EPRI Underground Distribution Systems Reference Book

CHAPTER 16 Transformers and Equipment

Authors: Stephen L. Cress, Reviewers: Gordon Hayslip, Snohomish PUD
Kinectrics. Inc. John Igielski, Northeast Utilities
Ali Naderian, Kinectrics, Inc. Ken Ochs, We Energies
Joseph Somma, Consolidated Edison

Abstract:
This chapter reviews several aspects related to transformers, including transformer
losses, loading characteristics, selection criteria for pad-mounted transformers,
transformer cooling, interpretation of tests on transformers and oil, and capacitors.

Stephen L. Cress graduated in 1976 from The University of Toronto

with a Bachelor of Applied Science in Electrical Engineering. He is cur-
rently Department Manager – Distribution Asset Management at Kine-
ctrics Inc. Stephen has over 33 years’ experience in specialized technical
investigations, research, testing, and applications in the power distribu-
tion field based on his work at Federal Pioneer, Ontario Hydro
Research Division, and Kinectrics Inc. He has conducted major proj-
ects for North American distribution utilities dealing with: transformer
loading and sizing, transformer losses and efficiency, asset management, asset condition
assessment, life extension, distribution protection, equipment failure analysis (transform-
ers, switchgear, fuses, capacitors), standard testing, distribution modeling, and develop-
ment of utility-oriented engineering software. Stephen is the holder of a U.S. patent on
high-voltage current limiting fuses. He is a co-author of the CEATI reference books
Application Guide for Distribution Fuses and Engineering Guide for Distribution Overcur-
rent Protection. Stephen’s work in the development of probabilistic methods for calculat-
ing transformer loss evaluation, loss-of-life, and loading have been incorporated in the
commercially available CEATI TRANSIZE TM computer program. He has published
papers with international organizations such as IEEE, CIRED, and INTER-RAM, and
articles in power industry magazines. He is the Chair of the harmonized CSA and
CNC\IEC TC32 Committees dealing with High Voltage Fuses, and a Professional Engineer
in the Province of Ontario.

Ali Naderian received his B.Sc. and M.Sc. degrees from Sharif Univer-
sity of Technology in 1998 and the University of Tehran in 2000, respec-
tively. During his studies, his part-time employment experience included
ISC (1997-1999) for testing of switchgear and circuit breakers, and ITS
(1999-2000) for designing and manufacturing of HV power transform-
ers. He was co-designer of a 3*300-kV cascade HV testing transformer.
He compared commercially available RTV coatings for outdoor insula-
tors in his PhD thesis during his research at the University of Waterloo,
Ontario (2003-2006). He has been a project manager of high-voltage testing at Kinectrics,
Inc. (formerly Ontario Hydro Research) since 2007, working on diagnostics of power
transformers, high-voltage cables, and outdoor insulators. He performs on-line and off-
line PD measurements for HV apparatus. His research interests include high-voltage test
techniques, dielectric frequency response, and partial discharge. He has published several
papers, is actively involved in IEEE transformer working groups, and is a registered engi-
neer in the Province of Ontario.

16-1
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

16.1 INTRODUCTION induced current, called eddy current, produces losses in

This edition of the Bronze Book covers only a subset of the core plates. The eddy current loss can be calculated
what will be a comprehensive look at underground dis- by Equation 16.2-1
tribution transformers. Included here are sections on
π2
transformer losses, loading characteristics, pad- Peddy = σ f 2 d 2 B 2V
mounted transformer selection criteria, interpretation 6 16.2-1
of tests on transformers and oil, as well as a discussion Where:
on capacitors. σ is the core conductivity.
ƒ is frequency.
The next edition will also include an overview of trans- d is the core thickness.
former types by application, unit components and core B is the peak value of the flux density.
construction, installation options, and insulation types. V is the core volume.
Additional topics will be transformer cooling, testing
and monitoring, and typical examples of failure root As per Equation 16.2-1, eddy current is controlled by
causes. using laminated core to cut large current loops at the
cross section of the core. The no-load loss is the sum of
The reader is encouraged to refer to other sources more hysteresis and eddy current losses, as shown in Equation
broadly covering this topic, including the Electric Power 16.2-2.
Distribution Handbook by Tom Short (2004), Power
Transformers, Principles and Applications by John Wind- P0 = Peddy + Ph 16.2-2
ers (2002), and the ABB Distribution Transformer Guide
(2002). 16.2.2 Load Losses
Load losses (or copper losses or resistive losses) are pri-
16.2 TRANSFORMER LOSSES marily a function of the winding design of the trans-
Losses in distribution transformers are categorized as former. They result from the load current flowing in the
load and no-load losses. Load losses vary with the primary and secondary windings.
square of the load on the transformer, whereas no-load
losses are continuous and constant regardless of load. Components of load loss are I2R and stray losses. For a
distribution transformer, I2R is in the range of 92-99%
16.2.1 No-Load Loss of the load loss. The proportion is lower for larger kVA
sizes. Load loss is affected by:
No-load losses (or excitation losses, iron losses, or core
losses) are inherent to the excitation of the transformer. • number of turns of winding
No-load losses are associated with the core design. They • mean length of the primary and secondary turns
include core loss, dielectric loss, and the loss in the wind-
• conductor cross section
ings due to exciting current. For distribution transform-
ers at 27.6 kV and below, the dielectric loss is negligible. • material of the conductor—i.e., copper or aluminum

The no-load loss in the transformer core is a function of Stray losses vary inversely as the temperature, thereby
the magnitude, frequency, and wavefor m of the making necessary the calculation of load loss at a spe-
impressed voltage. No-load losses are affected by volt- cific temperature such as 85°C. Stray losses have three
age fluctuations. When an AC voltage is applied to the components: conductor eddy currents, conductor circu-
terminals of the transformer, magnetizing current flows lating currents, and stray currents in the core wall and
through the winding, and a magnetic flux appears in the core clamps.
core. The predominant component is core loss, which is
composed of hysteresis and eddy current losses. The current, which is applied to the windings, creates
losses due to the winding resistance. The losses of a
The hysteresis loss is proportional to the frequency and transformer are losses incident to a specified load car-
dependent on the area of the hysteresis loop in the B-H ried by the transformer. Load losses in distribution-class
diagram, and, therefore, characteristic of the material transformers mainly include I 2 R loss in the windings
and a function of the peak flux density. due to load current.

The variable magnetic flux induces current running in Load loss follows Ohm’s law and can be decreased by
paths perpendicular to the direction of the flux. The reducing the number of winding turns, by increasing the
cross-sectional area of the turn conductor, or by a com-

16-2
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

bination of both. However, reducing the number of Transformer efficiency (η) is the ratio of a transformer’s
turns requires an increase of the flux—i.e., an increase useful power output to its total power input as indicated
in the core cross-section, which increases the iron weight in Equation 16.2-3 (IEEE 2006).
and iron loss. Therefore, a tradeoff has to be made
Pout Pin − Ploss
between the load loss and the no-load loss. η= = 16.2-3
Pin Pin
16.2.3 Total Loss
The following summarizing relationships are useful Where:
when considering the losses in distribution transformers: η is the efficiency.
Pin is the input power.
No Load Loss α Flux Density α 1/# Turns α 1/ Core Pout is the output power.
Cross Section Ploss is the total power loss of the transformer to
Load Loss α # Turns be introduced.
Impedance α Reactance α (# Turns)2
Initial Cost α Core Material α Winding Material In many jurisdictions, government energy agencies have
mandated minimum efficiency levels for liquid-filled and
Table 16.2-1 summarizes the components of load and dry-type distribution transformers (DOE 2007; NEMA
no-load losses. 2002). Table 16.2-3 provides an example of the accepted
efficiency levels for liquid-immersed distribution trans-
Figure 16.2-1 and Table 16.2-2 provide some typical
load and no-load loss values for distribution transform-
ers. Figure 16.2-1 illustrates how load losses vary with
load on the transformer.

16.2.4 Transformer Efficiency

Transformer efficiency is related to the amount of watts
losses that occur when the transformer is in operation.

Table 16.2-1 Components of Transformer Load and No-

Load Loss
Type No-Load Lossa Load Lossa
I2R from load I
Electric I2R from No-load I I2R from I supply-
ing Losses
Core Hysteresis Loss
Conductor Eddy
Core Eddy Current Loss
Magnetic Current from
Stray Eddy Current Loss in
Leakage fields
Internal components
Dielectric Dielectric Loss

a. Where I represents current, and I2R is the current Figure 16.2-1 Typical load and no-load losses of
squared times the conductor resistance. distribution transformers.

Table 16.2-2 Typical Losses for Power Distribution Transformers

Rating No-Load Loss Efficiencya
KVA Load Loss Watts Watts at 50% load
250 3800 880 0.9925
400 5500 1200 0.9932
667 7900 1700 0.9941
1000 11000 2300 0.9945
1500 15000 3000 0.9950
2500 23000 5000 0.9954

a. Calculated using Equation 16.2-4.

16-3
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

formers. An example of the minimum efficiency for dry- Present distribution transformers are, for the most part,
type distribution transformers is shown in Table 16.2-4 between 98% and 99.5% efficient. For the new trans-
(DOE 2007). These efficiency values are computed at formers, the guideline from (DOE 2007), presented in
50% of nameplate-rated load. Tables 16.2-3 and 16.2-4, should be followed. Because
virtually all-electric energy passes through distribution
Efficiency can be expressed directly as a function of the transformers, losses in these devices, though small, are
load and no-load losses as in Equation 16.2-4 (NEMA estimated to constitute as much as 2 to 3% of all energy
2002). The efficiency values computed using this for- generated.
mula are provided alongside the load and no-load losses
in the examples in Table 16.2-2. Generally transformers are at maximum efficiency when
they are 50% loaded. When transformers are lightly
KVA × Lp.u. loaded, the no-load losses form a large percentage of the
η= 16.2-4
KVA × Lp.u. + P0 + Lp.u.2 × PL power utilized, and therefore, the efficiency is low. As
the transformer is loaded to higher levels, the load losses
Where: dominate the efficiency. The maximum efficiency point
KVA is the transformer rated power. is the optimal point of lowest load and no-load losses. It
PL is the load loss. is determined by the design of the transformer and theo-
P0 is the no-load loss. retically could be designed to occur at any load percent-
Lp.u. is the per-unit load (the ratio of actual load age. It typically is designed to occur at 50%, because the
to the rated full load). average load tends to be about 50% of the peak load.
However, transformers with high no-load losses are

Table 16.2-3 Standard Levels of Efficiency for Liquid-immersed Distribution Transformers (DOE 2007)

Table 16.2-4 Standard Levels of Efficiency for Dry-type Distribution Transformers (DOE 2007)

16-4
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

most efficient at 60%-80% load, and transformers with tive permeability of the steel, thus reducing the losses,
low no-load losses are most efficient at about 40% load. but it also decreases the saturation magnetic flux den-
(See Figure 16.2-2.) sity, which increases the amount of material required in
the core. Together, these effects reduce the no-load loss
16.2.5 Reduction of Transformer Losses of the core, but the amorphous steel cores are larger,
Reduction of transformer losses and improvement in heavier, and more costly to produce. (Permeability
efficiency can be achieved by reduction of either load or increases by a factor of 4, but saturation flux density
no-load losses. For any given set of core and winding decreases by a factor of 0.75, requiring 1.3 times as
materials, reduction of load losses often leads to an much material in the core, so overall loss is lower by a
increase in no-load losses and vice versa. factor of 3.) On average, amorphous core loss values are
about 30% of that for high-efficiency silicon steel, and
Many factors of core design affect no-load losses and only 15% of that for older, less efficient steels.
can be altered to reduce these losses. Higher magnetic
flux density leads to higher losses. Larger gaps in cut Numerous questions have arisen regarding the mechani-
cores lead to higher losses. These gaps can be reduced cal robustness and long-term mechanical performance
by manufacturing techniques. The thickness of the of amorphous metals. Short-term testing programs have
enamel insulation on the winding conductors affects the not substantiated these beliefs, but the concern persists.
size of the core. High-quality enamel can be used in very
thin layers to reduce core size and no-load losses. More recently nano-crystalline steel has become avail-
Mechanical arrangement of the windings and taps also able for use in transformer cores.
affects the efficient use of space and the size of the core.
The best are based on an Fe-Zr-B alloy that is formed in
Traditionally cores have been made from grain-oriented an amorphous state and then annealed to produce very
silicon steel formed into thin sheets and wrapped into a small grain sizes. This approach makes the material less
rectangular shape. The loss decreases as the thickness of brittle and thereby decreases production costs. This steel
the sheets decrease. Standard grades are M-2 at 0.18 has even higher permeability and also higher saturation
mm, M-3 at 0.23 mm, M-4 at 0.27 mm, and M-6 at 0.35 induction than the amorphous materials, but it is not
mm. Losses also depend on the permeability of the steel yet available in manufactured transformer cores. The
alloy. Higher permeability leads to lower losses. The new steel has 17 times the permeability of steel and 0.89
permeability depends upon the alloy and the orienta- of the saturation flux density; so losses should be
tions of the grains. reduced by a factor of 15.

A large advance in technology occurred in the 1980s Load losses are caused primarily by the heating of the
with the development of amorphous steel cores. These windings by the passage of current (I2R losses). The cur-
cores are also built up by wrapping thin sheets or rib- rent is determined by the impedance of the load on the
bons, but the steel itself (such as Co-Fe-Si-B alloy) is transformer and the voltage levels and so is not under
quenched during manufacture to ensure that no grains the control of the transformer designer. The resistance
are formed in the steel. This process increases the effec- depends on the material used in the winding, the cross-
sectional area of the wires, and the number of turns.

Transformer windings are made of either copper or alu-

minum in round wires, square wires, or flat sheets. The
resistivity of aluminum is about 1.6 times larger than
that of copper, but aluminum has a lower cost. Many
different alloys of aluminum and copper are available.
In general, the lower resistance alloys are more expen-
sive and harder to work with in the manufacturing pro-
cesses, leading to higher initial costs.

In addition to choice of material, load losses are

affected by the cross-sectional area of the wire used.
Larger wires produce lower load losses, but then the
windings are larger, and this requires a larger core,
Figure 16.2-2 Transformer efficiency as a function of load. which increases the no-load losses.

16-5
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

Some load loss is caused by induced currents from adja- 16.2.7 Cost-of-Losses Formula
cent windings. These currents can be reduced by using The lifetime cost of a transformer depends on the capi-
continuously transposed conductor in the winding and tal cost of the transformer and the cost of the load and
thus reducing load losses. This approach also leads to no-load losses during its lifetime. The present value
higher initial costs. method is often employed to express the lifetime cost in
terms of a dollar value in the present year. Losses from
16.2.6 Transformer Short-circuit Impedance distribution transformers are a significant contribution
When provided a customer’s cost of no-load and load to distribution system losses, and their reduction repre-
losses, transformer manufacturers will use software that sents an opportunity for improving energy efficiency
performs hundreds of iterations, varying core, winding,
and tank options, to arrive at a transformer with an A cost-of-losses formula for purchasing purposes is
optimal balance of losses and initial cost. often employed to determine the lifetime costs for vari-
ous transformer options available to utilities. Compari-
The short-circuit impedance of a transformer is used to sons can then be made between more capital intensive
calculate the maximum short-circuit current and is low-loss transformers and less expensive higher-loss
needed for sizing circuit breakers, fuses, cables, and transformers.
other equipment connected to the secondary of the
transformer. The following paragraphs describe the general formula-
tion of a cost-of-losses formula. Table 16.2-5 defines the
Transformer impedance (or short-circuit impedance or quantities used in these equations.
impedance voltage) is the percent of per unit voltage
that must be applied to the primary side of a trans-
former, so that the rated current flows when the second- Table 16.2-5 Definition of Symbols for Cost of Losses
ary terminals are short-circuited. This impedance is Formula
formulated as Equation 16.2-5. CAP Capital cost ($)
CLL Present value of cost of load losses ($/W)
U CLL(m) Cost of load losses for month “m” ($/kW)
Z % = Z × 100 16.2-5
ZP CLY(y) Cost of load losses for year “y” ($)
CNLL Present value of cost of no-load losses ($/W)
As the no-load test result is available, the ohmic part of CNLL(m) Cost of no-load losses for month “m” ($/kW)
the impedance can be calculated using Equation 16.2-6, D Demand charge, monthly ($/kW)
and therefore, the inductive part of the impedance can D(m) Demand charge for month “m” ($/kW)
be derived by Equation 16.2-7. E Energy charge, monthly (¢/kWh)
EOP(m) Energy charge off-peak for month “m” (¢/kWh)
P3ϕ − load − loss EP(m) Energy charge on-peak for month “m” (¢/kWh)
R% = × 100 16.2-6
MVA3ϕ .106 FYG(y)
Factor for yearly load growth accumulated to year
“y”
X % = Z %2 − R%2 16.2-7 g(y) Growth of load for year “y” (%/100)
HOP(m) Hours off-peak for month “m” (h)
In a transformer having a tapped winding, the short-cir- HP(m) Hours on-peak for month “m” (h)
cuit impedance is referred to a particular tap. Unless i(y) Interest rate for year “y” (%/100)
otherwise specified, the nominal tap applies and is the j(y) Inflation rate for year “y” (%/100)
impedance (Z%) that is marked on the nameplate. The PVLC Present value of lifetime cost ($)
impedance voltage of distribution transformers with LL Load losses (W)
rated power below 630 kVA is usually 4% or less, and LSF Loss factor (average loss/peak loss)
this value is usually around 6% for 630 kVA up to NLL No-load losses (W)
2.5 MVA distribution transformers. NY Number of years in economic study period
p(y) Growth of power costs for year “y” (%/100)
For parallel operation of two or more transformers, PVF Present value factor for a period of years
short-circuit impedance is critical. If paralleled trans- PVF(y) Present value factor for year “y”
formers do not have the same short-circuit impedance, RATL Rated load for transformer (kVA)
the load will be shared in an unbalanced way such that Responsibility factor (load at system peak/peak
RF
load)2
one transformer can be overloaded and the transformer
UF Utilization factor (peak load/rated load)
can be underloaded.

