Three Phase Transformer Winding Configurations and

Differential Relay Compensation

Larry Lawhead, Randy Hamilton, John Horak

Basler Electric Company

Presented before the

60th Annual Georgia Tech
Protective Relay Conference
May 2-5, 2006
This page intentionally left blank.
 Three Phase Transformer Winding Configurations and
Differential Relay Compensation

Larry Lawhead, Randy Hamilton, John Horak

Basler Electric Company

Most engineers have some familiarity with two commonly known delta connections that give
either a +300 or - 300 phase shift of positive sequence voltages and currents, and just these two
configurations seem to cause extensive confusion. There are actually many other ways to
configure a wye or delta that give other phase shifts, and to further complicate matters, there is
the occasional zigzag winding application and the additional confusion over what occurs when
CTs are connected in delta. These alternate transformer winding configurations are sometimes
referred to by terms such as Dy# or Yy#, or Yd#, Dz#, and Yz#, and where the # can be,
seemingly, almost any hour of the clock, hence the term “around the clock” phase shifting is
sometimes heard. The paper will review the variety of possible winding configuration and give
examples of the nomenclature that is used with them and how these various phase shifts are
created.

The paper will also show how a transformer differential relay compensates for the effects of the
various transformer winding configurations, as well as account for delta CT configurations. Many
papers and instruction manuals refer to compensation in terms of phase shifting. This leads
engineers to have a vague and misleading understanding that the relay is somehow phase
shifting currents to compensate for the transformer phase shift. While a “sequence component
differential” relay might be able to work this way, most transformer differential relays work
outside of the sequence component domain and do some form of current balance calculation in
the ABC domain.

Basic Transformer Design Concepts

Before proceeding, let us review a few points on transformer design and review some
nomenclature that will be used in the paper.

Since this paper is aimed at describing transformers that use phase shifts and winding designs
that are more commonly found outside the US market, the phase and bushing names of U, V,
W, will be used generally, rather than A, B, and C, or H and X. The means of specifying phase
shift and transformer connection will be the D-Y-Z + clock method. For instance, a transformer
connection will be Dy1 rather than a DAB/Y, though some dual designations will be used for
clarity.

If one browses transformer sales literature, technical papers and books, and industry standards
such as IEEE C57.12.00 and IEC-60076-1, one will find many variations on the nomenclature
and figures used to show how the phases are identified in a three phase system. Figure 1 is a
composite of some of the identification methods that will be found. Besides the terms seen in
Figure 1, the terms R, S, and T are used in some sources to name the phases. Also, various
color schemes are used. Some use of the colors black, red, and blue, or brown, orange, and

Prepared by Basler Electric for presentation to the

Georgia Tech Protective Relay Conference, May 2006
2/14/06. Comments: [email protected]

1
yellow to reference the phases. White is sometimes used rather than blue, but white is also
commonly used for neutral in low voltage applications too. Green or bare copper is generally
used for a ground conductor.
Phase or Bushing Names Transformer Phasors
Examples from Industry IEC 60076-1
U V W u v w I II III H2
(i) (i) (i)
or X

H1 H3
Reference
Angle
X Y Z x y z (ii) (ii) (ii)

1U
Primary Secondary Primary
Reference
A B C a b c i ii iii Angle

(i) (i) (i)

1W 1V

Transformer Symbols
(ii) (ii) (ii)
Primary Secondary
1U1 1V1 1W1 2U1 2V1 2W1 Secondary

C57.12.10
Bushing Designations
H0 H1 H2 H3

1U2 1V2 1W2 2U2 2V2 2W2

X0 X1 X2 X3

Primary Secondary

Figure 1 - Example Transformer Symbology and Terminations

Many times the figures that are used to model a transformer and show its connections do not
clearly indicate which windings share a common core leg, and do not clearly indicate the polarity
of the coupled windings, at least to the first time reviewer. To make the winding polarity more
evident in this paper, a variation of the drawings above will be presented that makes it clearer
where the polarity of the windings is and which windings share a common transformer leg. See
Figure 2.

We need to define the term “Winding” to avoid confusion. In various sources the term winding
can refer to either a single continuous coil of wire on a single core leg, or collectively to a 3
phase set of windings connected to the 1U, 1V, and 1W bushings, or even the group of
windings associated with a particular phase. Herein, a “winding” is a single continuous coil of
wire on a single core leg, a “winding set” is the set of three windings that constitute the three
phases, each on different core legs, with a common terminal voltage level (e.g., 1U, 1V, and
1W), and a “phase set” will refer to the two or more windings that are found on a common core
leg. See Figure 2.

2
The windings in the transformer figures are given the terms W1-9. The winding sets associated
with the various voltage levels, each on a different core leg, are:
Winding Set 1: W1, W2, and W3
Winding Set 2: W4, W5, and W6
Winding Set 3: W7, W8, and W9 (only seen herein in zigzag transformers).

The windings sharing a common core leg are:

Core Leg 1: W1 and W4
Core Leg 2: W2 and W5
Core Leg 3: W3 and W6
(Windings W7, W8 and W9, used herein for zigzag windings, are hard to classify into a
particular leg since zigzag windings cross-connect core phase legs.)
(In the figures, core leg 1 is abbreviated as ΦA, core leg 2 as ΦB, and core leg as ΦC.)

