0.030 0.01
November 0.000 97.00 1398 61.1 41.6 0.030 0.01
December 0.000 97.00 1398 53.7 36.1 0.030 0.01

Maximum Loading= 0.814 Used Life= 0.01 years(46 hrs)

----------------------------------------------------------------------

Year 2030

% Loss Remain Remain Max. Max. Acc. Used

of Life Tensile DP Hot Spot Top Oil %LOF Life(yrs)
January 0.008 23.95 228 87.1 51.9 88.567 15.17
February 0.027 23.94 228 99.1 60.2 88.594 15.17
March 0.077 23.91 227 108.6 65.1 88.671 15.18
April 0.082 23.88 227 109.4 65.9 88.752 15.20
May 0.282 23.77 226 122.0 73.7 89.035 15.25
June 1.437 23.24 223 139.5 87.1 90.472 15.49
July 12.502 19.07 193 165.8 103.7 102.974 17.63
August 14.647 15.13 164 167.8 105.7 117.621 20.14
September 1.373 14.81 161 139.1 86.7 118.993 20.38
October 0.378 14.72 161 125.4 77.1 119.371 20.44
November 0.221 14.67 160 120.1 71.8 119.592 20.48
December 0.060 14.65 160 106.5 63.0 119.652 20.49

Maximum Loading= 1.435 Used Life= 20.49 years(179478 hrs)

----------------------------------------------------------------------

80
 Hot Spot and Top Oil Temperature

Hot Spot
Top Oil
160

150

140

130
Temperature (Degree C)

120

110

100

90

80

70

60

50

40

2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030
Year

Figure 7.1: Hottest-spot and top-oil temperature of case #1’s 25 MVA transformer (Path #41)

7.2.2 Case #2, Transformer Replacement with Moderate Load Growth Rate

The first four years of load growth is constructed using the spreadsheet and the probability tree
with the high, expected, and low load growth rate of 2.0, 1.75, and 1.0 %, respectively. The
choices compared are: (i) “replace now” or (ii) “delay replacement” at the end of the existing
transformer’s life. The transformer size ranges from 20-40MVA. Table 7.5 shows the result. The
Equivalent Revenue Requirement (ERR) reaches a minimum of $91,327 in the “delay replace-
ment” option with 0.8 pu peak load criteria. In this example, the delay replacement option is
economically favorable. The existing transformer is able to be in operation for 9 yrs. before its
life ends. The result is shown in Table 7.7. When the transformer’s life ends at 10th year, the
30.18MVA unit replaces the existing 18 MVA transformer. The load at that time is 24.1MVA.
This corresponds to 0.8 pu peak load criterion (= 24.1/30.18). The “replace now” option is eco-
nomically more expensive compared to the “delay replacement” option. It can be seen that the
maximum hottest-spot and top-oil temp. at 9th year is 150.6C, 96.5C respectively, which is too
high with a considerable amount of loss-of-life. The “replace now” option gives the minimum
ERR of $125,706 for a 28 MVA transformer, compared to the “delay replacement” of $91,327.
The utility can save $34,379 per year by choosing to overload the existing transformer.

81
 Table 7.5: Result of case #2 study

Transformer Delay Replacement Program

Top-oil rise over ambient at rated load = 45.0 C

Hottest-spot rise over top oil at rated load = 35.0 C
Oil time constant = 1.25 hours
Winding time constant = 5.0 minutes
Cooling mode is FA
---------------------------------

Equivalent Revenue Requirement(ERR) for Replace Now Option

Size ERR
20.00 $134,345
21.00 $132,260
22.00 $130,365
23.00 $128,753
24.00 $127,327
25.00 $126,051
26.00 $125,852
27.00 $125,737
28.00 $125,706
29.00 $125,762
30.00 $125,909
31.00 $126,150
32.00 $126,493
33.00 $126,945
34.00 $127,513
35.00 $128,205
36.00 $129,031
37.00 $129,999
38.00 $131,119
39.00 $132,402
40.00 $133,858

Equivalent Revenue Requirement(ERR) for Delay Replacement Option

Peak Load ERR

0.50 $101,798
0.60 $93,795
0.70 $91,569
0.80 $91,327
0.90 $91,858
1.00 $94,336

82
 Table 7.6: Detailed result of delay replacement with 0.8 per-unit peak load criterion on load
growth path #41, case #2 study