16-6
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

The basic form of the cost of losses formula, providing 1 ⎛ E ⎞

the present value of the lifetime cost (PVLC) of a trans- CLL = ⎜12D ∗ RF + 8760 ∗ LSF ⎟
formers, is as expressed in Equation 16.2-8. 1000 ⎝ 100 ⎠ 16.2-10

{ }
NY
∗∑ [UF ∗ FYG ( y )] ∗ PVF ( y )
2
PVLC = CAP + NLL ∗ CNLL + LL ∗ CLL 16.2-8
y =1

Where: (1 + p ) y −1
PVF ( y ) = 16.2-11
CAP is the capital cost or initial purchase price (1 + i ) y
of the transformer. NY
NLL is the no-load losses that occur continu- PVF = ∑ PVF ( y ) 16.2-12
ously when the transformer is energized, y =1

regardless of the loading. FYG ( y ) = (1 + g ) y −1 16.2-13

CNLL is the cost of no-load losses and is inde-
pendent of the loading and dependent on
the demand and energy charges. Time-of- Where:
use energy charges can be considered by UF, the utilization factor, is defined as the ratio of the
using on-peak and off-peak energy peak load to the transformer rated load. It repre-
charges, and considering the hours that sents the portion of the transformer rated load that
the transformer is on-peak or off-peak. is utilized when the transformer is at its highest
LL is the load loss at rated load. The value of loading.
load loss at rated load is a measured peak load
parameter, and load losses at other load- UF = 16.2-14
rated load
ings are derived from this value.
CLL is the cost of the load losses, and depends
on the demand and energy charge rates as RF, the peak responsibility factor, is used to adjust the load to
well as on the loading of the transformer reflect the proportion of the asset load that actually contrib-
throughout its life. utes to the peak load of the utility as a whole. That is, it indi-
cates how much the lo ad lo ss of the particu lar
Common cost-of-losses equations use flat-rate demand transformer contributes to the total demand. The
and energy charges and fixed annual economic factors, responsibility factor is the ratio of the transformer load
such as interest rate, to evaluate the lifetime costs of at system peak to the peak load, all squared.
load losses (CLL) and cost of no-load losses (CNLL). 2
⎛ load at system peak ⎞
The concepts of load factor, loss-factor, utilization fac- RF = ⎜ ⎟ 16.2-15
tor, and responsibility factor are used to describe the ⎝ peak load ⎠
loads on the transformer. A load growth factor can be
used to include the influence of rising loads on the LSF, the loss factor, is the ratio of the average loss to the
transformer losses over the transformer’s lifetime. Note peak loss. The loss factor can be derived from the load
that the load growth factor is 1 in the first year, and then factor. The load factor is a single value that character-
changes to a fixed factor at the start of the second year. izes the load profile. The load factor is the ratio of the
The present value factor includes the influence of eco- average load to the peak load.
nomic factors such as inflation of the cost of power and
interest rates. The growth in power costs factor is 1 in Load and loss factors are dependent on the shape of the
the first year and then changes to a fixed factor in the load profile. Loading profiles are different for indus-
second year. The rate of interest starts in the first year. trial/commercial, urban residential, and rural residential
Note that there are 8760 hours in a year. transformers. Industrial/commercial loads are steadier
both over the day and over the week. A typical load fac-
1 ⎛ E ⎞
CNLL = ⎜12D + 8760 ⎟ ∗ PVF 16.2-9 tor is 0.85. Residential loads are more variable, with typ-
1000 ⎝ 100 ⎠ ical load factors of 0.4 for a single transformer. Urban
residential transformers tend to be more heavily loaded
than rural transformers.

Theoretically the loss factor may have a value between

the value of the load factor and the load factor squared,
depending on the load profile shape. A common for-

16-7
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

mula that has been used to calculate loss factor from NY

load factor is as shown in Equation 16.2-16. PVF = ∑ PVF ( y ) 16.2-23
y =1

LSF = 0.85 * LDF 2 + 0.15 * LDF 16.2-16

FYG ( y ) = [1 + g ( y )] ∗ FYG ( y − 1) 16.2-24

Where FYG (1) = 1 + g (1) 16.2-25

LDF is the load factor of the daily load profile.
Where the variables are as defined in Table 16.2-5.
PVF, the present value factor, accounts for the changing
value of money and expresses the present worth of dol- With computer assistance the cost of losses formula can
lars spent in the future. be further expanded to replace the use of the load factor
concept and determine loads directly from daily and
Economic factors, of course, are generally not fixed over monthly profiles.
long time periods of time. Further, there is considerable
utility interest in applying variable or time-of-use rates. Figure 16.2-3 shows a general graph of costs versus
With addition of several parameters, the cost of losses transformer mass for a typical distribution transformer.
formula can be modified to consider these variable eco- There is an optimum value for total cost. If the loss eval-
nomic inputs. uation figures are submitted to the transformer manu-
facturers in the request for quotation, they can design a
Energy and demand charges can be expressed as being transformer with an optimal cost from the end user
dependent on the time of use, either on-peak or off- point of view. The result of this process should be the
peak. Economic factors and the load growth can be cheapest transformer in the useful life period—i.e., with
allowed in the equation to vary from year to year. Note the lowest total owning cost, optimized for a given
that either p(y) or j(y) must be set to zero for all years application.
y. To use the initial cost of power in the first year, set
p(1) to zero [or j(1) to zero, if you are not using p(y)].

Therefore the components of the cost of losses formula

can be further expressed as:

⎡ 1 12
⎤
CLL = ⎢ ∗ ∑ CLL( m ) ⎥
⎣1000 m =1 ⎦ 16.2-17

{ }
NY
∗∑ [UF ∗ FYG ( y )] ∗ PVF ( y )
2

y =1

CLL( m ) = D( m ) ∗ RF
⎡ EP ( m ) EOP ( m ) ⎤
+ ⎢ HP ( m ) ∗ + HOP ( m ) ∗
⎣ 100 100 ⎥⎦
∗LSF
16.2-18 Figure 16.2-3 Transformer mass vs. transformer lifetime
cost.
⎡ 1 ⎤12
CNLL = ⎢ ∗ ∑ CNLL( m ) ⎥ ∗ PVF 16.2-19
⎣1000 m =1 ⎦

CNLL( m ) = D( m ) + HP ( m )
EP ( m )
∗ + HOP ( m ) 16.2-20
100
EOP ( m )
∗
100
[1 + p( y )] ∗ [1 + j ( y )]
PVF ( y ) = ∗ PVF ( y − 1) 16.2-21
[1 + i ( y )

PVF (1) =
[1 + p(1)] ∗ [1 + j (1)] 16.2-22
[1 + i (1)]

16-8
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

16.3 LOAD CHARACTERISTICS FOR 24-hour load profile is modeled by a series of constant
TRANSFORMERS loads of a short duration, usually 1 hour. The equivalent
One of the main considerations for selecting the appro- load during the short time steps is determined by using
priate transformer is the characteristic of the load. Not the maximum peak load during the short-time period
only the number and type of loads, but the load pattern under consideration. An equivalent two-step overload
needs to be considered. cycle can be used for determining emergency overload
capability, as shown in Figure 16.3-1. The equivalent
Because load is a function of human behavior and life- two-step load cycle consists of a prior load and a peak
style variables, as well as the type and size of electric load. A constant load that generates total losses the
equipment and weather changes, load forecasting has same as a fluctuating load is assumed to be an equiva-
some level of uncertainty. lent load from a temperature standpoint. Equivalent
load for a specific part of daily load is expressed by
16.3.1 Load Types Equation 16.3-1.
Several types of loads occur on a distribution systems: N

• Domestic (residential): Mainly lights, fans, heaters, ∑L t 2

i i

refrigerators, air conditioners, ovens, small pumps, Leq = i =1

N
16.3-1

and other household appliances. ∑ ti

i =1
• Commercial: Lighting of shops, air-conditioning,
heating, and shop appliances. Where:
• Industrial: Medium and large motors. Li is various load steps in% or per unit.
N is the total number of load steps.
• Municipal (Public): Street lights, and traffic signals.
ti is the duration of each load step.
• Agricultural: Motors and pumps.
16.3.3 Peak Load
Commercial loads typically have a dedicated trans-
Equivalent peak load is the rms load obtained by Equa-
former; however, multiple residences are usually served
tion 16.3-1 for the limited period over which the major
by a single or three-phase transformer. Public loads usu-
part of the actual peak exists. If the peak load duration
ally need their own dedicated transformer due to the load
is over-estimated, the rms peak value may be consider-
size. The daily load profiles of these three load categories
ably below the maximum peak demand. To protect
are not usually matched. Commercial and industrial
against overheating due to high, brief overloads during
loads may at times e served on a spot network of multiple
the peak overload, the rms value for the peak load
transformers in parallel. Some service areas, mainly in
period should not be less than 90% of the integrated ½
metropolitan areas of loads including residential and
hour maximum demand.
commercial loads are serviced from distributed grids of
many transformers in parallel via network protectors.
Besides daily peak load, seasonal peak load needs to be
taken into account. Depending on the geographic loca-
Distribution transformers serving primarily residential
tion, and due to weather conditions, a winter peak or
loads regularly carry average loads that are only 15 to
summer peak can be expected.
25% of the transformer's rated capacity but also must be
designed to support peak morning and evening loads.
An example of a daily load profile with two peak loads
Because of the wide gap between peak and non-peak
is given in Figure 16.3-2.
loads, and the relatively limited amount of time that the
transformer is peak-loaded, average transformer load-
ing tends to be fairly low.

16.3.2 Load Profiles

Transformer loads generally follow cycles that repeat
daily, and may have seasonal variation during the year
and yearly growth. The daily load variation for many
utilities repeats every 24 hours and has two common
forms: a single hump shape (as shown in Figure 16.3-1)
or a double hump shape. A multistep load cycle calcula- Figure 16.3-1 Example of actual load cycle and
tion can be used to describe the load (IEEE 1995b). The equivalent load cycle of IEEE C57.91.

16-9
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

the actual maximum demand on the system as described

in Equation 16.3-4.

Load Diversity = ∑ Individual Maximum Demands

− System Maximum Demand
16.3-4

Diversity factor in a distribution system is the ratio of

the sum of the individual maximum demands of the var-
ious subdivisions of a system to the maximum demand
of the whole system under consideration (see Equation
16.3-5). Loads do not normally all peak at the same
time. Therefore, the sum of the individual peak loads is
greater than the peak load of the composite system.
Therefore, diversity factor is usually more than one.
Figure 16.3-2 Morning and evening peak loads (from
Pabla 2004).
DF =
∑ Individual Maximum Demands
16.3-5
System Maximum Demand
16.3.4 Average Load
Demand factor is the ratio of the maximum demand of
According to IEEE C57.91, the average continuous load a system, or part of a system, to the total connected
is the rms load obtained by Equation 16.3-1 over a cho- load on the system. Demand factor is always less than
sen period of the day. A period of 12 hours preceding one. “Demand factor” is a percentage by which the total
and following the peak is suggested to be considered for connected load on a service or feeder is multiplied to
the time interval of average load calculation. Time inter- determine the greatest probable load that the feeder will
vals (t) of 1 hour are suggested as a further simplifica- be called upon to carry. For example, in hospitals,
tion of the equation, which for a 12-hour period hotels, apartment complexes, and dwelling units, it is
becomes Equation 16.3-2. The dashed line in Figure not likely that all of the loads are connected to every
16.3-2 shows the average load cycle constructed from branch-circuit served by a service or feeder would be
the actual load cycle. “on” at the same time. Therefore, instead of sizing the
12 feeder to carry the entire load on all of the branches, a
Laverage (12h ) = 0.29 ∑L
i =1
2
i 16.3-2 percentage can be applied to this total load, and the
components sized accordingly. Equation 16.3-6 formu-
lates the size of a distribution transformer considering
In fact, the average load determines the kWh billing rev- the incorporated factors:
enue that will be obtained from serving the load,
whereas peak load determines how much system capac- M
N i × kWi × DFi × LFi
ity is required to serve that particular load group. ∑
i =1 PFi 16.3-6
S ( kVA) =
DivF
16.3.5 Load Factor
The ratio of the average demand over a time interval to Where:
the maximum demand over the same time interval is the S is the rated power of transformer.
load factor. N is the number of loads (appliances) of the
same type.
Average Demand Power( kW ) kW is the rated power of each load.
LDF % = × 100 16.3-3
Peak Load ( kW ) DF is demand factor.
LF is the load factor.
Load factor can be calculated daily, monthly, and annu- PF is the power factor of each load.
ally based on the load profile. M is the number of different type of loads.
DivF is the diversity factor.
16.3.6 Load Diversity, Diversity Factor, and
Demand Factor Table 16.3-1 suggests typical values for load factor,
Load diversity is the difference between the sum of the diversity factor, and demand factor of loads (Pabla
individual maximum demands of loads on a system and 2004).

16-10
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

16.3.7 Load Growth In general, the method to determine the maximum

Estimating load growth includes an element of specula- diversified load of a number of houses consists of the
tion. Load growth for each year into the future may be following steps:
estimated from known factors such as planned installa- • Define the type of houses based on major electrical
tion and geographically related load patterns. usage, such as space heating, water heating, and air
conditioning.
If the annual rate of load growth is available, the load
• Identify all loads in the type of home being consid-
growth can be calculated for the transformer useful life-
ered.
time interval. The modified transformer rating is as
shown in Equation 16.3-7. • Determine the value of all Connected Loads (Lk) and
the Maximum Non-Coincident Demand (MNCD).
ST = S (1 + i )n 16.3-7 • Determine the maximum peak load for each house
type.
Where:
S is the calculated power from Equation • Use demand factors to determine the Maximum
16.3-6. Coincident Demand (MCD) for groups of similar
i is the annual growth rate. types of houses.
n is the typical expected transformer life. • Develop charts of number of kW per home vs. num-
ber of homes, and total kW vs. number of homes.
16.3.8 Load Diversity Charts
Residential loads can be analyzed to determine the type
Load diversity considerations account for the fact that
of electrical equipment and its electrical load that would
not all loads connected to the distribution transformer
be connected in typical homes. Electrical equipment
will be drawing power at the same time. Many individual
used in residential homes may be general (e.g., clothes
loads are thermostatically controlled or cycling and
washer, microwave oven, stereo, hair dryer, etc.), high-
therefore are not likely to be turned on at the same
energy consuming (e.g., electric clothes dryer), or ther-
time—that is, not coincident. Transformer loading
mostatically controlled (e.g., refrigerator, air-condition-
needs to accommodate the diversified or coincident load
ing, heating). For an average home, major appliances
as opposed to the total connected load.
consume the most electrical energy (10.3-kWh/day).
Lighting would consume an average of 4.1-kWh/day.
For the purpose of characterizing loads on a distribu-
Homes with air-conditioning units would utilize
tion transformer, it is useful to determine the maximum
7.3-kWh for cooling and motor blower. Houses with
peak load that is likely to occur when a group of similar
electric heating use, would utilize on average about
load types are connected to the transformer. For
120 kWh/day, and average houses with electric water
instance, in practice, it is useful to know the ultimate
heating consume 14.7-kWh/day.
peak load that will result from connecting a number of
similar electrically heated residences to a distribution
Based on the electrical energy equipment and load,
transformer. The total diversified or coincident load on
houses can be classified into different major categories
the transformer will be less than the sum of the maxi-
such as:
mum peak demand of all the residences.
• Natural gas heated with no air conditioning
Table 16.3-1 Typical values for Demand Factor, Diversity • Natural gas heated with air conditioning
Factor, and Load Factor • Natural gas heated with air conditioning and electric
Demand Diversity water heating
Factor% Factor Load Factor%
• Natural gas heating and cooking with air condition-
Domestic 70-100 1.2-1.3 10-15
ing
Commercial 90-100 1.1-1.2 25-30
Industrial • Central electrical heating, electrical water heating
(less than 70-80 - 60-65 with no air conditioning
500 kW)
Industrial
(Above 500 85-90 - 70-80
The definitions and relations for maximum coincident
kW) load, maximum noncoincident load, connected loads,
Municipal 100 1 25-30
Agricultural 15-20 1-1.5 90-100

16-11
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

diversity factor, and demand factor are expressed in A second method for developing diversity charts is
Equations 16.3-8 and 16.3-9. using the “diversity factor” and the relation as shown in
Equation 16.3-10.
Maximum Coincident Demand ( MCD ) = DF . Lk
Maximum Diversified
16.3-8
ΣkWn
( DF1 ) . ⎡⎣1 − P N ⎤⎦ ( Coincident ) Demand =      16.3-10
( Fact )n
Div
( DF )N =   16.3-9
N [1 − P ] Where:
Where: Σ kWn is the sum of the maximum non-diversified
DF is the demand factor. load.
DF1 is the demand factor for one house (ratio of
Maximum Demand to Total Connected Table 16.3-2 is an example of a diversity chart for 1 to 20
Load for one house). houses for different scenarios including air conditioned,
Lk is the total connected load. electric heating, natural gas appliances, etc. The refer-
N is the number of houses. ence size of the house is a range of 1250 to 1750 square
P is the probability that one house has the feet. Larger or smaller homes or with a mix of loads
same Coincident Loads as other houses would require appropriate adjustments to these load
within the same time period. factors. Utilities should develop their own diversity
charts based on their regional loading data.
With these relations, demand factors for different condi-
tions can be established. The demand factor approach was used in Table 16.3-2,
where:
Demand factor for N = 1 is 0.64.
Probability factor is 0.7.