The windings by themselves are passive devices and which winding is given the number W1,
W2, or W3 (or W4, W5, W6, W7, W8, or W9) will not matter initially, but it will matter when we
connect the second set of windings to the various bushings since this will affect phase shift of
positive (and negative) sequence voltages across the transformer.

Figure 2 - Basic Transformer Representation

The paper will show the balanced transformer output current in terms of the flux in the core that
is due to load current, which invites discussion on how load flux distributes in a core. There are
two issues with core design that affect differential relaying.

First, we normally think in terms of power flow in a transformer as staying within a phase set
(i.e., current in a single phase on the secondary of a three phase transformer will generate
current in the primary only of that same transformer leg), but this is not always the case.

3
Normally when current flows in the secondary winding, it generates a flux in that leg and an
equal but opposing flux is generated by load current flowing in the primary on that same phase
leg so that the level of excitation flux remains constant. However, in a three phase transformer
where all phases are magnetically linked, there is some coupling and power transfer between
the phases, especially in the 3 legged core design. The inter-phase coupling provides a
mechanism where an external fault on one phase can cause current on unfaulted phases and is
part of the limitation on sensitivity of a transformer differential relay.

The highest inter-phase coupling occurs in a three legged core transformer. Note that in a 3
legged core all flux that leaves a leg must pass through another leg or air. Examine Figure 2.
Assume some current flows in the W4 winding. This current will tend to create flux in a direction
that demagnetizes the W4 core, resulting in lower back e.m.f. on the W1 primary. Since in this
core the flux in the ΦA leg divides between the ΦB and ΦC legs, the W2 and W3 windings could
supply the current/flux to counter the W4 load. However, W4 is generally wired directly under or
interleaved within W1, so the W1/W4 coupling is stronger than W2/W4 or W3/W4 coupling. Also, it
can be shown that the source that will supply the load current with the least reactive current flow
is W1. Hence virtually all the current required to counter the W4 load flux will come from W1. If
current did come from W2 or W3 the transformer differential relay would tend toward tripping.

In the 4 legged and core form designs, there are flux paths that allow flux to flow without
entering other phases, so each phase is more independent of what is occurring in the other
phases and is indicated in Figure 2 as the zero sequence flux path. In the 5 legged / 4 core
designs, where each phase winding encircles two cores, used mostly on low cost distribution
transformers, the two outside phases are fairly independent of each other, but the inside phase
can be affected by events in either of the two outside phases.

Directly related, a significant difference between core designs that can affect differential
protection is whether the core can support zero sequence voltage. If there is no independent
zero sequence flux path, as in the three legged core design, and zero sequence voltage is
applied to the windings, the excitation flux is forced to travel through air to complete the flux
circuit. Excitation current will be very high and can make a transformer differential operate. This
high excitation current is sometimes masked in differential relays by schemes that filter zero
sequence current from the differential relay. This can be done by connecting CTs in delta, or in
numeric relays, by algorithms that measure zero sequence current entering the windings and
subtract it from the differential analysis. The ability to remove the zero sequence current
mathematically in modern numeric differential relays has reduced the motivation to connect
differential relay CTs in delta in most applications.

While the connection diagrams to follow will generally show a three legged core, this is only for
convenience. The connection diagrams are applicable to either 3x1 phase designs or any three
phase core. In all designs, the windings for any given phase are still on a common leg of the
core.

Positive Sequence Phasing and Even More Possible Winding Configurations

The entire paper assumes positive sequence phase rotation and that all the configuration
diagrams were built on positive sequence phase rotation. There are actually more connections
than listed if one wishes to do phase rotation changes within the transformer by swapping two
phases on only one side of the transformer and leaving the third in place in a way that would
convert positive sequence voltage on one side of the transformer to negative sequence voltage
on the other side. This seems a fairly remote possibility.

4
Transformer Winding Configurations
There are 4 basic winding configurations for the typical large 3 phase power transformers: delta,
wye, auto, and zigzag. Within the delta, wye, and zigzag groups, there are multiple means of
creating the winding configuration. There is some help available in that only a few versions of
each show up in regular practice, but this does not rule out any version showing up in one’s
work (or from a relay manufacturer’s point of view, showing up in the relay application) so the
possibility should not be excluded from one’s plans. Scott T and mid-winding tapped delta
transformers are not considered herein.
External vs. Internal Connections
The paper mainly is attempting to analyze the internal connections to the transformer, i.e., what
occurs behind the bushings and is unchangeable by the end user. However, the diagrams also
can be seen as representations of some forms of external phase connection shifts. For
example, in typical practice in the U.S., phases A, B and C connect to bushings H1, H2 and H3
and X1, X2 and X3, respectively. However, in some applications it may be convenient to shift
connections and connect A, B and C to H2, H3 and H1 and X2, X3 and X1, respectively.