18.00MVA Path No.41

Year MVA UIk Tk CCk COLk CFk RRk

1 18.00 85000 6800 25925 35888 227 62040
2 18.00 74800 5984 24038 37488 170 61696
3 18.00 64600 5168 22151 39169 113 61433
4 18.00 54400 4352 20264 40937 57 61258
5 18.00 44200 3536 18377 42945 0 61322
6 18.00 34000 2720 16490 45066 0 61556
7 18.00 23800 1904 14603 47307 0 61910
8 18.00 13600 1088 12716 49674 0 62390
9 18.00 3400 272 10829 52176 0 29005
10 30.18 463233 37059 99595 40242 0 139837
11 30.18 449336 35947 97024 41882 2162 141068
12 30.18 435439 34835 94453 43606 2085 140144
13 30.18 421542 33723 91882 45419 2007 139308
14 30.18 407645 32612 89311 47326 1930 138567
15 30.18 393748 31500 86740 49331 1853 137925
16 30.18 379851 30388 84169 51442 1776 137387
17 30.18 365954 29276 81598 53664 1699 136961
18 30.18 352057 28165 79027 56003 1621 136652
19 30.18 338160 27053 76457 58467 1544 136468
20 30.18 324263 25941 73886 61062 1467 136414
21 30.18 310366 24829 71315 63796 1390 136500
22 30.18 296469 23718 68744 66678 1312 136734
23 30.18 282572 22606 66173 69715 1235 137123
24 30.18 268675 21494 63602 72918 1158 137677
25 30.18 254778 20382 61031 76295 1081 138406
26 30.18 240881 19270 58460 79857 1004 139320
27 30.18 226984 18159 55889 83615 926 140430
28 30.18 213087 17047 53318 87580 849 141748
29 30.18 199190 15935 50747 91766 772 143285
30 30.18 185293 14823 48176 96183 695 145054

Levelized Revenue Requirement = $91,772

7.3 Summary

Simulation results for two case studies are presented here. Case studies: #1 relate to the problem
of sizing a new transformer, whereas, #2 deals with the problem of an existing system and to
evaluate the “delay replacement” option of an existing 18 MVA power transformer. .Both cases
#1 and #2 assume moderate (1.75%) load growth.

Load growth rate is a very important factor that needs to be considered in transformer sizing. The
analyses and results yield that the most economical transformer’s sizes for Case #1 (moderate
load growth) is 25MVA.

83
In Case #2 (moderate load growth), the rating of the existing transformer is assumed to be 18
MVA with 0.25 pu remaining life. The system is experiencing overload and utility engineers are
trying to look into both “replace now” or “delay replacement” option. In both cases, the most
economical choice goes to “delay replacement” option. In Case #2, the delay replacement option
yields the lower ERR of $91,327 when the new transformer with the initial loading at 0.8 per-
unit is selected. The replace now option gives the much higher ERR of $125,796 for a new 28
MVA transformer. The existing transformer can be allowed to overload for the next 9 years until
it needs replacement. The “delay replacement” clearly saves utility $34,379 annually.

These case studies utilize typical load profiles, ambient temperature, economic and financial da-
ta. However, the optimization techniques presented in this report can be applied to any real situa-
tions.

Table 7.7: Life cycle study result case #2’s existing 18 MVA transformer (path #41)

Program Transformer Overloading Version 1.0

Life Cycle Study of Transformer using IEEE Clause 7 Thermal model

Year 2001

% Loss Remain Remain Max. Max. Acc. Used

of Life Tensile DP Hot Spot Top Oil %LOF Life(yrs)
January 0.001 29.67 266 64.4 40.4 75.001 12.84
February 0.002 29.67 266 73.8 47.2 75.002 12.84
March 0.004 29.67 266 80.0 50.3 75.006 12.84
April 0.004 29.67 266 80.8 51.1 75.010 12.84
May 0.011 29.66 266 90.1 57.1 75.021 12.85
June 0.056 29.64 265 104.6 68.8 75.077 12.86
July 0.374 29.46 264 123.9 81.5 75.450 12.92
August 0.453 29.25 263 125.9 83.5 75.903 13.00
September 0.053 29.23 263 104.2 68.4 75.955 13.01
October 0.016 29.22 263 93.5 60.4 75.971 13.01
November 0.009 29.22 263 88.2 55.2 75.980 13.01
December 0.003 29.22 263 77.9 48.2 75.983 13.01

Maximum Loading= 1.131 Used Life= 13.01 years(113974 hrs)

----------------------------------------------------------------------
Year 2009

% Loss Remain Remain Max. Max. Acc. Used

of Life Tensile DP Hot Spot Top Oil %LOF Life(yrs)
January 0.003 20.90 206 77.7 47.1 97.186 16.64
February 0.009 20.90 206 88.6 54.7 97.194 16.64
March 0.023 20.89 206 96.8 58.9 97.217 16.65
April 0.024 20.88 206 97.6 59.7 97.241 16.65
May 0.080 20.85 206 108.8 66.8 97.321 16.66
June 0.400 20.72 205 125.1 79.5 97.721 16.73
July 3.182 19.71 198 148.6 94.5 100.903 17.28
August 3.776 18.57 190 150.6 96.5 104.679 17.92
September 0.381 18.45 189 124.7 79.1 105.060 17.99
October 0.108 18.42 189 112.2 70.2 105.168 18.01
November 0.061 18.40 188 106.9 64.9 105.229 18.02
December 0.018 18.40 188 94.7 56.8 105.247 18.02