Table 16.3-2 Diversity Chart for 1 to 20 Detached Houses

Transformer Peak Load (kW) for Detached Houses (1250 to 1750 ft2)
Peak
Season Number of Houses
1 2 3 4 5 6 7 8 9 10
Demand Factor 0.64 0.55 0.47 0.41 0.36 0.31 0.28 0.25 0.23 0.21
Natural Gas
Summer 8.6 14.9 19.4 22.7 25.2 27.1 28.6 29.9 30.9 31.8
Heated – No A/C
Natural Gas
Heated – Central Summer 10.2 17.9 23.9 28.9 32.9 36.4 39.4 42.1 44.8 47.2
A/C
Natural Gas
Heated – Natu-
Summer 8.8 15.5 20.9 25.2 28.9 32.2 35.0 37.7 40.2 42.5
ral Gas Stove -
Central A/C
Natural Gas
Heated – electric
Summer 12.8 22.9 31.2 38.1 44.1 49.5 54.4 59.0 63.4 67.6
Water Heater -
Central A/C
Electric Space
Winter 14.8 27.7 39.1 49.6 59.4 68.8 77.8 68.6 95.2 103.7
and Water Heat

16-12
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

16.4 PAD-MOUNT TRANSFORMER SELECTION been used for transformers with 65°C average winding
temperature rise.
16.4.1 Loading Criteria and Transformer Rating
The rated kVA of a transformer is the output that can Top oil temperature alone should not be used as a guide
be delivered for the time specified at rated secondary in loading transformers, because the difference between
voltage and rated frequency without exceeding the spec- top oil and hot-spot copper temperatures varies with
ified temperature-rise limitations and within the limits different designs and with load. Transformers may be
established in the design spec. operated above average continuous hottest-spot temper-
atures (95°C for 55°C rated transformers and 110°C for
Selection of a transformer with an appropriate rating to 65°C rated transformers) for short times, provided they
serve to load should be done by considering several fac- are operated over much longer periods at temperatures
tors, including: below 95°C and 110°C, respectively. According to Equa-
tion 16.4-1, 110°C is the sum of the following: average
• Transformer internal temperatures, such as hottest winding rise (65°C), ambient (30°C), and hot spot rise
spot in the winding, top oil temperature, and average (15°C).
winding temperatures,
• Transformer loss of life, and Two characteristic modes of operation can be identified
• Total lifetime cost of the transformer with respect to the aging of insulation:
• Normal operation—corresponds to the normal life
Hottest-spot, Top Oil Temperatures, and Average Winding
expectancy where the deterioration under varying
Temperature
conditions of load and ambient temperature is nor-
Transformer loading causes heat to be generated due to
mal.
the winding and core losses, which results in a tempera-
ture rise of the oil and solid insulation. In addition, ele- • Overload operation—which is permitted when neces-
vated loading increases the presence of oxygen, moisture, sary without risking the reliability of the transformer.
and their byproducts, and will accelerate the process of
insulation aging. It is, therefore, important to ensure that Loading of transformers above nameplate is a contro-
the temperature rise is kept within the design limits. It is versial subject. Transformers, at some time, may have to
possible to relate normal and abnormal loading to the be overloaded during power system emergencies, in
transformer hottest-spot temperature in order to under- order to preserve system reliability. The maximum con-
stand how loading affects the life of the insulation. tinuous load-carrying capacity of the transformer
depends on its rating, on the temperature of the cooling
The hot-spot winding temperature is the principal factor medium, ambient temperature, and the level of accepted
in determining the degradation of the transformer due insulation aging governed by the effect of temperature
to loading and hence has major bearing on the trans- and time.
former life. The hottest-spot temperature can be consid-
ered as the sum of the temperature of the cooling Overload capacity of a transformer is the maximum
medium, the average temperature rise of the copper, and load for which the transformer can be subjected for a
the hot-spot allowance. It is given by Equation 16.4-1 particular duration and considering a particular ambi-
ent temperature.
θ H = θ A + ΔθT + Δθ H
16.4-1
ΔθT = θT − θ A The overload capacity depends on the average winding
temperature rise that has been used to design the trans-
Where: former. This temperature can be 55°C or 65°C, depend-
θΑ is the average ambient temperature. ing on the standard or request of end user at purchase
ΔθΤ is the top-oil rise over ambient temperature. time.
ΔθΗ is the winding hottest-spot rise over top-oil
temperature. When transformer purchase specifications include over-
loadability requirements for specific load profiles, in
It is not possible to measure the hottest-spot tempera- duration, frequency, and magnitude of overload, the
ture directly in a traditional transformer because of the manufacturer will adjust the design accordingly to guar-
hazards in placing a temperature detector at the proper antee such overload operation as normal, and can also
location. Standard allowances for hottest-spot rise over do so with no loss of life as specified. This design adjust-
top-oil temperature have been obtained from laboratory ment usually results in a more substantial design and/or
tests. A hottest-spot allowance at rated load of 15°C has lower loss unit.

16-13
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

According to IEEE C57.91, normal life expectancy will bushings, leads, soldered connections, and tap changers;
result from operating continuously with hottest-spot and heating of associated equipment such as cables, cir-
conductor temperature of 110°C or with an equivalent cuit breakers, fuses, disconnecting switches, and current
daily transient cycle. Distribution transformer tests indi- transformers are examples of associated equipment.
cate that the normal life expectancy at a continuous hot- Any one of these may constitute the practical limit in
test-spot temperature of 110°C is 20 years. load-carrying ability.
Long-term and Short-time Emergency Overloads If the loading strategy is based on the average winding
The permissible loading of transformers for normal life temperature, as a typical value, for each degree Celsius
expectancy depends on the design of the particular in excess of 5°C that the average winding test tempera-
transformer, its temperature rise at rated load, tempera- ture rise is below 65 °C, the transformer load may be
ture of the cooling medium, duration of the overloads, increased above rated kVA by 1.0%. The 5°C margin is
the load factor, and the altitude above sea level if air is taken to provide a tolerance in the measurement of tem-
used as the cooling medium. ANSI-IEEE C57.92 perature rise. The load value thus obtained is the kVA
(ANSI/IEEE 1981) has developed several permissible load, which the transformer can carry at 65°C rise.
overload graphs for different types of transformers with
respect to a number of factors. Figure 16.4-1 shows a For a very short-time loading that is less than ½ hour, it
typical overload capability curve for oil-immersed trans- is possible to load transformers up to 300%, with the
formers from ANSI C57.92 for ambient temperature of maximum hottest spot of 200oC and top-oil temperature
30oC and oil temperature rise of 65oC. For example, a of 120oC. If the high loading factor continues more than
liquid-filled transformer with a 50% continuous equiva- ½ hour, the insulation aging takes place. It should be
lent base load at 30°C ambient temperature could be clearly understood that, while the insulation aging rate
loaded to 120% of full load nameplate rating for five information is considered to be conservative and helpful
hours without excessive loss of insulation life. in estimating the relative loss of life due to loads above
nameplate rating under various conditions, this infor-
Overloading of transformers should not be practiced mation is not intended to furnish the sole basis for cal-
without investigation of the various limitations culating the normal life expectancy of transformer
involved, other than winding and oil temperature. Oil insulation. The uncertainty of service conditions and
expansion; pressure in sealed- type units; heating of the wide range in ratings covered should be considered
in determining a loading schedule. As a guide, utilities
consider an average loss of life of 4% per day in any one
emergency operation to be reasonable.
Percent Loss-of-Life due to Loading
Aging or deterioration of insulation is a function of
time and temperature. When cellulose ages, the cellulose
chains are cut in a process called chain scission, reduc-
ing the average length of the cellulose chains and result-
ing in shorter fibers. This can be measured by Degree of
Polymerization, or so-called DP. The rate of degrada-
tion is very slow at room temperature. At elevated tem-
peratures, however, the rate of degradation increases
exponentially, effectively doubling for approximately
every 8°C increase in temperature. Because the tempera-
ture distribution in most apparatus is not uniform, the
part that is operating at the highest temperature will
ordinarily undergo the greatest deterioration. Therefore,
it is usual to consider the effects produced by the highest
temperature, or the hottest spot.

Traditionally NEMA developed graphs of % of loss of

life of transformers versus the hottest spot temperature,
Figure 16.4-1 Permissible overload for varying periods of as shown in Figure 16.4-2. The basis of the aging factor
time for oil-filled transformers with 65oC rise based on the modeled by IEEE is the exponential curve of aging ver-
initial load, normal life expectancy, ambient = 30oC (ANSI sus temperature.
C57.92).

16-14
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

IEEE C57.91-1995 (IEEE 1995) has a well-defined types. This standard defines “insulation aging rate”,
model for transformer aging and life of insulation. It FAA, as shown in Equation 16.4-3.
includes a per unit life model to calculate the aging of
⎛ A A ⎞
transformers, as shown in Equation 16.4-2. ⎜
⎜θ
− ⎟
+ 273 θ HS + 273 ⎟⎠ 16.4-3
FAA = e ⎝ HS ,R

PerUnit Life = Ae θH + 273 16.4-2

Where FAA is the insulation aging rate, θHS,R is the refer-
ence hot spot temperature for the insulation, and θHS is
where θΗ is the winding hottest spot in °C, A = 2 × 10−18 the hot spot temperature at which aging is evaluated.
and B is a constant equal to 15,000 for most insulation
A curve of FAA versus hottest-spot temperature for a
65°C rise insulation system is shown in Figure 16.4-3.
FAA has a value greater than 1 for winding hottest-spot
temperatures greater than the reference temperature
110°C and less than 1 for temperatures below 110°C.
Reduced Life Expectancy with Heavy Loading
IEEE C57.91 has defined a method to calculate the
reduced life expectancy based on “aging accelerated fac-
tor”, FAA as shown in Figure 16.4-3. The reduced life
expectancy, RLF , is calculated from Equation 16.4-4.

Feq × t
% RLF = × 100 16.4-4
Normal Life
N
( ∑ FAAn Δtn )
Feq = n =1
N
16.4-5

∑ Δt
n =1
n

Where:
Feq is equivalent aging factor for the total time
period.
N is the total number of intervals.
FAAn is aging acceleration factor for the tempera-
Figure 16.4-2 Loss of life versus temperature for different ture that exists during the time interval
time periods, 65oC rise time (NEMA TR-98-1964). Δ tn .
t is the time period in hours.

Figure 16.4-3 Insulation’s aging acceleration factor (IEEE C57.91-1995).

16-15
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

Normal Life is defined by manufacturer. As a bench- roof gratings, a higher ambient temperature than the
mark for a distribution transformer, normal life is 20 outdoor air is expected. The amount of increase
years for a well-dried, oxygen-free 65oC average winding depends on the design of the manholes and vaults, net
temperature rise insulation at a reference temperature of opening area of the roof gratings, and the adjacent sub-
110oC. surface structures. Therefore, the increase in effective
ambient temperature for expected transformer losses
Unusual Service Condition must be determined before loading limitations can be
A number of factors related to transformer loading are estimated.
considered unusual service conditions such as:
• Increase of ambient temperature Total Lifetime Cost
As discussed in Section 16.2.7, “Cost of Loss Formula,”
• Installation in a height more than 1000 m (3300 ft) the transformer cost has three components: capital
investment, no-load loss, and load loss. If the end-user
The design of distribution transformers usually consid- provides the energy price with the purchase request, the
ers ambient temperature of 30oC. If the average of ambi- designer can develop a transformer design that will min-
ent temperature increases, the loading should be imize the total lifetime cost including the cost of losses.
lowered to keep the normal life expectancy. A guideline The result of this process is the cheapest transformer in
provided by IEEE C57.91 suggests a load de-rating of the useful life period—i.e., with the lowest total owning
1.5% for each o C up to 50 o C. The load is allowed to cost—optimized for a given application.
increase by 1% for each oC lower than 30o C. Average
ambient temperatures can be considered to cover The following considers the total cost of losses for trans-
24-hour periods. The maximum ambient temperature in formers loaded at different fractions of their rating.
24 hours should not be more than 10°C above the aver- Typically a transformer is designed to have a minimum
age temperature. loss when operated at about 50% of rating. However, a
larger transformer operated at a lower fraction of rat-
The effect of the decreased air density due to high alti- ing, may have a smaller cost of losses than a smaller unit
tude is to increase the temperature rise of transformers, operated at 50% of rating. This circumstance will be
because they are dependent upon air for the dissipation particularly true in situations with significant annual
of heat losses. If the transformer is installed at a height load growth.
of 1000 m (3300 ft) above sea level, a de-rating factor
needs to be considered as shown in Figure 16.4-4. The present value of the total cost of losses can be cal-
culated by calculating the loss in each of the next 40
Note that if enough information has been delivered to years and then applying a discount factor to account for
the transformer designer, the effect of de-rating due to inflation, and the cost of capital or the expected rate of
high ambient temperature or high altitude level is usu- return on capital investment. The losses in any one year
ally considered by the designer. Therefore, the name- are calculated as the sum of load and no-load losses.
plate ratings do not need to be de-rated. The no-load power loss is simply the no-load loss
expressed as a percent of rating times the rating of the
For transformers installed in subsurface manholes and transformer. Because the no-load loss is constant, this
vaults of minimum size with natural ventilation through power loss is simply multiplied by the hours in a year to
obtain the energy loss.

The load losses in a transformer vary with the load. The

manufacturer usually states load losses at rated load as
a percentage of transformer rating. The value of loss at
other loads can be estimated by multiplying by the ratio
of the loads squared, because the loss increases with the
square of the current. This procedure ignores the
decrease in loss at lower temperatures caused by the
decrease in resistance as the temperature decreases
(approximately 25% from 90ºC to 20ºC), because this
decrease is small compared to the quadratic decrease.
Figure 16.4-4 Permissible KVA loading and ambient The peak power loss is calculated at the peak load, and
temperature for altitude above 1000 m (ANSI C.57.12.00). the energy loss is calculated at the peak loss (at peak
load) multiplied by the loss factor to give the energy

16-16
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

loss. The loss factor can be an input parameter, or it can on the time for which the peak load occurs, the previous
be calculated from the load factor using an assumed load condition, and the thermal time constant of the
load profile by the empirical equation LF = 0.85(LD2) + transformer. Short time peaks of up to 200% of rating
0.15 LD, where LF is the loss factor and LD is the load can be justifiable.
factor. If the exact load profile of a transformer is
known, such as hourly load for a year, then the loss fac- Optimal transformer sizing can be determined using the
tor can be calculated from the load data, and the loss Diversity Chart and Figure 16.4-5. Using the peak load
calculation will be exact. calculation from Table 7 of IEEE C57, the first vertical
intercept with a transformer plot determines the most
The input parameters to the calculation procedure are: optimum size in terms of lifetime ownership cost.
• Load loss for each transformer rating
Transformer size selection, at any specific load level, is
• Noload loss for each transformer rating controlled by the thermal load limit, not by the cost of
• Cost of losses (kW and kWh) losses. This conclusion depends on the ratio of no-load
• Real discount rate loss to load loss for the particular set of transformers. It
will be true as long as the difference in no-load loss from
• Annual load growth rate one transformer size to the next is larger than the load
• Load factor loss of the smaller size transformer when loaded near its
• Loss factor rating.

Figure 16.4-5 shows the present value of the cost of The overall conclusion is that a utility cannot reduce
losses over a 40-year life versus peak load in the first transformer losses by going to a larger size transformer
year. Single–phase, 4-kV polemount transformer data that will have lower load losses. The minimum loss costs
are used in this example for sizes ranging between 10 are achieved if the smallest possible transformer is
and 100 kVA to provide the widest data coverage. The selected based on thermal loading limits.
lowest losses are often for a transformer that is severely
undersized. To make a reasonable limit on the loading, 16.4.2 Other Parameters for Transformer
the “thermal limits” are shown as vertical dashed lines Selection
based on IEEE C57 – Distribution, Power and Regulat- Selection of the appropriate transformer should also
ing Transformers (Table 7, 2-hour peak load duration at include consideration of:
10 °C and 30° C [winter and summer operation] 65 ° C • Preferred power ratings
rise.). For winter and summer operation, the peak limit
was set at 1.87 and 1.57, respectively. This is not a firm
• Short-circuit capacity
limit, because the loss of life of a transformer depends • Noise level

Figure 16.4-5 Cost of ownership vs. initial load – 4 kV pole transformers –

single phase.

16-17
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

Preferred Power Ratings The symmetrical short-circuit current can be calculated

Despite the selection of an exact power rating that may as follows:
be optimal for an application, distribution transformers
are generally produced in a number of preferred ratings. U
I SC =
3( ZS + ZT )
16.4-7
Preferred continuous kVA ratings of single-phase and IR I
three-phase distribution and power transformers based = ≅ R
ZS % + ZT % ZT %
on an average winding rise by resistance of 65o C are
defined as following: Where:
ISC is the symmetrical short-circuit current.
Single-Phase (kVA): 5, 10, 15, 25, 37.5, 50, 75, 100, 167, IR is the rated current.
250, 333, 500, 800,1250, 1600, 2500, 3300 ZS is the system impedance connected to the
transformer.
Three-Phase (kVA): 15, 30, 45, 75, 112.5, 150, 225, 300, ZT is the transformer short-circuit impedance.
500, 750, 1000, 1500, 2000, 2500, 3750, 5000
Based on IEEE Std C57.12.00, multi-winding trans-
To reduce inventory, some utilities seek to further limit formers shall be considered to have system fault power
the ratings of the transformers that they purchase. supplied at no more than two sets of un-faulted termi-
Short-circuit Capacity nals rated greater than 35% of the terminal kVA of the
Another of the important factors for selecting a trans- highest capacity winding.
former is the short-circuit capacity. Transformers Noise Level
should be designed and constructed to withstand the Transformers in service cause sound, which may cause
mechanical and thermal stresses produced by external discomfort to people in the environment in the long
short circuits. The external short circuits shall include term. This is mainly the problem of power transformers.
three-phase, single line-to-ground, double line-to- However, it can be an issue for large distribution trans-
ground, and line-to-line faults on any one set of termi- formers too. Sound can be defined as the pressure varia-
nals at a time. tion in air that the human ear can detect. The normal
range of hearing of a healthy young person is from
IEEE Std C57.12.00 limits determine the short-circuit approximately 20 Hz to 20 kHz. The weakest sound that
current duration of distribution transformers as shown an ear can detect is dependent on the frequency.
in Equation 16.4-6.

1250 Sound pressure level, LP, expressed in dB, is defined in

ts = if S ≤ 500kVA Equation 16.4-8
I2 16.4-6
ts = 2 if S > 500kVA p2
LP = 10 log 16.4-8
p02
Accordingly, the above standard has determined the
short-circuit withstand capability of distribution trans- Where:
formers based on the symmetrical short-circuit current po is the reference level equal to 20μPa.
shown in Table 16.4-1. p is the sound pressure measured by a micro-
phone.

Table 16.4-1 Short-circuit Withstand Capability (ANSI/IEEE C57.12.00)

Withstand Capability per Unit
of Base Current
Single Phase (kVA) Three Phase (kVA) (Symmetrical)
5-25 15-75 40
37.5-110 112.5-300 35
167-500 500 25
Should be calculated using trans-
Above 500 kVA
former impedance only.

16-18
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

To provide a feeling, a quiet living area has a sound option, it is suggested to order transformers designed at
pressure level of about 45 dB, and a city street with 3 dB below NEMA standard sound levels.
heavy traffic can have 95 dB sound pressure.
Methods are available to the transformer designer to
The dominant generating source of transformer sound control the transformer noise:
is core magnetization. When the magnetic flux changes, • Reducing the core flux density from 1.5 T - 1.6 T to a
the magnetic domains change their directions. There- range of 1.2 T-1.3 T. This can be done either by
fore, when excited by a sinusoidal flux, the core sounds. increasing the core cross section, or by increasing the
In three-phase cores, the changes of magnetic domain number of turns in the winding.
for each core limb do not occur simultaneously, which
means that the whole core is subjected to pulsating dis- • Making a heavier framework for the core
tortions. Comprehensive investigations are made to cor- • Inserting pad of damping material between core lay-
relate human perception of loudness at various ers, or between active part and tank
frequencies and sound pressure. To imitate the response
curves of the human ear, three different filters are Dimensions and Relation between KVA and Size
inserted in the measuring equipment, named A- There is a certain fundamental relationship between the
weighted, B-weighted, and C-weighted filters. They imi- KVA rating of transformers and their physical size. A
tate the curves going through 40, 70, and 100 dB, rather obvious relationship is the fact that large trans-
respectively. For transformers, the frequency spectra of formers of the same voltage have lower loss than smaller
the audible sound consists primarily of the even har- units.
monics of the power frequency; thus, for a 60-Hz power
system, the audible sound spectra consists of tones at As a typical scaling rule, the length, width, and height
120 Hz, 240 Hz, 360 Hz, 480 Hz, etc. A transformer
are scaled as . Where D represents all
“hum” is usually in the range of 100 Hz to 300 Hz.
Depending on other nearby ambient noise, the trans- directions of the dimension.
former sounds might not be noticeable.
To overcome the limitation of the transformer size,
The noise of a transformer is defined as the A-weighted manufacturers have several options, some of which
sound pressure level measured in dB at a specified mea- result in a tradeoff in transformer performance:
suring surface with a sound level meter, and then con-
verted to a sound power, LW, with the formula shown in • Reducing the size of core by using Hi-B material or
Equation 16.4-9. changing the flux density design value, which results
higher core loss and noise
LW = LP + LS 16.4-9
• Reducing the space between windings
Where: • Reducing the oil volume by using thermally upgraded
LS is the measuring surface level in dB. insulation

Table 16.4-2 can be used as a guideline for the noise Each of the above solutions may affect other design
level of distribution transformers up to 5 MVA. As an parameters, which need to be fully evaluated before
manufacturing.

Table 16.4-2 Average Sound Power Level for Distribution Transformers (NEMA TR-1, 1993)
Power(kVA)/Sound
0-50 51-100 101-300 301-500 700-1000 1600 2000 2500 3000 4000 5000
Power (dB)
Oil-Type 48 51 55 56 57 60 61 62 63 64 65
Dry-Type
50 55 58 60 64 66 66 68 68 70 71
Self-cooled (open)
Dry-type
50 55 57 59 63 65 65 66 66 68 69
Self-cooled (sealed)

16-19
 Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

16.5 TRANSFORMER COOLING transfer properties, flash and fire points, dielectric
breakdown, oxidative stability, decomposition, water
16.5.1 Mineral Oils and Alternative Ester Oils solubility, long-term aging, sludging, climatic effects,
Traditionally, transformer dielectric insulating fluid has economics, and maintenance relative to standard
been a refined naphthenic mineral oil that is stable at approved mineral oils. Table 16.5-1 provides a compari-
high temperatures and has excellent electrical insulating son of many of these parameters for mineral oils and
properties. Transformers for indoor use either have been natural ester oils.
a dry type, or have used a less-flammable liquid.
As is evident from Table 16.5-1, natural ester dielectric
Up to the 1970s, polychlorinated biphenyls (PCBs) were oils offer several advantages over mineral oils. These
used as a dielectric fluid, because they are not flamma- advantages include their availability from renewable
ble. PCBs are toxic, and under incomplete combustion, domestic sources, their nontoxicity, their being readily
can form highly toxic products such as furans. Starting biodegradable, and their being non-carcinogenic. Natu-
in the early 1970s, concerns about the toxicity of PCBs ral esters have a higher flash point (i.e., lower volatility),
led to their being banned in many countries. Recently superior thermal conductivity, and no sulphur content,
nontoxic, stable silicon-based or fluorinated hydrocar- and offer a significant reduction in damage to cellulose
bons have been used, where the added expense of a fire- insulation.
resistant liquid offset the additional building cost for a
transformer vault. Any two adjacent conductors form a capacitor. In an
ideal capacitor, the phase difference between an applied
In the early 20th century, there was interest in seed-oil- AC voltage and the current is 90°, and the power dissi-
based coolants, but compared to mineral oils, these had a pated is zero. If the dielectric between the conductors is
higher pour point and inferior resistance to oxidation. less than ideal, the phase difference will be less than 90°,
Synthetic esters found specialty applications where high and some power dissipation will occur. To keep this loss
flash point and lower pour point were desired. However, low, it is desirable to have the dielectric as near to ideal
the high cost of synthetic esters limited widespread use. as is practical.
In the early 1990s, natural esters were revisited due to
environmental regulations. The natural ester products For insulating oils, the value for this characteristic is
developed, shared many of the desirable products of the called the power factor or loss tangent (dissipation fac-
synthetic esters, and were more economical. Combus- tor) and is expressed as a percentage at a specified tem-
tion-resistant vegetable oil-based dielectric coolants and perature. These values are determined experimentally
synthetic pentaerythritol tetra fatty acid esters are becom- and represent trigonometric functions of the angle of
ing increasingly common as alternatives to mineral oil. phase difference. With the particular functions used, a
value of zero would represent a 90° phase difference and
Transformer insulating fluids can be compared based on the ideal condition; therefore, low values are desirable.
features such as: availability, their effect on losses, heat In Table 16.5-1, it can be seen that the natural esters

Table 16.5-1 Transformer Oil Comparison

Thermal Conductivity cal/(cm.sec.

Dielectric BreakdownASTM D1816

Power Factor Dissipation Factor

Specific Heat Cal/g/oC @100oC

Sludging (Oxidation Stability)

ASTM D924 % @ 25oC/100oC

Climatic Effects, Pour Point

Minimum/gap/ impulse kV

Neutralization Number
Flash Point ASTM D93

Fire Point ASTM D92

Long Term Aging,

Specific Gravity

Biodegradation

Water Content

Projected Life
ASTM D2440
(closed cup)

ppm @ 15oC
ASTM D974
mg. KOH/g
(open cup)

ASTM D97
72 h/164h

Aquatic

Years
Type

oC

oC

oC

oC
%

Natural 120%±33%
0.15/3.0 0.92 0.6 4.0 x 10-4 330 ≥350 ≥30/≥20/NA ≤0.2/≤0.2 0.04 75 20+ -18
Ester Oil after 28 days
Mineral 0.45 28% to 49%
0.003/0.06 0.885 3.0 x10-4 145 160 >35/>28/>180 0.01/0.01 ≤0.03 45 20+ -47
Oil @ 20oC after 28 days

16-20
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

have higher power dissipation factor values than the The natural ester oils may not meet some criteria of
mineral oils. standards such as pour point, water content, and sludg-
ing. A separate set of acceptance criteria may be needed
Other disadvantages of the natural esters are higher oxi- for these oils or limits of application (e.g., outdoor
dation, pour points, and water retention. Oxidation and transformers not employing external cooling radiators,
sludging are the weakest points of ester oils. circulating pumps, etc.).

Exposure to atmospheric oxygen can lead to sludging, Compared to standard mineral transformer oil, ester
acid by-products, and finally polymerization of the oils. oils are more costly. The capital cost of a new trans-
Natural esters are often supplemented with anti-oxi- former filled with the new oils is estimated at 1.25 to
dants to address this limitation. In North America, 1.30 times the same transformer containing mineral oil.
transformers are normally sealed, which limits exposure For this price differential, a number of advantages are
to oxygen. Once manufactured, the oils are shipped with cited, usually the higher flash point and lower life cycle
nitrogen blanketing in the container void to prevent oxi- environmental cost (i.e., spills and end-of-life disposal).
dation during transport and storage. Oxidation and
contamination of oil can cause the power factor of an As with any transformer asset, periodic sampling and
oil to rise, so determination of this property may pro- analysis of the oil are recommended as a preventative
vide useful information about used electrical insulating measure and would be part of the life-cycle cost. Even
oil. Because these values vary with temperatures, com- though toxicity of ester oils is low, the rules for cleanup
parisons must always be made at the same temperature. of spills are the same as any other substance. The only
When oils are applied properly, oxidation is a low con- difference is the cleanup cost should be lower because
cern. When specifying ester oils, one should confirm special precautions are not needed compared to hazard-
that the transformer is sealed. ous substances. In terms of medical issues, the MSDS
sheets call only for standard precautions when working
The higher temperature pour point is not deemed a with the eyes —to avoid getting oil in the eyes, inhaling
problem for small, sealed transformers because the the mists, or handling oil if hot.
dielectric properties are maintained. As the oil warms
up after the transformer is energized, its fluid properties Information about long-term aging of the ester oils is
are restored. For outdoor transformers, use in trans- not well known, because the products have not been on
formers with mechanical oil circulation or internal the market long. The longest time in service is about 10
switches may be an issue in very cold climates. Indoor years. However, some aging tests have been performed,
transformers with controlled ambient above the pour and field-sampling tests have been conducted by the
point do not have these restrictions. U.S. EPA. Since 1996, more than 17,000 transformers
have been built with natural ester fluid, primarily distri-
Water content is used to monitor a dielectric fluid’s bution low-power, pad-mounted, and pole-mounted
quality. It is an indicator of possible oil deterioration, types, ranging from 10 kVA up to 10 MVA. In 2001, the
which could, for instance, lead to dielectric breakdown. first medium-power transformer (50 MVA) was retro-
The values used are based on the relative saturation of filled with natural ester oil. Accelerated aging tests per
water in the dielectric fluid. The relative saturation is IEEE C57.100™ (IEEE 1999) show that the paper-
based on the amount of water dissolved in the oil aging range is significantly slower when aged in natural
divided by the total amount of water that the oil could esters vs. mineral oil. Full-scale tests per C57.100
hold at that temperature. The dielectric strength of oil resulted in units lasting between three and four times
starts to fall when saturation reaches about 50%. For the required standard average life. Based on these
petroleum-based oils, 50% at room temperature is 30 to results, it has been calculated that the natural ester
35 mg/kg. Esters hold 500-600 mg/kg water at room tested has a 21oC higher thermal index than mineral oil.
temperature. In a closed system, the affinity of the ester The improved thermal index means longer life at a given
oils for water has been observed to be a desirable trait. temperature or the ability to operate at higher tempera-
Mineral oil lacks this property, leaving water to migrate tures for a given life. An ASTM standard and an IEEE
to the kraft insulating papers. Moisture in the paper maintenance guide have been developed for ester oils.
causes it to age. Residual acid in the paper catalyzes
hydrolysis and degradation of the cellulose results.

16-21
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

16.6 INTERPRETATION OF TEST RESULTS Dissolved Gas Analysis

A number of measurements and tests can be performed A simple interpretation method for dissolved gas analy-
on distribution transformers to assess the condition of sis (DGA) results is the “Key Gas Method,” as shown in
the oil, the solid insulation, the windings, and the trans- Table 16.6-2. It should be noted that small amounts of
former internal construction. Though many of the tests H2, CH4, CO, and CO2 are generated by normal aging.
are relatively simple to perform, interpretation of the
meaning of the test results requires some expertise. IEEE Std. C57.104-1991, “IEEE Guide for the Interpre-
tation of Gases Generated in Oil-Immersed Transform-
16.6.1 Oil Tests Interpretation ers” (IEEE 1991) introduces a four-condition DGA
guide to classify risks to transformers with no previous
Standard methods are available to assess the quality of
problems. This guide uses combinations of individual
oil; however, these oil tests are not commonly used on
gases and total dissolved combustible gas concentration
small rating distribution transformers. IEEE, ASTM,
(TDCG). Table 16.6-3 summarizes the DGA key gas
and other standards do not specify interpretation of the
limits suggested by IEEE. However, these numbers have
oil test results specifically for distribution transformers,
been generated based on power transformer units, and
such as pad-mounted or network transformers. The sug-
no data is available for distribution transformers in this
gested numbers in Table 16.6-1 provide a guideline for
or other standards. Table 16.6-3 assumes that no previ-
interpreting oil test results.
ous tests on the transformer for dissolved gas analysis
have been made or that no recent history exists. If a pre-
Table 16.6-1 summarizes the oil quality test standards
vious analysis exists, it should be reviewed to determine
and recommended limits according to the standards for
if the situation is stable or unstable.
“service-aged insulating oil.”

Table 16.6-1 Recommended Oil Quality Tests for Service-aged Insulating Oil (IEEE Std C57.106-2006)
Interfacial ten- Neutralization
Test Dielectric Strength Dissipation Factor sion (IFT) number (acidity) PCB
ASTM D1816 -97 (1
ASTM D924-99
mm gap) ASTM D-971-91 ASTM D974-92 ASTM 4059-91
Standard @ 25 0C
ASTM 877 [min] [max] [max]
[max]
[min]
1816:23kV
Limit (new oil) 0.1% 35 mN/m 0.03 mg KOH/g 2
877: 26 kV
Limit (service aged
1mm gap:23kV 0.5% 24 mN/m 0.2 mg KOH/g 50
oil)

Table 16.6-2 Key Gas Interpretation Method

Key Gas Secondary Gas Fault Pattern Possible Root Cause
Aging of insulation, possible carbon particles in oil,
Low-energy partial
H2 CH4 and minor C2H6 and C2H4 poor grounding of metal objects, loosed lead, float-
discharge
ing metal, or contamination
Paper insulation destroyed, metal discoloration, oil
C2H4 CH4 and minor H2 and C2H6 Oil overheating
heavily carbonized.
Poor contacts in leads, weakened insulation from
C2H2 H2 and minor CH4 and C2H4 High energy Arcing
aging, carbonized oil.
CO If the fault involves and oil-impreg- Overloading or cooling problem, bad connection in
Conductor overheating
CO2 nated structure CH4 and C2H4 leads, stray magnetic flux, discoloration of paper.

Table 16.6-3 IEEE Dissolved Key Gas Concentration Limits (in ppm)
Status H2 CH4 C2H2 C2H4 C2H6 CO CO2 TDCG
Condition 1 100 120 35 50 65 350 2500 720
Condition 2 101-700 121-400 36-50 51-100 66-100 351-570 2500-4000 721-1920
Condition 3 701-1800 401-1000 51-80 101-200 101-150 571-1400 4001-10000 1921-4630
Condition 4 >1800 >1000 >80 >200 >150 >1400 >10000 >4630

16-22
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

Condition 1: TDCG below 720; satisfactory operation; of the results of a number of failure investigations with
any individual combustible gas exceeding specified levels the gas analysis for each case. Another ratio method is
should prompt additional investigation. the “Doernenburg method,” which is very similar to the
Rogers method with 5 ratios. Another DGA interpreta-
Condition 2: Action should be taken to establish a trend tion technique proposed by IEC 60599 is based on the
quarterly. Duval triangle. This method provides a coded list of
faults detectable by DGA of a faulty transformer.
Condition 3: High level of decomposition; immediate
action should be taken to establish a trend monthly. CIGRE (International Council on Large Electric Sys-
tems), one of the leading worldwide organizations on
Condition 4: Continued operation could result in failure electrical power systems, has reported phenomena
of the transformer. Immediate action required to called “stray gassing,” which occurs when some types of
remove the transformer from service. insulating oils are heated at relatively low temperatures
(100 to 120°C), producing hydrogen or hydrocarbons.
In interpreting DGA, relative gas concentrations are This gas formation seems to reach a plateau after some
found to be more useful than actual concentrations. If a time and then stops. Under certain conditions, stray
possible fault is suspected, a scheme developed by Rog- gassing may interfere with DGA evaluation. CIGRE
ers (IEEE 1991) and later simplified by the IEEE, can be has found that at 120°C, the main gas produced, in gen-
used to define transformer condition. The three-ratio eral, is hydrogen, followed by methane. The production
version of the Rogers Ratio Method uses the following of hydrogen is temperature dependent.
ratios: R 1 = C 2 H 2 / C 2 H 4 , R 2 = CH 4 /H 2 , R 3 = C 2 H 4 /
C2H6. Development of saturated hydrocarbons without fault is
a common issue that can easily be misinterpreted using
Figure 16.6-1 is the flowchart recommended by IEEE to the Rogers or Duval methods. Typical for these cases is
interpret the Rogers Ratio Method. It is important to the production of ethane, ethylene, and methane in high
mention that the gas ratio method is for determining the amounts. The ratio of ethane to ethylene, and especially
possible fault type, not for detecting the presence of a ethylene to propylene, may be higher than 10. Ethane,
fault. The validity of this method is based on correlation ethylene, and methane increase steadily in the first years

Figure 16.6-1 IEEE recommendation for Rogers Ratio Method.