Maximum Loading= 1.315 Used Life= 18.02 years(157870 hrs)

----------------------------------------------------------------------

84
8. Conclusions
The research is aimed primarily to deal with two major issues, routinely encountered by utility
planning engineers, however, without any clear answer:

(i) Economically optimize transformer sizing in a new project with load growth uncertain-
ties, and

(ii) The most cost-effective transformer replacement strategy in an existing system.

Even though the optimization technique developed here is independent of the type of transformer
(size, cooling, design, applications, etc.), the numerical examples demonstrate the principles are
limited to oil-cooled power transformers, ranges typically between 10-100MVA. This method
could easily be expanded to other types of transformers and varied applications.

To address the first issue, an improved optimization method has been proposed over the conven-
tional “loss evaluation” (“Total Owning Cost”) method. Besides, the economic considerations
traditionally adopted (use of “A” and “B” factors), the proposed method also includes the trans-
former thermal model, its loss-of-life evaluation subjected to cyclic load and the ambient tem-
perature profile. To take into account the uncertainties in future load growth projection probabili-
ty tree method is introduced. The economic evaluation utilizes the “Minimum Revenue Re-
quirement” method, which is appropriate for investor-owned utilities (IOUs).

The second problem also utilizes the same tools developed in this research, namely, the thermal
model, the economic model and the probabilistic approach to load growth.

Regarding the thermal model, overloading and remaining life expectancy estimation of a trans-
former, IEEE [1] and IEC [2] standards provide guidelines. These standards require a thorough un-
derstanding of the transformer design and heat-transfer equations. IEEE Clause 7 (and the IEC
standard) is simple, however, has some limitations.

In order to facilitate the use of these standards by practicing engineers and to avoid long and
complex calculations, as an integral part of this research, an interactive computer program has
been developed. Based on the research performed, the following conclusions are derived:

1) The IEEE classical model (Clause 7) is simple, easy to follow and requires minimum in-
formation. However, it has a number of limitations. It also requires less computation time. On
the other hand, IEEE Annex G model is more complex, requires additional information from the
manufacturer. This model requires small time step because it employs finite element forward
marching technique and hence, is computationally demanding. A direct comparison of the two
IEEE models reveals very close agreement for the hottest-spot temperature at steady-state and
reasonable overloading conditions. It is only at heavy overload and FOA type cooling, that the
results are different. The IEC model is similar to the IEEE classical model and is relatively sim-
ple. The major difference is the rated temperature rise and forced cooling type calculation.

2) The deterioration of insulation follows the “Reaction Rate Theory”. The hottest-spot
temperature and the exposed time reduce the mechanical tensile strength and the degree of poly-

85
merization of insulation. The normal insulation life of 158,000 hours, corresponding to the “Re-
tained Tensile Strength” of 20% and the “Degree of Polymerization” of 200, is utilized in this
research as the one “per-unit” life for power distribution transformer. There exists a good corre-
lation between the RTS and the corresponding DP.

3) The thermal behaviors of the transformer are studied independent of any economic evalu-
ation. Because of the nature of this application, very little effort has been made to combine the
thermal model and the economics together with other uncertainties, namely, load growth, trans-
former failure, and the variation of the ambient temperature and the load cycle. This research has
successfully combined a number of these variables into one optimization technique.

4) The proposed optimization technique has been tested for two real commonly faced appli-
cations by the planning engineers: (i) buy a new transformer, and (ii) optimize the transformer
utilization in a retrofit design. The results obtained from the simulations are quite satisfactory.

5) As a part of this research, a software has been developed and tested. This could be very
easily used by the planning engineers as a “design tool.” However, more work is needed and
other cases must be tested.

6) The results from the proposed optimization technique should also be verified from the
field data, when possible.

8.1 Contributions of this Research:

Major contributions of this research can be outlined as follows:

 Study in depth the three thermal models (two IEEE models and the IEC model). A sim-
plified guideline for transformer overloading has also been added in the Appendix A.

 Following the concepts of Per-Unit Life, Relative Aging factor, Equivalent Aging, and
end-of-insulation-life criteria, two simple equations have been developed from some ex-
perimental data available in the published literature to estimate the transformer remaining
life.