R1= C2H2/ C2H4 , R2= CH4/H2 , R3= C2H4/ C2H6

16-23
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

after commissioning, while the amounts of hydrogen factor (>1.0% at 25 °C) in oil can be caused by the pres-
and ethylene stay constant and low. Such behavior has ence of free water, which could be hazardous to the
been observed in new transformers as well as in old operation of a transformer. Oxidation, free water, wet
ones. The interpretation of DGA usually indicates a hot particles, contamination, and material incompatibility
spot below 150°C; however, the transformers are failure- are all possible sources of high power factor in oil.
free (Duval 2004).
Polychlorinated Biphenyl (PCB)
Dielectric Test According to IEEE Std 62-1995, low polychlorinated
The dielectric test measures the voltage at which oil biphenyl (PCB) concentration (<50 ppm) generally indi-
breaks down electrically. This test can give a good indi- cates an extremely low risk (according to the U.S. EPA),
cation of the amount of contaminants such as dirt, and the oil is classified as noncontaminated. A moder-
water, and oxidation particles. The IEEE guide for insu- ate PCB concentration (50 ppm to 500 ppm) causes the
lating oil equipment prefers the ASTM D-1816 (ASTM oil to be classified as contaminated. Any concentration
2004) dielectric test method rather than the ASTM D- above 500 ppm is considered as if it were pure PCB.
877 (ASTM 2002), because the electrodes are closer to Local governmental regulations and environmental leg-
those in real application, and the test is more sensitive to islations may require specific values of even lower than
moisture than the ASTM D-877. If ASTM D 877 is 50 ppm. Some regulations do not allow moderate con-
used instead of ASTM D1816 for dielectric strength, the centration (50 ppm to 500 ppm) near sensitive areas
limit is 26 kV rather than 23 kV. If a 2 mm gap is used such as a hospital, food or feed processing plant, senior
for ASTM D1816, 40 kV is recommended. care facility, pre-school/daycare, or a school. The term
“Non-PCB” means PCB free from origin of manufac-
It should be noted, however, that high dielectric strength ture and tested out at less than 1 ppm PCB.
is no guarantee that the oil is not contaminated. Tests
on oil from a failed transformer are not indicative of the If a high level of PCBs was detected, the oil needs to be
oil quality just before failure, because carbon and debris retrofilled. To reduce the PCB concentration in the core
from the failure will be suspended in the oil. and coil of a PCB-contaminated transformer, the con-
taminated oil is drained out, and new replacement oil is
Although rarely performed, carbon and other particu- put in its place—a process called “retrofilling.” The only
late matter can be removed by filtration methods prior time that it would be logical to retrofill a transformer to
to dielectric testing. reclassify it to non-PCB status is if the transformer has a
reasonable life expectancy. As a routine, all transformers
Power Dissipation Factor
that come out of service should be sampled and analyzed
The dissipation factor is a measure of the power lost
for PCBs before they are repaired, disposed, or recycled.
when an electrical insulating liquid is subjected to an ac
Retrofills cannot reach a level of 1 ppm; more likely, less
field. The power is dissipated as heat within the fluid. A
than 50 ppm PCB is more reasonable due to leach-back
low-value dissipation factor means that the fluid will
from 10% typical retained oil volume in saturated insula-
cause little of the applied power to be lost. The test is tion. Less than 50 ppm PCB is not the same as non-PCB,
used as a check on the deterioration and contamination and in some ways is handled differently.
of insulating oil because of its sensitivity to ionic con-
taminants. ASTM D924 (ASTM 2008a) is a reference
for this test. This test may be satisfactorily performed in Table 16.6-4 Maximum Suggested Dissipation Power
the field, as well as in a laboratory environment. A Factors for Different Categories of New and Service Aged
visual check should be performed to ensure that the Oils (IEEE Std 62)
sample does not contain air bubbles due to agitation Power Factor Power Factor
during transport. Type of Oil @25 oC @100 oC
New oil as received 0.05 0.3
The maximum recommended levels of percent power New oil in new trans-
0.15 1.5
former
factor for different categories of new and service aged
New oil after filling the
oils are shown in Table 16.6-4, according to IEEE Std transformer, prior to ener- 0.1 -
62-1995 (IEEE 1995a). gizing
Service-aged oil 0.5 -
High levels of power factor (>0.5% @ 25 °C) in oil are
of concern, because contaminants can collect in areas of
high electrical stress in the winding. Very high power

16-24
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

Acid Number and Inter Facial Tension (IFT) Power Factor

ASTM D974 (ASTM 2008b) is the traditional color- In general, power factor measurement equipment comes
change indicator method of titrating the acids with a with three basic modes of operation: grounded speci-
mild (0.1 N) KOH solution. On some service-aged liq- men test, grounded specimen with guard, and
uids, the color may be so dark as to impair the ability of ungrounded. The three measurement modes allow mea-
the technician to determine the indicator color change surement of the current leaking back to the test set on
in ASTM D974, so ASTM D664 (ASTM 2009) is used each lead, individually and together. In general, a power
instead. IEEE maximum acceptance value for acid num- factor of less than 1% is considered good; 1-2% is ques-
ber is 0.2 mg KOH/g. tionable; and if it exceeds 2%, action should be taken.
Practically, the evaluation is not only based on a single
Acceptable limits for IFT vary with operating voltage. power factor data point but is also based on the history
For a service-aged oil, the minimum acceptance value is of the change in power factor. Values obtained at the
24 mN/m. For oils in service, a decreasing value indi- time of the original tests are used as benchmarks to
cates the accumulation of contaminants, oxidation determine the amount of insulation deterioration on
products, or both. subsequent tests.

16.6.2 Transformer Tests Interpretation The power factor of an insulation system should not
increase with an increase in applied ac voltage. If it does
Insulation Resistance and Polarization Index increase as the ac voltage is increased, there is a problem
The purpose of the transformer insulation resistance test in the insulation system. Another value of the power fac-
is to measure the condition of a “major” insulation sys- tor measurement is that it will detect voids in the insula-
tem—i.e., the insulation between a winding and ground tion system that may be causing high partial discharges.
(core) or between two windings. IEEE C57.125-1991 rec- Table 16.6-6 is a guideline to interpret the insulation
ommends 500 V, 1000 V, or 2500 V DC to be applied to power factor test. The tests can be done, respectively, on
the transformer winding. The resistance of each mea- high-voltage winding to ground, high- to low-voltage
1.5UW winding, and low-voltage winding to ground.
surement should not be smaller than R = . R is
KVA Table 16.6-6 Power Factor Test Interpretation
in MΩ measured at 20 0C, and UW is the winding voltage Power Factor Insulation Condition
in kV. If the winding is Y-connected, then UW is the Dangerous
Above 2.0%
phase-to-ground voltage. If it is Delta-connected, then wet transformer
UW is equal to phase-to-phase voltage. KVA is the rated 1.0 – 2.0 Investigate
power of the winding under the test. Megaohm meter 0.5 – 1.0 Deteriorated
test results below this minimum value would indicate Less than 0.5 Good
probable insulation breakdown.
Turns Ratio
If a transformer passes the insulation resistance test, The purpose of a turn-ratio test basically is to diagnose
before applying any overvoltage test, it is recommended a problem in the winding turn-to-turn or shorted multi-
to do a Polarization Index (PI) test. The polarization turn insulation system in a transformer. This test detects
index is a ratio of the Megohm resistance at the end of a primarily inner winding short circuits. A very low volt-
10-minute test, to that at the end of a 1-minute test at a age ac source is used to determine the turn ratio. Two
constant voltage. Another common way for PI calcula- windings on one phase of a transformer are connected
tion is the ratio of resistance readings that are taken 15 to the instrument, and the internal bridge elements are
and 60 seconds after connecting the voltage. Table varied to produce a null indication on the detector, with
16.6-5 is a guide to interpretation of the PI test results. exciting current also being measured in most cases.
Measured ratios should compare with ratios calculated
Table 16.6-5 Test Interpretation from nameplate voltage to within 0.5%, but should
Polarization Index Insulation Condition compare even closer to actual benchmark values. Out-
Less than 1 Dangerous of-tolerance readings should be compared with prior
1.0 - 1.1 Poor tests. The turn-ratio test may also detect high-resistance
1.1 - 1.25 Questionable connections in the lead circuitry or high contact resis-
1.25 - 2.0 Fair tance in tap changers by higher excitation current and a
Above 2.0 Good difficulty in balancing the bridge.

16-25
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

Winding Resistance former tank wall. If multiple sensors are used, the PD
Winding resistance is used to indicate the winding con- can be located based on the arrival time of the pulses at
ductor and tap changer contact condition. The test the sensors.
requires an ohmmeter capable of accurately measuring
resistance in the range of 20Ω down to fractions of an In the field, the test can be done on-line or off-line. For
ohm. Resistance measurements can be used to check for the off-line test, a three-phase source is required to
proper connections and to determine if an open-circuit apply the voltage. On-line PD measurement can be
condition or a high-resistance connection exists in par- employed using acoustic sensors, via busing tap, or
allel conductor windings. On three-phase transformers, through high-frequency current transformers (HFCT)
measurements are made on the individual windings located either on bushing tap or in the neutral of trans-
from phase to neutral, when possible. On delta connec- former. Figure 16.6-2 shows a PD resolved pattern on
tions, there will always be two windings in series, which the left, recorded using an HFCT sensor via neutral
are in parallel with the winding under test. Therefore, on cable. A classification technique is employed to separate
a delta winding, three measurements must be made to the contributions of PD from those generated by distur-
be able to calculate each individual winding resistance. bances. Each PD pulse waveform is acquired, and the
Winding resistance varies with oil temperature. Because so-called equivalent time-length and bandwidth are
the resistance of copper varies with temperature, all test evaluated and plotted on the TF map, as shown in Fig-
readings must be converted to a common temperature ure 16.6-2 (b).
to give meaningful results. Most factory test data is con-
verted to 85oC. This has become the most commonly 16.7 CAPACITORS
used temperature. Variations of more than 5% may indi-
cate a damaged conductor in a winding. 16.7.1 Purpose of Capacitors
Partial Discharge Loads on electric utility systems include two compo-
For large power transformers, the partial discharge (PD) nents: active power (measured in kilowatts) and reactive
test is performed in the laboratory as a routine test, power (measured in kilovars). Active power is gener-
although a PD test is not required for quality control of ated, whereas reactive power can be provided by either
distribution transformers. However, the PD test is well generation or capacitance. Distribution systems have
known as a diagnostics tool and can be employed to VAR requirements, because distribution power lines and
detect minor and progressing problems leading to a cat- loads are primarily inductive. Uncompensated distribu-
astrophic fault inside a transformer. The two commonly tion systems operate at lagging power factor, drawing
used PD detection methods are: detection of the acous- reactive power from generation.
tic signals, and measurement of the electrical signals
produced by the PD. The acceptable PD limits for new Fixed and switched capacitors are inexpensive means of
transformers are dependent on the voltage and size of providing VAR compensation for distribution systems
the transformers and range from 100 to 500 pC. PD and thus correcting power factor and reducing system
pulses generate mechanical stress waves that propagate losses. Shunt capacitors supply reactive current to
through the surrounding oil. To detect these waves, oppose the out-of-phase component of the current
acoustic emission sensors are mounted on the trans- required by an inductive load. A shunt capacitor draws

Figure 16.6-2 PD measurement using HFCT.

16-26
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

leading current, which counteracts the lagging compo- between the foil. Recently manufactured capacitors have
nent of the current at the point of its installation. all-film-insulating layers. The rolls are packed tightly in
the can, and the can is filled with a dielectric fluid. The
When shunt capacitors are applied, the magnitude of packs are connected in series and parallel using tabs
the source current can be reduced, the power factor can connected to the foil to obtain the desired capacitance.
be improved, and the voltage drop can also reduced. Connection to capacitor elements is generally by means
Capacitors can provide effective cost-reduction by of mechanical crimps or ultrasonic welds.
deferring or eliminating investment in new plant.
Capacitors aid in minimizing operating expenses and Terminal leads are connected to the tabs and exit
allow utilities to serve new loads and customers with a through the bushings to form the exterior connections.
minimum system investment Advantages of installing Capacitor bushings are generally processed porcelain
shunt capacitors in distribution systems are as follows: and are welded to the top of the case and the hermeti-
• Released system capacity. The installation of shunt cally sealed system.
capacitors decreases the reactive power demand from
the generation. Thus, generation, transmission, and Capacitors nameplates generally include the following
distribution substation capacities are released. information:

• Reduction in losses. The reactive components of line • Name of manufacturer

currents are reduced from the point of the capacitor • Unique serial number
installation back to the generator. Dollar savings are • Catalog number
realized from peak power and energy loss reductions.
• Year of manufacture
• Improvement in voltage regulation. The demand
capacity of distribution feeders is often limited by the • Rated capacitance
voltage drop along the line rather than by the thermal • Rated rms voltage
ampacity of the conductor. The installation of shunt • Number of poles
capacitors will improve the voltage profile of the
feeder. An additional benefit from improving the • Rated frequency Rated BIL
voltage profile is the ability to practice conservative • Amount of fluid, indicate flammable or not flamma-
voltage reduction (CVR) or peak shaving from which ble
further demand savings can be achieved.
Capacitors are rated for line-to-line voltage in the event
Depending on the uncorrected power factor of the sys- that they are applied on ungrounded or poorly
tem, the installation of capacitors can increase substa- grounded systems.
tion capability for additional load by as much as 30%,
and can increase individual circuit capability, from the Capacitor units are capable of continuous operation
voltage regulation point of view, approximately 30 to over an ambient range of -50°C to +55°C, provided that
100%. Furthermore, the current reduction for trans- the following limitations are not exceeded:
formers, distribution lines, and equipment can reduce • 135% of nameplate KVAR
the load on these kilovoltampere-limited apparatus and
consequently delay new facility installations. • 110% of rated voltage rms, including harmonics
• 180% of rated current rms, including fundamental
The preceding benefits can be achieved by both fixed and harmonic currents
and switched capacitors. With a variable capacitor, the
benefits can be further enhanced by closely matching 16.7.3 Application of Capacitors at Stations and
the VAR requirements of the load. If control of a vari- on Feeders
able capacitor can be achieved quickly, transient-free Capacitors are used in distribution stations or on distri-
switching and voltage flicker reduction are additional bution feeders. Station capacitors are rack mounted in
benefits. large banks. Capacitors installed on feeders are usually
in pole-top banks with necessary group fusing. The
16.7.2 Description of Capacitors maximum bank sizes are about 1800 KVAR at 15 kV
Distribution capacitors are typically housed in rectan- and 3600 KVAR at higher voltage levels. Usually, utili-
gular, sealed, metal cans, which can be made of stainless ties do not install more than four capacitor units on
steel. The cans contain rolled packs of aluminum foil, each feeder. Approximately 60% of capacitors are
with layers of insulating paper, and/or plastic film,

16-27
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

applied to feeders, 30% to the substation bus, and 10% • On feeders with light load, where the minimum load
on the transmission systems. per phase beyond the capacitor bank does not exceed
150% of the per-phase rating of the capacitor bank
Capacitors can be applied as fixed or switched units. • On feeders with single-phase breaker operation at the
Switched units have capacitor bank controllers that sending end
switch several capacitor banks. Such controllers can
switch capacitor banks at the point of installation or • On fixed-capacitor banks
based on a user-specified time schedule. There are also • On feeder sections beyond a sectionalizing-fuse or
controllers that switch capacitors on the zero crossing of single-phase recloser
voltage in order to reduce transients. The components • On feeders with emergency load transfers.
required for a switched capacitor installation operating on
VAR conditions are as follows: Usually, grounded-wye capacitor banks are employed
• Capacitors only on four-wire, three-phase primary systems. Other-
• Oil switch wise, if a grounded-wye capacitor bank is used on a
three-phase, three-wire ungrounded-wye or delta sys-
• Surge arrester tem, it furnishes a ground current source that may dis-
• Current transformer turb ground relays.
• Potential transformer
The optimum amount of capacitor kilovars to employ is
• Transducers to convert CT and PT values into suit- generally the amount at which the economic benefits
able signals for capacitor controller obtained from the addition of the last kilovar equals the
• Capacitor bank status installed cost of the kilovars of capacitors. The methods
• Local/remote switch status used by the utilities to determine the economic benefits
derived from the installation of capacitors vary from
• Local/remote relay control company to company, but usually they all determine the
total installed cost of a kilovar of capacitance.
Capacitors can be applied at almost any voltage level.
Individual capacitor units can be added in parallel to The economic benefits that can be derived from capaci-
achieve the desired kilovar capacity and can be added in tor installation can be itemized as:
series to achieve the required kilovolt voltage.
• Released generation capacity.
A three-phase capacitor bank on a distribution feeder • Released transmission capacity
c a n b e c o n n e c t e d i n d e l t a , g ro u n d e d - w y e, o r • Released distribution substation capacity
ungrounded-wye. The type of connection used depends
upon: • Reduced energy (copper) losses

• System type—i.e., whether it is a grounded or an • Reduced voltage drop and consequently improved
ungrounded system voltage regulation
• Fusing requirements • Released capacity of feeder and associated apparatus
• Capacitor-bank location • Postponement or elimination of capital expenditure
due to system improvements and/or expansions
• Telephone interference considerations
• Revenue increase due to voltage improvements
A res o na n ce c on di t i on m ay o c cu r i n de lt a a nd
ungrounded-wye banks when there is one- or two-line The total yearly benefit due to the installation of capaci-
open-type fault that occurs on the source side of the tor banks can be summarized as

∑
capacitor bank. The resonance occurs due to the main-
Δ$ = Δ$G + Δ$T + Δ$DS + Δ$DF + Δ$LR + Δ$EC
tained voltage on the open delta, which backfeeds any
transformers located on the load side of the open condi- 16.7-1
tion through the series capacitor. As a result of this con- Where:
dition, the single-phase of distribution transformers on Δ$G = annual benefit from generation capacity
four-wire system s may be damaged. Therefore, released above capacity at original pf,
ungrounded-wye capacitor banks are not recommended ($/yr).
under the following conditions: Δ$T = annual benefit from transmission capacity
released above capacity at original pf,
($/yr).

16-28
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

Δ$DS = annual benefit from distribution station selected to match the remaining load characteristics
capacity released above that at original pf, from hour to hour.
($/yr).
Δ$DF = annual benefit from distribution feeder A rule of thumb is often used to determine the size of
capacity released above that at original pf, the switched capacitors. Switched capacitors are added
($/yr). until:
Δ$LR = annual benefit from reduction in energy
k var from switched + fixed capacitors
losses, ($/yr). ≥ 0.70
Δ$CE = annual additional revenue from increased k var of peak reactive feeder load
16.7-3
consumption by voltage improvement,
($/yr). The kilovars needed to raise the voltage at the end of the
feeder to the maximum allowable voltage level at mini-
The total benefits obtained should be compared against mum load is the size of the fixed capacitors that should
the annual equivalent of the total cost of the installed be used. On the other hand, if more than one capacitor
capacitor banks. The total cost of the installed capacitor bank is installed, the size of each capacitor bank at each
banks can be found from location should have the same proportion, that is:
$TIC = ΔQc ⋅ $IC ⋅ ic 16.7-2
k var of load center kVA of load center
= 16.7-4
k var of total feeder kVA of total feeder
Where:
$TIC = annual equivalent of total cost of installed The resultant voltage rise must not exceed the light-load
capacitor banks, $/yr. voltage drop. The approximate value of the percent volt-
ΔQc = required amount of added capacitance, age rise is:
KVAR.
$IC = cost of installed capacitor banks, $/KVAR. Qc ⋅3ϕ Xl
ic = annual fixed charge rate applicable to % VR = 16.7-5
10 ⋅VL2− L
capacitors. Where:
% VR = percent voltage rise.
If only fixed-type capacitors are installed, the utility will Qc⋅3φ = three-phase reactive power due to fixed
experience an excessive leading power factor and volt- capacitors applied, KVAR.
age rise at low-load conditions. Therefore, some of the X = line reactance, Ω.
capacitors should be installed as switched-capacitor l = length of feeder from sending end to fixed-
banks, so they can switched off during light-load condi- capacitance location, mile.
tions. Thus, the fixed capacitors are sized for light load VL-L = line-to-line voltage, kV.
and connected permanently. Switched capacitors can be
switched as a block or in several consecutive steps as the If the fixed capacitors are applied to the end of the
reactive load becomes greater from light-load level to feeder, and if the percent voltage rise is already deter-
peak load, and sized accordingly. mined, the maximum value of the fixed capacitors in
KVARs can be determined from:
A system analysis is required in choosing the type of
capacitor installation. As a result of load flow program 10(%VR ) VL2− L
runs on feeders or distribution substations, the system’s Max Qc ⋅3ϕ = 16.7-6
Xl
lagging reactive loads (i.e., power demands) can be
determined, and the reactive power in KVARs can be
plotted against time of day. This plot is called the reac- The %voltage rise equation above can also be used to
tive load duration curve, and is the cumulative sum of the calculate the percent voltage rise due to the switched
reactive loads (e.g., fluorescent lights, household appli- capacitors. Therefore, once the percent voltage rises due
ances, and motors) of consumers and the reactive power to both fixed and switched capacitors are found, the
requirements of the system (e.g., transformers and regu- total percent voltage rise can be calculated as:
lators). Once the daily reactive load duration curve is ∑ % VR = % VRF + %VRSW 16.7-7
obtained, then the size of the fixed capacitors can be Where:
determined to meet the minimum constant reactive load Σ % VR = total percent voltage rise.
requirements. The remaining kilovar demands of the % VRF = percent voltage rise due to fixed capac-
loads are met by the generator or preferably by the itors.
switched capacitors. Switched capacitor sizes can be % VRsw = percent voltage rise due to switched
capacitors.
16-29
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

Another rule of thumb sometimes used is that: The total itive line loss from the total inductive line loss. Use
amount of fixed and switched capacitors for a feeder is the capacitor kilovars determined in steps 3 and 6,
the amount necessary to raise the receiving-end feeder and find the line loss in each line section due to
voltage to maximum at 50% of peak feeder load. capacitors.
8. To find the distance to capacitor location, divide total
16.7.4 Capacitor Location inductive line loss by capacitive line loss per thou-
Once the kilovars of capacitors necessary for the system sand feet. If this quotient is greater than the line sec-
are determined, the location of the capacitors needs to tion length:
be determined. The rule of thumb for locating the fixed • Divide the remaining inductive line loss by capaci-
capacitors on feeders with uniformly distributed loads is tive line loss in the next line section to find the loca-
to locate them approximately at two-thirds of the dis- tion.
tance from the substation to the end of the feeder. For
• If this quotient is still greater than the line section
the uniformly decreasing loads, fixed capacitors are
length, repeat step 8a.
located approximately halfway out on the feeder. The
location of switched capacitors is often determined by 9. Construct a voltage profile for the feeder. If the pro-
voltage regulation requirements, and they are usually file shows that the voltages are inside the recom-
located on the last one-third of the feeder away from the mended limits, then the capacitors are installed at the
source. location of minimum loss. If not, then use engineer-
ing judgment to locate them for the most effective
The best location for capacitors can be found by opti- voltage control applications.
mizing power loss and voltage regulation. A feeder volt-
age profile study is required to determine the most Some summary rules that can be used in the application
effective location for capacitors and a voltage that is of capacitor banks include the following:
within recommended limits. Usually, a 2-V rise on cir- 1. The location of fixed shunt capacitors should be
cuits used in urban areas and a 3-V rise on circuits used based on the average reactive load.
in rural areas are the maximum voltage changes that are 2. There is only one location for each size of capacitor
allowed when a switched-capacitor bank is placed into bank that produces maximum loss reduction.
operation. A general iteration process is summarized as
3. One large capacitor bank can provide almost as much
follows:
savings as two or more capacitor banks of equal size.
1. Obtain circuit and load information:
4. When multiple locations are used for fixed-shunt-
• kilovoltamperes, kilovars, kilowatts, and load capacitor banks, the banks should have the same rat-
power factor for each load ing to be economical.
• desired corrected power of circuit 5. For a feeder with a uniformly distributed load, a
• feeder circuit voltage fixed-capacitor bank rated at two-thirds of the total
• a feeder circuit map that shows locations of loads reactive load and located at two-thirds of the distance
and presently existing capacitor banks out on the feeder from the source gives an 89% loss
reduction.
2. Determine the kilowatt load of the feeder and the
power factor. 6. The result of the two-thirds rule is particularly useful
when the reactive load factor is high. It can be
3. Determine the kilovars per kilowatts of load neces-
applied only when fixed shunt capacitors are used.
sary to correct the feeder-circuit power factor from
the original to the desired power factor. 7. In general, particularly at low reactive load factors,
some combination of fixed and switched capacitors
4. Determine the individual kilovoltamperes and power
gives the greatest energy loss reduction.
factor for each load or group of loads.
8. In actual situations, it may be difficult, if not physi-
5. Determine the kilovars on the line.
cally impossible, to locate a capacitor bank at the
6. Determine the line loss in watts per thousand feet due optimum location; in such cases, the permanent
to the inductive loads determined in steps 4 and 5 location of the capacitor bank ends up being sub-
above. Multiply these line losses by their respective optimum.
line lengths in thousands of feet. Repeat this process
for all loads and line sections, and add them to find 16.7.5 Capacitor Protection Considerations
the total inductive line loss.
The main function of capacitor protection is to electri-
7. If there are capacitors presently on the feeder, per- cally remove failed capacitors from the distribution sys-
form the calculation of step 6, but subtract the capac- tem.

16-30
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

The protection must isolate a faulted bank or individual The number of capacitor groups in series is an impor-
shunt capacitors without interrupting service on the tant factor in determining the appropriate type of fuse.
remainder of the circuit. When the capacitor does fail, The impedance of the series groups limits the current
the protection should rapidly remove it from the system into a faulted unit and thus determines the magnitude
to avoid case rupture. If the protection has been of the available fault current into a single shorted can.
properly coordinated, it should also operate before any As a general rule, the fault current through the fuse,
other upstream protective devices. While fulfilling this when the unit that it is protecting becomes shorted,
fault-clearing role, the capacitor protection must also should not be less than 10 times the rated capacitor cur-
remain immune to a number of “normal” transient rent. This available fault current is also affected by
conditions such as energizing inrush, discharging/out- whether or not the neutral is grounded.
rush, parallel switching outrush, and lightning surges.
The number of series units in a capacitor installation
To ensure that capacitor protection will fulfill these also affects the overvoltage that healthy units are
functions, a number of issues must be considered as out- exposed to after the short-circuit failure of one series
lined in the following sections. unit. This factor is discussed later in the section entitled
“Overvoltage Protection.”
Location Constraints
Within substations, capacitors are usually individually The number of cans in a parallel group is also an impor-
fused. Capacitor fuses will typically be installed on out- tant consideration in choosing appropriate protection.
door steel structures, which permits the use of any out- Energy stored in the capacitors in parallel with a faulted
door protection option. However, it is also possible to will be discharged into the faulted unit. This discharge
purchase capacitor banks with under-oil fuses installed must be withstood by the fuses on the good cans. When
inside each capacitor unit, and fuses available in encap- a large bank is desired, it may be better to use a double-
sulated designs may be specified for this application. Y construction so as to retain the use of expulsion fuses.
There is also a minimum number of capacitors that can
For feeder installations, capacitors are most often be connected safely in parallel in a group. Below this
located on overhead systems, due to the inherent capaci- critical number, individual capacitor fuses must be rated
tance of underground cable systems, so the protection at such a large percentage of the total phase current that,
equipment can be located at a pole-top location. Out- in the event of failure of a unit, the magnitude of the
door protection options can, therefore, be specified, fault current is insufficient to produce rapid fuse clear-
ranging from distribution cutouts through solid mate- ing. Figure 16.7-1 illustrates the effect of the number of
rial power fuses to current-limiting fuses. series sections and the number of parallel units in a sec-
Bank Configuration tion on the available current through a shorted unit.
Although capacitor units can be connected in several
different configurations, the majority of power capaci-
tor equipment installed on primary distribution feeders
is connected three-phase, either grounded-wye or
ungrounded-wye (delta and single-phase connections
are usually made only on low-voltage circuits). A num-
ber of advantages can be derived from the grounded-
wye type of connection. With the grounded-wye con-
nection, tanks and frames are at ground potential,
which provides additional personnel safety. Grounded-
wye connections facilitate faster operation of the series
fuse in the event of a capacitor failure. Grounded capac-
itors can divert some line surges to ground and, there-
fore, exhibit a certain degree of self-protection from
transient voltages and lightning surges. The grounded-
wye connection also provides a low-impedance path for
harmonics.

In general, phase-neutral rated capacitors should be

used on grounded capacitor banks, and phase-phase
rated banks should be used on ungrounded-wye or delta Figure 16.7-1 Current through a shorted unit versus the
number of units per section for a grounded-wye bank.
systems.

16-31
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

Individual versus Group Protection For a three-phase grounded-wye application, the con-
Capacitor protection practices at distribution voltages tinuous current will be:
can be divided into two basic protection techniques:
individual protection and group protection. 3ϕ kVAR
I C = 1.35 16.7-9
3 kV LL
Individual protection is commonly used for large capac-
itor banks, which are normally located at the distribu- Whenever possible, the lowest rated fuse that can con-
tion substation. In these installations, each capacitor tinuously carry this current should be selected, because
unit is protected by its own individual fuse; backup pro- this provides maximum sensitivity for high impedance
tection in the form of a circuit breaker or higher-rated faults and the greatest protection against tank rupture.
fuse is normally provided to protect capacitors against However, a fuse selected in this way will be more vulner-
bus faults ahead of the individual fuses. Fuses in this able to transient surges. Note that the continuous cur-
case are of the bus-mounted type. rent-carrying capability is not necessarily the same as
the rated current of the fuse. Some fuses will continu-
Group protection is commonly used to protect pole- ously carry currents above their rating. K and T links,
mounted capacitor banks, which are normally located for instance, will carry 150% of their rated current.
on the primary feeders. In this case, only one fuse per
phase is used, and each fuse protects all capacitor units Transient Currents
that are located in that phase. A capacitor fuse must withstand, without damage, the
transient currents and voltages due to lightning surges,
Continuous Current as well as transient currents during energization and de-
Although capacitors are considered constant current energization of the capacitor bank. In addition, it must
devices, in actual operation they are subject to overcur- withstand discharge currents and parallel switching
rents. These are caused by overcapacitance, operation at transients.
higher than rated voltage, and system harmonics.
ANSI/IEEE Standard C37.99 (ANSI/IEEE 1990) Figure 16.7-2 illustrates the various types of transient
allows a manufacturing tolerance of +15% on the rated currents and provides a reference for the symbols used
reactive power of capacitors at rated voltage and fre- in the equations that follow. Note that the capacitors on
quency at 25oC. Also, capacitor banks can be operated one phase are shown with individual protection.
at up to 10% overvoltage (though typical system volt-
ages do not exceed 6% of nominal voltage). These two
factors may combine to increase continuous current by
up to 25%. Harmonic currents depend on system condi-
tions and are difficult to predict; however, practice dic-
tates that an allowance of 5 to 10% of rated current
should be used. Ungrounded-wye or delta connected
capacitors need less margin for harmonic currents,
because there is no path for third or multiples of third
harmonic current.

In capacitor application, it is common to consider that

the continuous current may be equivalent to 135% of
the capacitor rated current for grounded-wye connected
banks, and 125% for ungrounded-wye banks. This
accounts for the effect of overvoltage conditions, capac-
itance variations, and harmonic currents.

The continuous current for an individual grounded-wye

connected capacitor can be calculated as follows:

1ϕ kVAR
I C = 1.35 16.7-8
kV LN

Figure 16.7-2 Illustration of parameters used in

capacitor transient current equations.

16-32
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

In individually fused substation capacitor banks, tran- Where:

sients due to lightning surges will typically be of little R1 = resistance for an individually fused capaci-
concern, because of the reliable substation shielding and tor unit, (ohms).
because the large number of capacitors in parallel will C = capacitance of a single unit, (F).
effectively share the transient current. Currents generated V = voltage, (V).
by switching transients are also controlled in substation
applications through the use of current-limiting reactors Table 16.7-1 provides some typical I2t values for single
and switch-closing resistors. Individually fused units are capacitors discharging from full voltage. When capaci-
generally exposed to only one significant form of tran- tors are connected in parallel, the actual discharge I2t
sient current—that is, discharge or outrush currents. from healthy units into a failed unit is typically 66% of
the tabulated values.
For feeder capacitors, where group capacitor protection
Table 16.7-1 I2t for Capacitor Discharge
is typically used, a number of transient considerations
are of concern, including energizing transients or inrush I2t (times 1000 A2s)
currents, parallel switching (outrush) transients, de- Unit 100 150 200 300
Volts KVAR KVAR KVAR KVAR
energizing transients, and transients due to lightning
2400 10.4 18.6 25.0 -
surges.
4160 8.9 15.9 21.0 -
4800 8.5 15.0 20.2 -
The various forms of transient currents experienced by
7200 6.9 12.3 16.6 25.4
capacitors in normal operation are described in the
8320 6.3 11.1 15.4 23.2
following paragraphs.
12470 4.5 8.1 11.5 17.8
Discharge or Outrush Into Faulted Capacitor 13280 4.2 7.5 10.9 17.2
Discharge currents in capacitor banks occur when one 13800 4.0 7.3 10.6 16.8
parallel unit fails and the remaining good capacitors 14400 3.9 6.9 10.3 16.2
discharge into the faulted unit. To prevent spurious fuse 16000 3.5 6.1 9.2 15.0
blowing and the disruptive failure of the capacitor case, 19920 2.5 4.5 7.2 12.5
the fuses on the healthy units must be capable of
withstanding these outrush currents. A typical discharge To prevent excessive outrush into faulted capacitors, the
transient waveform is shown in Figure 16.7-3. total parallel-stored energy should not exceed the
energy capability or joule rating of either the capacitor
The approximate I2t for the outrush current (Io in Fig- unit or the fuse. According to ANSI C37.99 and to
ure 16.7-3) from a capacitor to a failed unit can be esti- manufacturing recommendations, the calculated energy
mated by the following: of the bank must not exceed 15,000 Joules (4650 KVAR
in parallel) for all-film capacitors or 10,000 Joules (3100
2 1 V 2C KVAR) for paper-film capacitors. In cases when the
I t = . 16.7-10
2 2 R1 calculated value of the parallel-stored energy surpasses
the limitation capability of expulsion fuses, two possible
solutions are suggested: reconfiguration of the capacitor
bank, or the use of current-limiting fuses (CLFs).
Inrush Current
When a capacitor is energized, there is an initial inrush
current (Iin in Figure 16.7-3). This is a short-duration,
high-frequency damped sinusoidal current whose char-
acteristics depend on the capacitor size, the point on the
voltage wave at which energization occurs, and the
impedance of the supply circuit.

For adequate protection, the melting I2t of the fuse must

be higher than that of the capacitor inrush current.
Figure 16.7-3 Typical discharge transient waveform.

16-33
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

With acceptable accuracy, the I2t of the inrush current dependent on the surge impedance of the discharge
can be calculated using the following relationship: path, which is a function of the equivalent capacitance
of the two banks, the total inductance of the discharge
2 2 2
I t = 2.65 I L I sc 1+ K A s 16.7-11 path (the inductance of the conductors between the two
banks and the inductance of the capacitor banks them-
Where: selves), and the magnitude of the voltage at the instant
Isc = fault current at capacitor bank location the second bank is energized.
(kA).
IL = capacitor bank line current (A). The minimum equivalent circuit inductance (inductance
K = X/R at the bank location. of the discharge path) L required between the two
capacitor banks to prevent spurious fuse operation can
As can be seen from this equation, the inrush I2 t is a be calculated using the expression shown in Equation
function of capacitor phase current, available short-cir- 16.7-12.
cuit current at the point of application, and the X/R
ratio of the source impedance at that point. K 2 V 4 C e3
L= Henrys
(I t )
2 2 16.7-12
If N parallel capacitors on a phase are individually
fused, then the inrush current through each fuse would
be the value from the inrush I2t equation divided by N. Where:
K = constant equal to 3.7, which represents a
Figure 16.7-4 shows graphically the inrush I2t as a func- typical inrush damping factor of 0.81.
tion of capacitor phase current for a number of system V = peak value of the line-to-ground voltage
short circuit currents and X/R ratios. when the capacitor bank is energized, (V).
Ce = equivalent capacitance of the discharge
Parallel Switching Transients path, (Farads).
Parallel switching transients, which are also commonly
referred to as back-to-back switching transients, occur In Equation 16.7-12, the equivalent capacitance of the
when de-energized capacitor banks are switched into discharge path Ce may be derived from Equation 16.7-13.
service in the vicinity of a previously energized capacitor
bank. The energized capacitors discharge high-magni- C 1 2.65 kVAR
Ce = = • 16.7-13
tude, high-frequency currents (Ip) into the unit being 2 2 2
V lg
switched on, over a period lasting several milliseconds
after the parallel is established. These discharge currents Where:
are only significant when individual capacitor units are KVAR= single-phase KVAR.
installed in close proximity on the same distribution Vlg = line-to-ground voltage, (V).
feeder. The high-frequency transient outrush current
from the already energized capacitor bank is solely I2t= high-frequency surge withstand capability of the
capacitor bank, defined in Equation 16.7-14.

I t = ( I t ) 60 Hz • F PLD • F HFSW
2 2
16.7-14
Where:
FPLD = preload adjustment factor.
FHFSW = high-frequency surge-withstand I2t factor.

For a specific conductor size and configuration, with a

known inductance per unit length, the corresponding
minimum line length between the two capacitor banks
can be calculated as shown in Equation 16.7-15.