 A Windows based, object oriented program has been developed to calculate the hottest-
spot temperature, the top- and the bottom-oil temperature for each model. The program
also calculates the loss-of-insulation-life, the remaining life, and energy losses following
the methodology developed in this research.

 A method of optimal transformer sizing (New Design) and economic replacement alter-
natives (Retrofit Applications) are presented. It employs the minimum revenue require-
ment evaluation technique, and calculation of transformer end-of-life based on the ther-
mal model, and the probability of future load growth rate.

 In order to undertake the future uncertainties, the load growth probability- tree structure
is employed. The uncertainties of load and ambient temperature are also studied using
Monte Carlo simulation.

86
  The program enumerates the revenue requirement for each transformer size. The mini-
mum value of revenue requirement is the most economical alternative. Examples of dif-
ferent case studies are also presented in this research.

8.2 Future Work:

 The program is slow in optimization studies. Faster programming technique needs to be

employed to include uncertain future load growth.

 In order to simplify, the present work assumes “Straight Line” method of depreciation of
the transformer. Other depreciation method shall be explored. A survey of depreciation
method currently adopted by the utilities would be very useful.

 The calculation of loss-of-insulation-life is based on calculating its mechanical properties

(RTS and DP) after aging at constant high temperature. The mathematical modeling is
based on very limited test data. The model should be updated when more real data is
available.

 The business climate is changing the way utilities are operating. The financial model with
profit-oriented objective should be further reviewed. Alternate economic models may al-
so be implemented.

 “Loss Ratio (R)” used in the research is obtained from various transformer design. The
electrical efficiency of the transformer varies by this loss ratio and the daily load curve.
Sensitivity analysis should be performed for transformer buyer to minimize transformer
losses.

 Further validation of the optimization techniques from the real data collected from the in-
dustry is recommended.

 Theory and techniques for other transformer applications (Dry-Type, Large Power Trans-
formers, and Smaller Oil-Cooled Distribution Transformers) should be developed.

87
Bibliography
[1] IEEE Guide for Loading Mineral-Oil-Immersed Transformers, IEEE C57.91-1995.
[2] Loading Guide for Oil-Immersed Power Transformers, IEC Publication 354, 1991.
[3] R.L. Bean, N. Chackan Jr., H.R. Moore, and E.C. Wentz, Transformers for the Electric
Power Industry, McGraw-Hill, 1959.
[4] IEEE Guide for Application of Transformer Connections in Three-Phase Distribution
Systems, ANSI/IEEE C57.105-1978 (Reaffirmed 1987).
[5] R. Feinberg, Modern Power Transformer Practice, John Wiley & Sons, 1979.
[6] P.K. Sen, "Transformer Overloading," International Journal of Power and Energy Systems,
Vol. 19(1), pp. 52-56, 1999.
[7] V.M. Montsinger, "Loading Transformers by Temperature," AIEE Transactions, Vol. 49,
pp. 776-792, 1930.
[8] T.W. Dakin, "Insulation Deterioration Treated as a Chemical Reaction Rate Phenomenon,"
AIEE Transactions, Vol. 66, pp. 113-122, 1947.
[9] A.R. Bassetto and J. Mak, "Maintenance Practices to Improve Loading and To Extend the
Life of Power Transformers," Substation Equip. Diagnostics Conference IV, EPRI, 1996.
[10] W.J. McNutt, "Insulation Thermal Life Considerations for Transformers Loading Guides,"
IEEE Transactions on Power Delivery, Vol. 7(1), pp. 193-401, January 1992.
[11] S. Pansuwan and P.K. Sen, "Overloading, Loss-of-Life, and Assessment Guidelines of Oil-
Cooled Transformers: A Comparison of The IEEE and IEC Standards," International
Journal of Power and Energy Systems, Vol. 22(3), pp. 165-173, 2002.
[12] IEEE Guide for loading Mineral-Oil-Immersed Power Transformers Up to and Including
100 MVA with 55 C or 65 C Average Winding Rise, ANSI/IEEE C57.92-1981.
[13] A.C. Franklin and D.P. Franklin, The J&P Transformer Book, 11th ed., Butterworths, 1983.
[14] W.M. Flanagan, Handbook of Transformer Design & Applications, 2nd ed., McGraw-Hill,
1993.
[15] R.B. Kaufman, J.L. Pierce, and E.R. Uhlig, "The Effect of Transformer-Oil-Preservation
Methods on the Dielectric Strength of Oil," IEEE Transactions on Power Apparatus and
Systems, Vol. PAS-97, pp. 1315-1321, February 1958.
[16] F.W. Heinrichs, "Bubble Formation in Power Transformer Windings at Overload
Temperatures," IEEE Transactions on Power Apparatus and Systems, Vol. PAS-98(5), pp.
1576-1582, September/October 1979.
[17] W.J. McNutt, G.H. Kaufmann, A.P. Yitols, and J.D. MacDonald, "Short-Time Failure Mode
Considerations Associated With Power Transformer Overloading," IEEE Trans. on Power
Apparatus and Systems, Vol. PAS-99, pp. 1186-1197, May/June 1980.