L - Lb
Distance = 16.7-15
Lc
Where:
Lb = self inductance of the two banks in paral-
lel supplied by the capacitor manufac-
Figure 16.7-4 Energizing inrush I2t for grounded-wye
banks. turer, (H).
Lc = conductor inductance per foot (H/ft).

16-34
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

When capacitor banks are separated by several pole- to 25 A. In addition, the location of the arrester between
spans, nuisance operation of capacitor fuses due to the fuse cutout and the capacitor must be avoided.
parallel switching is not a major concern, and the calcu-
lation shown in Equation 16.7-15 can be used to con- Available Fault Current
firm this.
The system fault current available at the capacitor loca-
Figure 16.7-5 provides an example of the capacitor dis- tion, the type of connection (such as delta or wye, neu-
charge I 2 t for various KVAR units that are spaced a tral grounded, or ungrounded), the number of series
varying number of 150-foot spans on a line with 795 groups, kVA rating of the bank, and the number of
ACSR conductor. capacitors in parallel are all factors that should be taken
into consideration by the protection engineer when
De-Energizing Transients determining the proper protection for the capacitor
De-energizing transient currents (Id) can occur when bank.
opening a capacitor switch. When the capacitor switch is
being opened, the capacitor tries to maintain the poten- When capacitors are connected grounded-wye or delta,
tial that it had before the contacts where opened. If the any capacitor failure will cause the system fault current
switch restrikes, the oscillatory current discharge has a to flow through the faulted capacitor. The capacitor
high peak value. De-energizing transients are more must withstand the short-circuit current flow until the
likely to occur in circuits with voltages above 25 kV. In circuit is interrupted by the fuse. When multiple-series
these cases, restrike-free switches must be installed. groups of capacitors are used, as a general rule system
fault current will not flow through a faulted capacitor,
De-energizing transients can be estimated with Equa- and expulsion fuses can be employed.
tion 16.7-16.
With the wye configuration, the neutral can be either
2 2 2
I t = 10.6 I L I sc 1+ K A s 16.7-16 grounded or floating. When grounded, the fault current
through a failed capacitor is the available system line-to-
Where:
ground fault current. For the delta connection, line-to-
Isc = fault current at capacitor bank location
line system fault current will flow through the failed
(kA).
capacitor.
IL = capacitor bank line current (A).
K = X/R at the bank location. In an ungrounded-wye capacitor bank, the fault current
High-Frequency Transients is limited to three times normal line current. Available
Capacitor fuses are commonly exposed to high-fre- fault current to the failed unit and interrupting duty on
quency transients due to lightning surges. These surges the fuse are, therefore, reduced. The fuse, however, must
are more likely to damage low-current rated links. When be small enough to detect this low-level fault current.
group protection is employed, spurious fuse blowing can Furthermore, while the faulted capacitor is in the cir-
be reduced by utilizing a slow-clearing T tin link of up cuit, the neutral shift causes the voltage across the
capacitor in the unfaulted phases to increase to 1.73
times the rated voltage. Operation under these condi-
tions will result in failure of the healthy capacitors in a
short time. The fuse must operate as quickly as possible
to remove this overvoltage.

It must be ensured that the available fault current does

not exceed the interrupting rating of the selected fuse.
The available fault current, along with the following con-
siderations of capacitor rupture hazard, are used to
determine whether CLFs are required for an application.

Capacitor Case Rupture Hazard

Capacitor case rupture will occur if the total energy
applied to the capacitor under short-circuit conditions
Figure 16.7-5 I2t from parallel switching of capacitors is greater than the ability of the capacitor case to with-
on a distribution feeder. stand such energy.

16-35
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

A capacitor unit internally consists of a number of series through I2t of the fuse must always be less than the min-
groups of parallel-connected packs. Capacitor failure imum rupture I2 t of the capacitor. I2t coordination of
usually starts with the breakdown of one pack, which the capacitor minimum rupture I2t curve and the fuse
then shorts out the group. The capacitor current total clearing I2t curve will determine whether expulsion
increases, as does the voltage in the remaining series fuses are suitable to protect against case rupture at high
groups. This increased voltage will eventually lead to the fault levels, or whether CLFs are necessary.
dielectric failure of another pack, causing another
increase in current and voltage across the remaining Capacitor manufacturers supply the minimum rupture
good groups. This process will continue until all the I2t information for their units. Some typical values are
groups have failed, and the capacitor acts as a bolted provided in Table 16.7-2. Comparison of the tabulated
fault. The process may take hours or longer, during values with I2t curves for expulsion fuses, as illustrated
which time current escalates in discrete steps. It is desir- in Figure 16.7-7, can be used to determine the fault cur-
able that the capacitor fuse operate before all the series rent limit for expulsion fuse protection. If a capacitor
groups have failed, because the then remaining good minimum rupturing I2t is about 1,000,000 A2s, as illus-
groups will limit the fault current and the possibility of trated in Figure 16.7-7, then expulsion fuses will provide
case rupture will be minimized. protection with fault currents up to 8000 A. For a
capacitor with a lower minimum rupturing I 2 t of
The cause of capacitor case failure is attributable to the 100,000 A2s, expulsion fuses would only provide protec-
development of excessive internal pressure sufficient to
stress the capacitor case beyond its mechanical limits.
When a capacitor dielectric fails, the resulting arc cre-
ates internal pressure from heat and a gas generated in
the liquid dielectric of the unit. The pressure varies
depending on the magnitude of the fault current to the
failed unit and the time that it is allowed to flow. This
force can swell the sides of the capacitor case or rupture
the case—that is, anything from opening a seam or
bushing seal, to violent bursting, endangering adjacent
equipment.

Case-rupture curves are essential to the correct selection

of fuse links for overcurrent protection of any capacitor
installation. These curves, which are available from
capacitor manufacturers and standards, illustrate the
probability of case rupture for various time and current
relationships.

Capacitor case rupture for newer all-film capacitor

designs is generally defined by a single-case rupture
curve (see Figure 16.7-6). This is possible, because all-
film units fail to short circuit in a more predictable man-
ner, and thus have a more well-defined rupture thresh-
old than older paper-film capacitors.
Figure 16.7-6 Case-rupture curves for shunt capacitors
Capacitor fuses must have a time-current clearing char- (150, 200, and 300 KVAR all-film capacitors).
acteristic that will ensure rapid isolation of a faulted
capacitor without case rupture. For adequate protec-
Table 16.7-2 Typical All-film Capacitor Minimum
tion, the fuse total-clearing time current curve (TCC)
Rupture I2t
must lie to the left of the single case-rupture curve of the
Capacitor
capacitor. For high fault currents, case-rupture curves Unit
All-film Units All-film Units
must be compensated for asymmetry. 100 KVAR 150, 200, 300 KVAR
Voltage
Above 9000
112,500 A2s 450,000 A2s
Capacitor case rupture must be considered, not only by V
using the TCCs, but by ensuring that the maximum let- Below 9000 V 250,000 A2s 1,000,000 A2s

16-36
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

tion to about 3000 A. Note from Figure 16.7-7 that a ing of a single capacitor per phase, a failure on one
CLF can be used for protection at higher fault levels. phase increases the voltage on the other phases to 1.73
times the rated voltage. If the faulted capacitor is not
Overvoltage Protection removed promptly, this overvoltage can cause a second
The fuse link must protect healthy capacitors against capacitor failure.
the possibility of being operated at excessive overvoltage
during failure of an adjacent unit. Capacitors are IEEE Standards 18 and 1036 provide recommended
designed to operate normally at specific 60-Hz nominal limits for the duration of power frequency overvoltages
voltage, which is listed on the unit nameplate. However, on capacitors in service. These limits range from a dura-
a 10% overvoltage is allowed without causing any dam- tion of six cycles for an overvoltage 2.2 times rated volt-
age to the capacitor. age, to a duration of 30 minutes for an overvoltage of
1.25 times rated voltage. The fuse on a faulted unit must
Table 16.7-3 illustrates the overvoltage that is experi- operate fast enough to limit the duration of an overvolt-
enced by good units during the failure of a capacitor on age on the healthy units. One general rule for selecting
another phase. The table illustrates that the overvoltage fuses is to require the fuse to operate within 5 minutes at
is also a function of the number of units in series on the 95% of the fault current. K-type links operate more
faulted phase. In ungrounded-wye installations consist- quickly at high currents than equally rated T-links and
thus offer reduced overvoltage durations.

Fuse manufacturers generally publish selection tables

that reflect consideration of all the factors mentioned
above, and permit the direct selection of the capacitor
bank fuse, thereby eliminating the need to perform com-
plex calculations or graphical studies. However, for the
purpose of explanation, a step-by-step procedure to
select fuses for capacitor protection is provided in the
following section.

16.7.6 Application of Capacitor Fuses

In summary of the preceding sections, a capacitor fuse
must be selected to:
• carry continuous capacitor current,
• isolate a faulted capacitor,
• withstand transient currents,
• have sufficient interrupting capacity to interrupt the
maximum fault current at the point of application,
• limit the energy let-through to a faulted unit to mini-
Figure 16.7-7 Limit for capacitor protection by expulsion mize the possibility of capacitor case rupture,
fuse.
• protect healthy units against prolonged overvoltages.
Table 16.7-3 Overvoltage on Healthy Capacitor Units During Fuse Type
Short-Circuit Failure of a Series Unit The required interrupting duty of the protection device
can be established through an assessment of the avail-
Number of Voltage on Each Phase During a Short-Circuit
Series Failure of One Series Unit on Phase "a" (per-unit
able system fault current and the capacitor bank
Groups nominal phase voltage) configuration. Reviewing the available fault current and
Grounded Wye Ungrounded Wye the appropriate case-rupture curves for the capacitor
Va Vb Vc Va Vb Vc units will dictate whether 1/2-cycle fault clearing is ade-
1 - 1.00 1.00 - 1.73 1.73 quate or if fractional-cycle clearing is necessary.
2 2.00 1.00 1.00 1.50 1.15 1.15
3 1.50 1.00 1.00 1.29 1.08 1.08
4 1.33 1.00 1.00 1.20 1.05 1.05
The fuse-interrupting duty is established based on the
5 1.25 1.00 1.00 1.15 1.04 1.04 phase configuration (grounded wye versus ungrounded
wye versus delta), the capacitor unit arrangement

16-37
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

(series-parallel combinations), and the protection In parallel with selecting the fuse speed is selection of
arrangement (group versus individual). the fuse rating. This step is completed by comparing the
TCCs of short-listed fuses with the fuse withstand char-
The phase configuration determines the maximum level acteristics for outrush into faulted units, parallel switch-
of fault current that will flow for the condition where all ing outrush, and energizing inrush as appropriate, and
series branches of a bank phase are shorted. with the case-rupture curves for the specified capacitor
units. Where current-limiting fuses are specified, the
The capacitor unit arrangement is used to determine the minimum-melting I2t of the various fuse ratings can be
proportion of this fault current that will flow for the compared directly with the relevant withstand values,
failure of a single capacitor unit (impedance divider and the total let-through I2 t of the fuses can be com-
principle), along with the total I2t discharged into a sin- pared directly with the case-rupture I2t. The fuse-melt-
gle unit from the unfaulted parallel units. If the total I2t ing TCC must remain above and to the right of the
available from the system and from parallel units over relevant withstand curve throughout the current range,
the first half-cycle is found to exceed the case-rupture and should be selected to stay as close as possible to this
I 2 t, then current-limiting fuses will be required. This curve. By virtue of this selection preference, the designer
step will, therefore, determine whether a distribution can ensure that the probability of case rupture is as low
cutout, a solid-material power fuse or a current-limiting as possible.
fuse is to be used.
Example Capacitor Protection
Fuse Current Rating and Speed
A three-phase, 600-KVAR, grounded-wye connected
The idealized goal for capacitor fuse application is the capacitor bank is installed in a pole-top configuration.
selection of the fastest fuse that will avoid damage from The capacitor bank is configured with two single-phase,
the range of transient conditions. It is evident that these all-film construction, 100-KVAR capacitor units in par-
are conflicting requirements that must be reconciled allel connected in each phase. The capacitor units have
when choosing the fuse rating and speed ratio. With- voltage ratings of 7.2 kV. Assume the maximum line-to-
standing continuous current and transients would sug- ground fault current is 800 A rms symmetrical. The per-
gest use of a slow speed fuse, whereas reducing case phase load current of the capacitor bank is 27.8 A.
rupture and overvoltage require a fast fuse. Some typi- Select a primary protective device for this application.
cal practices are provided as examples in the following
paragraphs. According to the capacitor manufacturer, the capacitor
bank can be adequately protected with expulsion fuses if
In group protection applications requiring high continu- the maximum available fault current does not exceed 3.1
ous current ratings (above 25 A), K-type links provide kA rms symmetrical, which is the case.
adequate withstand to transient currents, while keeping
the melting time as short as practical and providing To accommodate the highest anticipated capacitor bank
maximum protection against case rupture. Good coor- current, the fuse continuous current is selected based on
dination is obtained using a K fuse link having con- the following bank tolerances: 10% overvoltage, 15% in
tinuous current capability of at least 165% of the capacitance, and 10% in harmonics. Consequently, the
capacitor current rating for grounded-wye banks and minimum continuous current that the fuse must carry is
150% capacitor current rating for ungrounded-wye determined as shown in Equation 16.7-17.
banks. In group protection applications where a fuse
with a low current rating is required (below 25 A), I c = 1.35 x 27.8 = 37.53 A 16.7-17

slower T-links may be preferred. The small T-links have

higher immunity to transients and lightning surges and Because K and T links are 150% rated, the continuous
may be particularly advantageous in areas with expo- current value must be divided by 1.5. As a first approxi-
sure to lightning activity (rural areas with little tree shel- mation, a 25T fuse link mounted on a 200 A distribu-
ter and high iso-keraunic levels). If power fuse or tion cutout is selected to protect each group of
current-limiting protection is required, the slowest avail- capacitors. The cutout has a rated voltage of 7.2 kV, and
able E-rated or C-rated fuses should be used. an interrupting capability of 10 kA rms symmetrical
(based on X/R ratio of 4), which exceeds the maximum
For individual capacitor protection, where low-current- available short-circuit current at the capacitor bank
rated links are generally required, T-links have the location.
advantage of withstanding greater outrush current.
Where higher current ratings are required, K-links offer The ability of the fuse to withstand the energizing
faster clearing for improved case-rupture protection. inrush currents associated with the capacitor bank is

16-38
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

determined by comparing the unloaded high-frequency • Use a partial-range CLF in series with the expulsion
surge-withstand I2t capability of the fuse with the I2t of fuse, or use a full-range CLF.
the transient inrush current. The I2t of the inrush cur-
rent when the capacitor bank is energized is calculated Table 16.7-4 provides some examples of fuse link ratings
as shown in Equation 16.7-18. selected for individual protection of all-film capacitors
with expulsion fuses. Table 16.7-5 provides some exam-
I t = 2.65 • 27.8 • 0.793 • 1+ 4 = 241 A s
2 2 2 16.7-18 ples of CLFs selected to protect individual all-film

Data available from the fuse manufacturer indicates that

the unloaded high-frequency surge-withstand I2t for sil-
ver-copper eutectic element links is approximately 45%
of the 60-Hz minimum melting I2t value. Using the min-
imum-melting TCC of the selected fuse (25T), the cur-
rent at 1 second is 200 A, which gives a minimum-
melting I2t of approximately 40,000 A2s. The fuse high-
frequency, surge-withstand I2t is calculated as shown in
Equation 16.7-19.

I 2 t hf = 0.45 x 40,000 = 18,000 A 2 s 16.7-19

This means that the transient inrush current associated

with energizing an isolated capacitor bank will not
cause nuisance blowing of the expulsion link selected.

The next step is to verify that the selected fuse can effec-
tively protect the individual capacitor units against case
rupture. This step is accomplished by comparing the
total-clearing TCC of the fuse with the case-rupture
curve of the capacitor unit. Figure 16.7-8 shows a plot
of the case-rupture curve of a 100-KVAR all-film capac-
itor unit along with the fuse total-clearing TCC.

The fuse total-clearing TCC lies below and to the left of

the capacitor case-rupture curve for all current values
up to approximately 3900 A. This crossover point indi-
cates the maximum short-circuit current for capacitor Figure 16.7-8 Example of capacitor protection.
bank protection. Because the maximum available fault
current at capacitor location (800 A) is lower than the Table 16.7-4 Typical Individual Expulsion Fuse Ratings
fault-current value at the point of intersection of the for All-Film Capacitors
two curves, the fuse will always clear the circuit prior to
case rupture. The selected fuse will clear the fault cur- Capacitor Fuse
rent in approximately 0.07 seconds. Unit Voltage
Voltage Rating
Rating (kV) Capacitor Unit Rating (KVAR)
If the maximum available fault current was greater than 50 100 150 200 300 400
the maximum fault-current for capacitor protection, the Fuse Link Rating
selected expulsion fuse would not provide adequate pro- 2400 8.7 20T 40K 65K 80K - -
tection to the capacitor bank, and one of the following 4800 8.7 12T 20T 30T 40T - -
alternatives would be considered: 7200 8.7 10T 15T 20T 25T 40T 50T
• Move the capacitor bank to a location where the 7960 8.7 10T 15T 20T 25T 40T 50T
available fault current is lower. 8320 8.7 10T 15T 20T 25T 40T 50T
14400 15.0 a 10T 15T 20T 25T 30T
• Use larger capacitor units.
• Individually fuse the capacitor units. a. For high-voltage 50-KVAR units, fuses with appro-
priate current ratings will not withstand the out-
rush I2t.

16-39
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

Table 16.7-5 Typical Individual CLF Ratings for All-Film capacitors. Table 16.7-6 provides examples of expulsion
Capacitors fuses for group protection of all-film capacitors.
Capacitor
Unit Fuse 16.8 HIGHLIGHTS
Voltage Voltage
Rating Rating Capacitor Unit Rating (KVAR) Efficiency and Components of Transformer Loss
50 100 150 200 300 400
• Loss in transformers is due to two causes: load loss
CLF Current Rating
and no-load loss. Physically, two main components of
2400 8.3 30 65 90a - - -
transformer loss are: electric (I2 R) and magnetic
4800 8.3 18 30 45 65 - -
(core hysteresis and core eddy current loss). Trans-
7200 8.3 18 25 30 40 65 80a
former efficiency is related to the amount of watts
7960 8.3 18 18 30 40 65 80a
losses that occur when the transformer is in opera-
8320 15.5 10 18 25 35 50a 80a
tion. The percentage of power that is available on the
14400 15.5 - 10 18 25 30 50a
secondary side of the transformer, as a percentage of
a. Indicates parallel fuses. the power input on the primary, is termed the effi-
ciency.