88
[18] G.H. Kaufmann, "Impulse Testing of Distribution Transformers Under Load," IEEE
Transactions on Power Apparatus and Systems, Vol. PAS-96, pp. 1583-1595, September
1977.
[19] F.W. Heinrichs, D.E. Truer, and J.D. Phillips, The Effect of Gassing During Overloads on
the Impulse Strength of Transformer Insulation.: Paper A79434-2, B.C., 1979.
[20] W.J. McNutt and G.H. Kaufmann, "Evaluation of a Functional Life Test Model for Power
Transformers," IEEE Transactions on Power Apparatus and Systems, Vol. PAS-102(5), pp.
1151-1162, May 1983.
[21] W.J. McNutt and G.H. Kaufmann, "Mathematical Modeling of Bubble Evolution in
Transformers," IEEE Transactions on Power Apparatus and Systems, Vol. PAS-104(2), pp.
447-487, February 1985.
[22] G.H. Kaufmann, "Gas Bubbles in Distribution Transformers," IEEE Transactions on Power
Apparatus and Systems, Vol. PAS-96, pp. 1596-1601, September 1977.
[23] W.J. McNutt and O.R. Compton, "A Refined Mathematical Model For Prediction of Bubble
Evolution in Transformers," IEEE Transaction on Power Delivery, Vol. 4(1), pp. 391-398,
January 1989.
[24] T.O. Rouse, Loss of Transformer Life Detection Program.: Final Report, Niagara Mohawk
Power Corp., Janurary 1981.
[25] IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed Transformers.:
IEEE C57.104-1991.
[26] Y. Inoue, K. Suganuma, M. Kamba, and M. Kikkawa, "Development of Oil-Dissolved
Hydrogen Gas Detector For Diagnosis of Transformers," IEEE Transactions on Power
Delivery, Vol. 5(1), pp. 226-232, January 1990.
[27] Y. Leblanc, R. Gilbert, M. Duval, and J. Hubert, "Static Headspace Gas Chromatographic
Determination of Fault Gases Dissolved in Transformer Insulating Oils," Journal of
Chromatography, no. 633, pp. 185-193, 1993.
[28] S.R. Lindgren, S.C. Pyke, and M.A. Reynolds, On-line Transformer Load and Fault Gas
Analyzer for Determination of Reliable Overload Capacity.: CIGRE, session 1996, 12-110.
[29] C. Bengtsson, "Status and Trends in Transformer Monitoring," IEEE Transaction on Power
Delivery, Vol. 11(3), pp. 1379-1384, July 1996.
[30] R. Batruni and M.A. Lebow, "Determining the Effect of Thermal Loading on the Remaining
Useful Life of a Power Transformer from Its Impedance Versus Frequency Characteristic,"
IEEE Trans. on Power Delivery, Vol. 11(3), pp. 1385-1390, July 1996.

[31] B.C. Lesieutre, W.H. Hagman, and J.L. Kirtley, "An Improved Transformer Top Oil
Temperature Model for Use in An On-Line Monitoring and Diagnostic System," IEEE

89
 Transactions on Power Delivery, Vol. 12(1), pp. 249-256, Janurary 1997.
[32] J. Aubin and T. Langhame, "Effect of Oil Viscosity on Transformer Loading Capability at
Low Ambient Temperatures," IEEE Transactions on Power Delivery, Vol. 7(2), pp. 516-
524, April 1992.
[33] L.W. Pierce, "An Investigation of the Thermal Performance of an Oil Filled Transformer
Winding," IEEE Transaction on Power Delivery, Vol. 7(3), pp. 1347-1358, July 1992.
[34] L.W. Pierce, "Predicting Liquid Filled Transformer Loading Capability," IEEE
Transactions on Industry Applications, Vol. 30(1), pp. 170-178, January 1994.
[35] J.A. Jardini, C.M. V. Tahan, S.U. Ahn, and E.E. Ferrari, Distribution Transformer Loading
Evaluation Based on Load Profiles Measurement. Los Angles, CA: 14th IEEE PES
Transmission and Distribution Conference and Exposition, September 1996.
[36] A. Papoullis, Probability, Random Variables, and Stochastic Process, 3rd Ed., McGraw-
Hill, 1991.
[37] P.J. Spinney and G.C.Watkins, "Monte Carlo Simulation Techniques and Electric Utility
Resource Decision," Energy Policy, Vol. 24, pp. 155-163, 1996.
[38] IEEE Loss Evaluation Guide for Power Transformers and Reactors.: IEEE C57.120-1991.
[39] P.K. Sen, S. Pansuwan, and J.P. Nelson,. Chicago, IL: 61st American Power Conference,
1999, pp. 973-978.
[40] D.L. Nickel and H.R. Braunstein, "Distribution Transformer Loss Evaluation: I-Proposed
Techniques," IEEE Transaction on Power Apparatus and Systems, Vol. PAS-100(2), pp.
788-797, February 1981.
[41] D.L. Nickel and H.R. Braunstein, "Distribution Transformer Loss Evaluation: II-Load
Characteristics and System Cost Parameters," IEEE Transaction on Power Apparatus and
Systems, Vol. PAS-100(2), pp. 798-811, February 1981.
[42] Booth & Associates Inc., Distribution System Loss Management Manual.: NRECA
Research Project 90-7, National Rural Electric Cooperation Association, 1991.
[43] E. Paul Degarmo, W.G. Sullivan, J.A. Bontadelli, and E.M. Wicks, Engineering Economy,
3rd ed, Prentice Hall, 1997.
[44] C.A. Collier and C.R. Glagola, Engineering Economic and Cost Analysis, 3rd ed., Addison-
Wesley, 1998.
[45] M.O. Locks, Reliability, Maintainability, & Availability Assessment, 9th ed., Haydeb Book
Company, 1978.
[46] W. Nelson, Applied Life Data Analysis, John Wiley & Sons, 1982.
[47] IEEE Guide for Failure Investigation, Documentation, and Analysis for Power
Transformers and Shunt Reactors, IEEE C57.125-1991.

90
[48] M.R. Patel, Wind and Solar Power Systems, CRC Press, 1999.

91
Appendix A: Simplified Transformer Overloading Guidelines
Loading power transformers beyond the nameplate rating is commonly practiced by most utili-
ties. A simplified and general transformer overloading guidelines described below is also de-
picted in Figure A.1 and summarized in Table A.1 for further clarity and clear understanding.

 Understand Transformer Nameplate Rating and Design Fundamentals - Transfor-

mer Classification (Distribution and Power); Cooling; Average Winding Temperature
Rise; Insulation Type (thermally upgraded vs. kraft paper) and Class; Allowable Hottest-
Spot Temperature and Design Limits, Insulation Life vs. Transformer Life; etc.

 Determine End-of-Insulation Life Criteria and the Normal Insulation Life Value -
RTS and DP or other. Typical industry standard for transformer life is between 20 to 40
yrs. 30 yrs., is the most commonly used number.

 Moisture Content - Doubling of moisture content reduces insulation life by half.

 Determine the Ambient Temperature - Worst possible condition over 24 hrs. period
and estimate suitable correction. For every 1C ambient temperature decrement, loading
capacity can be increase by 1% without any loss-of-life or vice versa.

 Normal Life Expectancy Loading - Average (24 hrs.) maximum hottest-spot tempera-
ture of 110C without exceeding the maximum value of 120C with no additional loss-of-
life. Normal life is the transformer’s life when it operates at a constant hottest-spot tem-
perature of 110C. No limit for loading beyond nameplate rating as long as the hottest-
spot temperatures do not exceed 110C.

 Planned Loading beyond the Nameplate Rating - Average (24 hrs.) maximum temper-
ature of 110C without exceeding the maximum value of 130C with limited loss-of-life.
Aging rate is double for every 6-8C hottest-spot temperature increment.

 Long-Time Emergency Loading - It is recommended that the maximum hottest-spot

temperature should not exceed 140C, otherwise substantial loss-of-life is expected.

 Short-Time Overloading - Usually last for a short-time (less than half-an hour), and the
hottest-spot temperature may go up to 180C with severe loss-of-life. Transformer failure
is expected due to the bubble and gas formation in the oil.

 Maximum Overloading at any time - Limits to 2 times the highest rating.

 Maximum Allowable Absolute Temperature - 180C (IEEE) and 160C (IEC).

 Bushing Overloading Capacities - 150C maximum bushing hottest-spot temperature

and/or 2 times rated bushing current as per IEEE.

92
  Bushing-type Current Transformer - Bushing-type current transformers have the top-
oil as their ambient, which is limited to 105C.

 Recommended Practice - For normal operation, for the winding hottest-spot tempera-
ture, in case of OA/FA or OA/FA/FA set the alarm between 115-120C and trip between
125C and 130C. For FOA cooling, it is recommended that both alarm and trip should be
set at lower values by 5-10C. At higher operating temperatures, expect significant loss-
of-insulation-life depending on the duration, frequency, and the moisture content.