Table 16.7-6 Typical Group Protection for All-Film Capacitors

System Capacitor Three- Rated Line
Line-to-Line Line Phase-Bank Current in
Voltage Voltage KVAR Amperes Typical Link Size
Grounded Wye Ungrounded Wye
150 20.8 20 T 20 T
300 41.6 40 K 40 K
4160 2400
450 62.5 65 K 65 K
600 83.3 80 K 80 K
150 10.4 10 T 10 T
300 20.8 20 T 20 T
450 31.2 30 K 30 K
8320 4800 600 41.7 40 K 40 K
900 62.5 65 K 65 K
1200 83.3 80 K 80 K
1350 93.8 80 K 80 K
150 6.9 8T 6T
300 13.9 15 T 12 T
450 20.8 20 T 20 T
600 27.8 25 T 25 T
12480 7200 900 41.7 40 K 40 K
1200 52.5 50 K 50 K
1350 59.0 65 K 65 K
1800 78.7 80 K 80 K
2400 105 100 K 100 K
150 6.3 6T 6T
300 12.6 12 T 12 T
450 18.8 20 K 20 K
600 25.1 25 K 25 K
13800 7960 900 37.7 40 K 40 K
1200 50.2 50 K 50 K
1350 56.5 50 K 50 K
1800 75.4 80 K 65 K
2400 100.5 100 K 100 K
300 6.9 8T 6T
450 10.4 10 T 10 T
600 13.9 15 T 12 T
900 20.8 20 T 20 T
1200 27.8 25 K 25 K
24900 14400
1350 31.2 30 K 30 K
1800 41.7 40 K 40 K
2400 55.6 50 K 50 K
2700 62.5 65 K 65 K
3600 83.3 80 K 80 K

16-40
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

• Generally transformers are at maximum efficiency IEEE C57.92 has developed several permissible over-
when they are 50% loaded. When transformers are load graphs for different types of transformers with
lightly loaded, the no-load losses form a large per- respect to a number of factors. For example, a liquid-
centage of the power utilized, and, therefore, the effi- filled transformer with a 50% continuous equivalent
ciency is low. As the transformer is loaded to higher base load at 30°C ambient temperature could be
levels, the load losses dominate the efficiency. The loaded to 120% of full load nameplate rating for five
maximum efficiency point is the optimal point of hours without excessive loss of insulation life.
lowest load and no-load losses. It is determined by
Total Lifetime Cost
the design of the transformer and, theoretically,
could be designed to occur at any load percentage. It • The transformer cost has three components: capital
typically is designed to occur at 50%, because the investment, no-load loss, and load loss. If the end-
average load tends to be about 50% of the peak load. user provides the energy price with the purchase
request, the designer can develop a transformer
• Regulations by Energy Departments often mandate
design that will minimize the total lifetime cost,
minimum efficiency levels for liquid-filled and dry-
including the cost of losses. The result of this process
type distribution transformers.
is the cheapest transformer in the useful life period—
Reduction of Transformer Losses i.e., with the lowest total owning cost, optimized for a
• Reduction of transformer losses and improvement in given application.
efficiency can be achieved by reduction of either load • Typically a transformer is designed to have a mini-
or no-load losses. For any given set of core and wind- mum loss when operated at about 50% of rating.
ing materials, reduction of load losses often leads to However, a larger transformer operated at a lower
an increase in no-load losses and vice versa. fraction of rating may have a smaller cost of losses
• More recently, nano-crystalline steel has become than a smaller unit operated at 50% of rating. This
available for use in transformer cores. The best of latter case will be particularly true in situations with
these steels are based on an Fe-Zr-B alloy that is significant annual load growth.
formed in an amorphous state and then annealed to • Transformer size selection, at any specific load level,
produce very small grain sizes. This process makes is controlled by the thermal load limit, not by the
the alloy less brittle and, thereby, decreases produc- cost of losses. This conclusion depends on the ratio of
tion costs. The alloy has even higher permeability and no-load loss to load loss for the particular set of
also higher saturation induction than the amorphous transformers. It will be true as long as the difference
materials, but it is not yet available in manufactured in no-load loss from one transformer size to the next
transformer cores. is larger than the load loss of the smaller size trans-
• Transformer windings are made of either copper or former when loaded near its rating.
aluminum in round wires, square wires, or flat sheets. • The overall conclusion is that a utility cannot reduce
The resistivity of aluminum is about 1.6 times larger transformer losses by going to a larger size trans-
than that of copper, but aluminum has a lower cost. former that will have lower load losses. The minimum
Many different alloys of aluminum and copper are loss costs are achieved if the smallest possible trans-
available. In general, the lower-resistance alloys are former is selected based on thermal loading limits.
more expensive and harder to work with in the manu-
Polarization Index Test
facturing processes, leading to higher initial costs.
• If a transformer passes the insulation resistance test,
• In addition to choice of material, load losses are
before applying any overvoltage test, it is recom-
affected by the cross-sectional area of the wire used.
mended to do a Polarization Index (PI) test. The
Larger wires produce lower load losses, but then the
polarization index is a ratio of the Megohm resis-
windings are larger, and this requires a larger core,
tance at the end of a 10-minute test, to that at the end
which increases the no-load losses.
of a 1-minute test at a constant voltage. Another
Long-term and Short-time Emergency Overloads common way for PI calculation is the ratio of resis-
• The permissible loading of transformers for normal tance readings that are taken 15 and 60 seconds after
life expectancy depends on the design of the particu- connecting the voltage. The following table is a guide
lar transformer, its temperature rise at rated load, to interpreting the PI test results.
temperature of the cooling medium, duration of the
overloads, the load factor, and the altitude above sea
level if air is used as the cooling medium. ANSI-

16-41
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

Application of Capacitors at Stations and on Feeders

Polarization Index Insulation Condition • Capacitors are used in distribution stations or on dis-
Less than 1 Dangerous tribution feeders. Station capacitors are rack
1.0 - 1.1 Poor mounted in large banks. Capacitors installed on feed-
1.1 - 1.25 Questionable ers are usually in pole-top banks with necessary
1.25 - 2.0 Fair group fusing. Capacitors can be applied as fixed or
Above 2.0 Good switched units. Switched units have capacitor bank
Power Factor Test controllers that switch several capacitor banks.
• In general, power factor measurement equipment • A three-phase capacitor bank on a distribution
comes with three basic modes of operation: grounded feeder can be connected in delta, grounded-wye, or
specimen test, grounded specimen with guard, and ungrounded-wye. The type of connection used
ungrounded. The three measurement modes allow depends upon:
measurement of the current leaking back to the test — System type—i.e., whether it is a grounded or an
set on each lead, individually and together. In gen- ungrounded system
eral, a power factor of less than 1% is considered — Fusing requirements
good; 1-2% is questionable; and if a power factor
— Capacitor-bank location
exceeds 2%, action should be taken. Practically, the
evaluation is not only based on a single power factor — Telephone interference considerations
data point, but is also based on the history of the • Ungrounded-wye capacitor banks are not recom-
change in power factor. Values obtained at the time mended under the following conditions:
of the original tests are used as benchmarks to deter- — On feeders with light load, where the minimum
mine the amount of insulation deterioration on sub- load per phase beyond the capacitor bank does
sequent tests. not exceed 150% of the per-phase rating of the
Purpose of Capacitors capacitor bank
• Fixed and switched capacitors are inexpensive means — On feeders with single-phase breaker operation at
of providing VAR compensation for distribution sys- the sending end
tems and thus correcting power factor and reducing — On fixed-capacitor banks
system losses. — On feeder sections beyond a sectionalizing-fuse or
• Advantages of installing shunt capacitors in distribu- single-phase recloser
tion systems are as follows: — On feeders with emergency load transfers
—Released system capacity • Usually, grounded-wye capacitor banks are
—Reduction in losses employed only on four-wire, three-phase primary sys-
—Improvement in voltage regulation tems. Otherwise, if a grounded-wye capacitor bank is
used on a three-phase, three-wire ungrounded-wye or
• Depending on the uncorrected power factor of the delta system, it furnishes a ground current source
system, the installation of capacitors can increase the that may disturb ground relays.
substation capability for additional load by as much
as 30%, and can increase individual circuit capability, • The optimum amount of capacitor kilovars to
from the voltage regulation point of view, approxi- employ is generally the amount at which the eco-
mately 30 to 100%. nomic benefits obtained from the addition of the last
kilovar equals the installed cost of the kilovars of
Description of Capacitors capacitors.
• Distribution capacitors are typically housed in rect- • The economic benefits that can be derived from
angular, sealed metal cans, which can be made of capacitor installation can be itemized as:
stainless steel. The cans contain rolled packs of alu- — Released generation capacity
minum foil with layers of insulating paper, and/or
— Released transmission capacity
plastic film, between the foil. Recently manufactured
capacitors have all-film-insulating layers. — Released distribution substation capacity
• Capacitors are rated for line-to-line voltage in the — Reduced energy (copper) losses
event that they are applied on ungrounded or poorly — Reduced voltage drop, and consequently,
grounded systems. improved voltage regulation

16-42
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

— Released capacity of feeder and associated appa- them approximately at two-thirds of the distance
ratus from the substation to the end of the feeder. For the
— Postponement or elimination of capital expendi- uniformly decreasing loads, fixed capacitors are
ture due to system improvements and/or expan- located approximately halfway out on the feeder. The
sions location of switched capacitors is often determined
by voltage regulation requirements, and they are usu-
— Revenue increase due to voltage improvements
ally located on the last one-third of the feeder away
• If only fixed-type capacitors are installed, the utility from the source.
will experience an excessive leading power factor and
• The best location for capacitors can be found by opti-
voltage rise at low-load conditions. Therefore, some
mizing power loss and voltage regulation. A feeder
of the capacitors should be installed as switched-
voltage profile study is required to determine the
capacitor banks so they can switched off during light-
most effective location for capacitors and a voltage
load conditions.
that is within recommended limits. Usually, a 2-V rise
• A rule of thumb is often used to determine the size of on circuits used in urban areas and a 3-V rise on cir-
the switched capacitors. Switched capacitors are cuits used in rural areas are the maximum voltage
added until: changes that are allowed when a switched-capacitor
bank is placed into operation.
k var from switched + fixed capacitors
≥ 0.70 • Some summary rules that can be used in the applica-
k var of peak reactive feeder load
tion of capacitor banks include the following:
• The kilovars needed to raise the voltage at the end of — The location of fixed shunt capacitors should be
the feeder to the maximum allowable voltage level at based on the average reactive load.
minimum load is the size of the fixed capacitors that — There is only one location for each size of capaci-
should be used. On the other hand, if more than one tor bank that produces maximum loss reduction.
capacitor bank is installed, the size of each capacitor
— One large capacitor bank can provide almost as
bank at each location should have the same propor-
much savings as two or more capacitor banks of
tion—that is:
equal size.
k var of load center kVA of load center — When multiple locations are used for fixed-shunt-
=
k var of total feeder kVA of total feeder capacitor banks, the banks should have the same
rating to be economical.
• The resultant voltage rise must not exceed the light- — For a feeder with a uniformly distributed load, a
load voltage drop. The approximate value of the per- fixed-capacitor bank rated at two-thirds of the
cent voltage rise is: total reactive load and located at two-thirds of the
Qc ⋅3ϕ Xl distance out on the feeder from the source gives
% VR = an 89% loss reduction.
10 ⋅VL2− L
— The result of the two-thirds rule is particularly
Where: useful when the reactive load factor is high. It can
% VR = percent voltage rise. be applied only when fixed shunt capacitors are
Qc⋅3φ = three-phase reactive power due to fixed used.
capacitors applied, KVAR. — In general, particularly at low reactive load fac-
X = line reactance, Ω. tors, some combination of fixed and switched
l = length of feeder from sending end to fixed- capacitors gives the greatest energy loss reduc-
capacitance location, mile. tion.
VL-L = line-to-line voltage, kV.
— In actual situations, it may be difficult, if not
• Another rule of thumb sometimes used is that: The physically impossible, to locate a capacitor bank
total amount of fixed and switched capacitors for a at the optimum location; in such cases, the per-
feeder is the amount necessary to raise the receiving- manent location of the capacitor bank ends up
end feeder voltage to maximum at 50% of peak feeder being sub-optimum.
load.
Capacitor Protection Considerations
Capacitor Location
• The main function of capacitor protection is to elec-
• The rule of thumb for locating the fixed capacitors on trically remove failed capacitors from the distribution
feeders with uniformly distributed loads is to locate system.

16-43
Chapter 16: Transformers and Equipment EPRI Underground Distribution Systems Reference Book

• The protection must isolate a faulted bank or — Transient currents, including discharge or outrush
individual shunt capacitors without interrupting ser- into faulted capacitor, inrush current, parallel
vice on the remainder of the circuit. When the capac- switching transients, de-energizing transients, and
itor does fail, the protection should rapidly remove it high-frequency transients
from the system to avoid case rupture. If the protec- — Available fault current
tion has been properly coordinated, it should also — Capacitor case-rupture hazard
operate before any other upstream protective devices.
While fulfilling this fault-clearing role, the capacitor — Overvoltage protection
protection must also remain immune to a number of Application of Capacitor Fuses
“normal” transient conditions such as energizing Selection of the appropriate fuse involves consideration
inrush, discharging/outrush, parallel switching out- of the following:
rush, and lightning surges.
• Fuse Type. Through an assessment of the available
• A capacitor fuse must be selected to: system fault current and the capacitor bank
— carry continuous capacitor current, configuration, the required interrupting duty of the
— isolate a faulted capacitor, protection device can be established. Reviewing the
— withstand transient currents, available fault current and the appropriate case-rup-
ture curves for the capacitor units will dictate
— have sufficient interrupting capacity to interrupt whether 1/2-cycle fault clearing is adequate or if frac-
the maximum fault current at the point of appli- tional-cycle clearing is necessary. This review will also
cation, determine whether an expulsion of a current-limiting
— limit the energy let-through to a faulted unit to fuse is required.
minimize the possibility of capacitor case rupture, • Fuse Current Rating and Speed. The idealized goal for
— protect healthy units against prolonged overvolt- capacitor fuse application is the selection of the fast-
ages. est fuse that will avoid damage from the range of
• To ensure that capacitor protection will fulfill these transient conditions. These conflicting requirements
functions, of the following issues must be considered: must be reconciled when choosing the fuse rating and
— Location constraints speed ratio. Withstanding continuous current and
transients would suggest use of a slow speed fuse,
— Bank configuration whereas reducing case rupture and overvoltage
— Individual versus group protection require a fast fuse.
— Continuous current

16-44
EPRI Underground Distribution Systems Reference Book Chapter 16: Transformers and Equipment

REFERENCES Duval, M. 2004. “Recent Developments in DGA Inter-

pretation.” CIGRE TF 15/12-01-11.
ABB. 2002. ABB Distribution Transformer Guide. ABB
Distribution Transformer Division. IEEE. 1991. C57.104-1991. “IEEE Guide for the Inter-
pretation of Gases Generated in Oil-Immersed Trans-
ABB. 2007. ABB Transformer Handbook. ABB Man- formers.”
agement Services Ltd.
IEEE. 1995a. Standard 62-1995. “Guide for Diagnostic
ANSI/IEEE 1981. ANSI/IEEE Std C57.92-1981. Field Testing of Electric Power Apparatus, Part 1: Oil
“IEEE Guide for Loading Mineral-Oil-Immersed Filled Power Transformers, Regulators, and Reactors.”
Power Transformers Up to and Including 100 MVA
with 55 Degree C and 65 Degree C Average Winding IEEE. 1995b. C57.91-1995, “IEEE Guide for Loading
Rise.” Mineral-Oil- Immersed Transformers.”

ANSI/IEEE. 1990. ANSI/IEEE C37.99. “Guide for IEEE. 1999. IEEE C57-100. “Standard Test Procedure
Protection of Shunt Capacitor Banks.” for Thermal Evaluation of Liquid-Immersed Distribu-
tion and Power Transformers.”
ANSI/IEEE. 2000. ANSI/IEEE C57.12.00-2000. “Gen-
eral Requirement for Liquid-Immersed Distribution, IEEE. 2006a. C57.106-2006. “IEEE Guide for Accep-
Power, and Regulating Transformers.” tance and Maintenance of Insulation Oil in Equip-
ment.”
ASTM. 2002. ASTM D877. “Standard Test Method for
Dielectric Breakdown Voltage of Insulating Liquids IEEE. 2006b. C57.12.90–2006. “IEEE Standard Test
Using Disk Electrodes.” Code for Liquid-Immersed Distribution, Power and
Regulating Transformers.”
ASTM. 2004. ASTM D1816-04. “Standard Test
Method for Dielectric Breakdown Voltage of Insulating NEMA. 1964. TR-98-1964. “Guide for Loading Oil-
Oils of Petroleum Origin Using VDE Electrodes.” Immersed Power Transformers with 65°C Average
Winding Rise.”
ASTM. 2008a. ASTM D924-08. “Standard Test
Method for Dissipation Factor (or Power Factor) and NEMA. 1993. NEMA TRI-1993. “Transformers, Regu-
Relative Permittivity (Dielectric Constant) of Electrical lators, and Reactors.
Insulating Liquids.”
NEMA. 2002. NEMA Standards Publication TP 1-
ASTM. 2008b. ASTM D974-08. “Standard Test 2002. Guide for Determining Energy Efficiency of Distri-
Method for Acid and Base Number by Color Indicator bution Transformers.
Titration.”
Pabla, A. S. 2004. Electric Power Distribution. McGraw-
ASTM. 2009. ASTM D664-09. “Standard Test Method Hill.
for Acid Number of Petroleum Products by Potentio-
metric Titration. Short, T. 2004. Electric Power distribution Handbook.
CRC Press.
DOE (Department of Energy). 2007. 10 CFR Part
431, Part III, 2007. “Energy Conservation Pro- Winders, J. 2002. Power Transformers, Principles and
gram for Commercial Equipment: Distribution Applications. Marcel Dekker.
Transformers Energy Conservation Standards;
Final Rule.”

16-45