Figure A.1: The thermal and electrical limits for various types of loading

Table A.1: Thermal and electrical limits for various types of loading

IEEE IEC
Type of Loading Current Winding Top-oil Current Winding Top-oil
(pu) Hot-spot Temp. (pu) Hot-spot Temp.
Normal Life Expectancy Loading 2 120 105 1.5 140 105
Planned Loading beyond the Nameplate
2 130 110 - - -
Rating
Long-Time Emergency Loading 2 140 110 1.5 140 115
Short-Time Emergency Loading 2 180 110 1.8 160 115

In case of unavailable winding hottest-spot temperature values, recommended values for the top-
oil temp. gauge settings (OA/FA, OA/FA/FA) are 100C for alarm and 110C for trip. For FOA,
the alarm and trip settings are lowered by 5-10C and consult with the manufacturer.
93
In order to evaluate the transformer overloading capacity and the possible loss-of-life, the fol-
lowing factors should be considered:

 Nameplate Rating
 Cooling Class, Cooling Design and Design Margin
 Operating Conditions - Altitude, Ambient Temperature and Seasonal Adjustment
 Loading Cycle with respect to the Maximum Rating – Initial Load, Equivalent Conti-
nuous Load and Overloading Requirements
 Planned Loading beyond Nameplate Rating
 Long-Term Emergency Loading
 Short-Term Emergency Loading
 Loss-of-Life Expectancy
 Other Limitations than Hottest-spot Temperature (CT, Bushings, etc.)
 Cooling Upgrade
 Operational and Routine Maintenance Practice
 Historical Loading Data

94
Appendix B: Delphi 4.0 (Pascal for Windows) Program's Screen Shots

Figure B.1: Transformer Type and Cooling

Figure B.2: IEEE Annex G Input Data File

95
 Figure B.3: Daily Load and Temperature Profile

Figure B.4: Monthly Load and Temperature Profile

96
Figure B.5: Annual Load Growth Data

97
Figure B.6 (a) (b): IEEE Annex G Run Data and Plot

98
Figure B.7 (a) (b): IEEE Clause 7 Run Data and Profile

99
Figure B.8: Life Cycle Analysis Calculated Data

Figure B.9: Life Cycle Study: Composite Chart

100
 Figure B.10: Utility Financial Data and Energy Cost Input Data

Figure B.11: Optimum Transformer Sizing Output Data for New Procurement

101
Figure B.12: Transformer Replacement Strategy Data

102
Appendix C: Comparison Between IEEE Loading Guide C57.91-1995 and IEC Loading Guide IEC 354-
1991

No. Item compared IEEE C57.91-1995 IEC 354-1991 Comments

Distribution Transfor-  2,500 kVA, 3-phase
1  500 kVA
mer Size  833 kVA, single phase
Medium For IEEE’s power transformer in
 100 MVA, 3-phase excess of 100MVA overloading,
Power Transformer  33.3 MVA, single phase refer to IEEE C57.115-1991
2  100 MVA
Size Large
> 100 MVA, 3-phase
> 33.3 MVA, single phase
24 hr. average: 30C (40C Max for air
3 Average ambient temp. 20C (40C Max.)
cooled)
Rated average winding rises in
Average winding temp. 65C (ONAN distribution transformer) IEC vary by cooling type.
rise at rated load 65C (thermally upgraded paper) 63C (ON.. power transformer)
4
TWR 55C (Kraft paper) 63C (OF.. power transformer) Rated average winding temp. rise
68C (OD.. power transformer) given here is maximum design
value.
55C (45C) for OA 55C (ONAN distribution transformer) Rated top oil temp. rises given
50C (40C) for FA,  133% 52C (ON.. power transformer) here are typical values.
Top oil temp. rise at 45C (37C) for FA, > 133% 56C (OF.. power transformer)
5 rated load 45C (37C) for FOA both direct and non- 49C (OD.. power transformer)
TTOR direct flow
Note: the temperature in parenthesis is for
55C average winding rise.

103
COMPARISON BETWEEN IEEE LOADING GUIDE C57.91-1995 AND IEC LOADING GUIDE IEC 354-
1991 (Contd.)
The standard preferably recommends using Using of direct measurement is recom- See IEC thermal diagram from
imbedded detector for measuring hot spot mended for higher accuracy. However, the Figure 3.4.
allowance. However, second alternative is simplified calculation is given as:
given by calculation as:
TGR  (TWR  TAVGOR ) H
TGR  TWR  TTOR  15(10) TGR is calculated from avg. winding rise
over avg. oil multiplied by H factor.
TGR is calculated from avg. winding rise H factor varies from 1.1 to 1.5 depending
over top oil plus 15(10) corresponding to 65C on transformer size and design.
and 55C winding rise respectively.
Hot spot allowance 23C (ONAN distribution transformer)
(over top oil rise at rated The third alternative is to assume that rated 26C (ON.. power transformer)
6 load) hot spot rise is 80C and 65C for 65C and 22C (OF.. power transformer)
TGR 55C average winding rise respectively. 29C (OD.. power transformer)
Therefore, rated hot spot allowance can be
found by subtracting rated top oil rise by rated
hot spot rise. The results are as follows:

25C (20C) for OA

30C (25C) for FA,  133%
35C (28C) for FA, > 133%
35C (28C) for FOA both direct and non-
direct flow

Rated hot spot temp. rise given

Rated hot spot 65C (55C winding rise)
7 78C here is maximum design value.
temp. rise 80C (65C winding rise)
Hot spot temp. rise
8 THS  TTO  TG THS  TTO  TG
THS

104
COMPARISON BETWEEN IEEE LOADING GUIDE C57.91-1995 AND IEC LOADING GUIDE IEC 354-
1991 (Contd.)
As IEC 76-2 does not consider
thermally upgraded insulation for
oil-immersed transformer, temper-
ature rise limits and improvement
98C in thermal behavior may be taken
Rated hot spot 95C (55C winding rise)
9 110C (when thermally upgraded paper is into account by agreement be-
temperature 110C (65C winding rise) tween the manufacturer and user.
used.)
A normal life expectancy at hot
spot temperature of 110C is used.
Rated hot spot temperature given
here is maximum design value.
ON.. cooling The oil exponential, n, of IEEE
n and IEC loading guide is the same
 K 2R  1
TTO  TTOR   for all cooling types.
 R 1 
n = 0.8 for ONAN distr. transformer
= 0.9 for ON. power transformer
n
 K 2R  1 OF.. & OD.. cooling
TTO  TTOR  
 R 1  Top oil rise is calculated from bottom oil
Top oil temp. rise & avg. oil rise.
10 Same equation is applied for all cooling type.
TTO n
n = 0.8 for OA  K 2R  1
TTO  TBOR   +
= 0.9 for FA  R 1 
= 1.0 for FOA
2(TAVGOR  TBOR)K 2m
n = 1.0
m = 0.8 for OF..
= 1.0 for OD..
2(TAVGOR  TBOR ) is a calculated value
of top oil rise over bottom oil.

105
COMPARISON BETWEEN IEEE LOADING GUIDE C57.91-1995 AND IEC LOADING GUIDE IEC 354-
1991 (Contd.)
The winding exponential, m, of
Hot spot allowance TG  TGRK 2m TG  TGR K 2m
IEEE and IEC loading guide is the
11 TG m = 0.8 for OA, FA, NDFOA m = 0.8 for ONAN, ON.. & OF.. same for all cooling types.
= 1.0 for DFOA = 1.0 for OD..
Top oil temp. rise follows exponential equa- Top oil rise in ON.. cooling and bottom oil IEEE’s oil time constant,  o , has
tion. rise in OF..&OD.. cooling also follow to be corrected if oil exponential,
exponential equation. n, is not equal to 1.0, but IEC does
TTO  TTO,i  (TTO,u  TTO,i )(1  e t /o ) TTO  TTO,i  (TTO,u  TTO ,i )(1  e t / o ) not mention this issue.
Top oil temp. rise where TBO  TBO ,i  (TBO ,u  TBO ,i )(1  e t / o )
12 TTO  o is oil time constant ( 3 hrs.) The second term of TTO for OF.. and
OD.. cooling, 2(TAVGOR  TBOR)K ,
2m

follows exponential equation but at faster

winding time constant. IEC recommends
neglecting this time constant.
For 110C rated hot spot No absolute life is mentioned in the IEC IEC doesn’t give absolute life.
 15000  standard but relative aging.. IEEE no longer specifies absolute
 
 THS  273 
Per Unit Life = 9.80 x10 18 e life, however, still gives various
 15000

15000  Relative aging rate is double for every absolute life from different end of
 
Relative aging rate (FAA) = e  383 THS  273  fixed 6C increase in hot spot temperature. life criteria for information only.
For 95C rated hot spot
(THS 98) / 6
 15000  Relative aging rate = 2
 
18  THS  273 
13 Loss of insulation life Per Unit Life = 2.00 x10 e
 15000 15000 
  
 368 THS  273 
Relative aging rate = e

Aging rate is double when hot spot tempera-

ture increases about 7-8C in the hot spot
temperature ranges from 110-140C.

106