Wye Winding Configurations

Given three windings, named W1, W2, and W3, and three transformer phase bushings named U,
V, W, and a neutral bushing N, and assuming a positive sequence rotation is to be maintained,
there are 6 different ways the windings and bushings can be interconnected, as shown in Figure
3. Of the 6 variations of wye connections listed, in actual practice almost all connections can be
seen in terms of Y0 or Y6. As previously mentioned, which winding is given the number W1, W2,
or W3 will not matter initially, but we need to keep track of which winding is connected to which
bushing so that when we connect the second set of windings we will know the positive
sequence phase shift across the transformer.

Phase Shifts Using Clock Notation

Note the use of Y# for naming each of the configurations. The # refers to the phase angle, as
viewed on a 12 hour clock, of winding W1 relative to the voltage applied to the U bushing with a
balanced 3 phase positive sequence voltage (UVW or ABC sequence). Note also the phasor
diagrams in Figure 3 that show the phase angle between the W1 winding and the voltage on the
U bushing. The relationship between the W1 voltage and the U bushing will become important
when one needs a common reference for determining phase shift across the transformer. One
finds the phase shift by lining up the W1 voltages on different sides of a transformer and seeing
how the U bushing voltages on each side compare.

5
 Figure 3 - Six Ways to Wire a Wye Winding

Delta Windings Configurations

Given a set of 3 windings W1, W2, and W3, and three transformer phase bushings U, V, W, and
again assuming that positive sequence rotation is to be maintained, there are 6 ways the
windings and bushings can be interconnected, as shown in Figure 4. Again, some of the
methods would not be seen in normal practice. The most common ones would be D1 (= DAB in

6
common US market terminology) and D11 (= DAC in common US market terminology). The D5
and D7 configurations are seen in some documentation of transformer connections in the
international market.

Figure 4 - Six Ways to Wire a Delta Winding

7
Auto-Transformer Winding Configurations
There is no flexibility in how an auto-transformer is wired. The only way to obtain a phase shift is
to rename the phases from one side to another, but this is quite uncommon; i.e., given a phase
set of W 1 and W 4, then if 1U is connected to W1 then 2U would almost invariable be connected
to the W4.

Figure 5 - An Auto Transformer Winding.

The auto-transformer will not be dealt with any further in this paper. The auto-transformer can
be treated as a simple wye-wye transformer for current differential calculations. Phase shift in
any delta tertiary connection would be treated as a separate delta/wye connection.

Zigzag (Interconnected Star) Winding Configurations

In much literature on transformers, the term “interconnected star” is used in lieu of “zigzag.”
However, most engineers in the power engineering business know with at least a little familiarity
what is meant by zigzag and this term will be used herein. Zigzag transformers do not see much
application in the US. Most engineers are familiar with a zigzag transformer from its application
as a ground reference on an ungrounded system. The wye/zigzag transformer is apparently
used in some distribution applications due to the lower cost of a high voltage primary wye
winding versus a similar voltage delta winding. When the primary is ungrounded wye, the
secondary needs to be zigzag in order to remain an effectively grounded secondary. The zigzag
winding also is used for obtaining non-standard phase shifts. It is capable of delivering
secondary power with 00, +/-300 or +/-600 phase shifts relative to the primary, depending on
primary winding configuration.

A zigzag winding is a series connection of two windings whose voltages are 600 out of phase.
The two windings are typically the same voltage magnitude, but custom phase shifts can be
created if the voltage magnitude of the two windings differs. There are two basic ways to create
a zigzag winding:
Connect the U leg in series with V leg (called a ZUV or a ZAB) or
Connect the U leg in series with W leg (called a ZUW or a ZAC)
The polarity marks of the two windings either face toward one another or face away from one
another. In order to get other phase shifts, some configurations may invert polarity connections,

8
change which side of the winding the bushings are connected to, or introduce a 1200 phase shift
by swapping the U, V, and W bushings connections.

The inter-phase coupling that a zigzag winding creates results in primary current patterns that
may be quite unlike what one expects with delta and wye primary windings. Unless a delta
tertiary is provided, the primary of a three legged core zigzag transformer cannot supply current
that will create zero sequence flux in the core. Even if the primary is connected in delta, the
winding cannot support ‘circulating I0’ since this creates zero sequence flux. For example:
Assume a zigzag secondary and a 1 per unit line to ground fault on the zigzag side. If the
primary were wye (see Figure 18), then the current is 1pu on one phase lead, and -1pu on
another, similar to a delta/wye bank response. However, if the primary were delta (see Figure
19), the current in the lines feeding the primary will see 2 pu current on one phase and -1 pu on
the other two phases, similar to what would be seen for a “double delta” transformation. If power
is to flow from the zigzag side to the wye or delta side, a zero sequence flux path or a buried
tertiary may be necessary for the transformer to carry any unbalanced loading, even for phase
to phase unbalance on a delta primary.

A zigzag winding used as a grounding bank is shown in Figure 6. When a phase U line to
ground load is applied, the current in the load will want to return via windings W4 and W3.
Current in W4 and W3 will demagnetize the two respective core legs, which will in turn allow
phase to ground voltage on the W and V to cause current flow in W1 and W6 to restore a flux
balance condition. The net effect is that equal current flows in every phase of the zigzag bank,
which constitutes zero sequence current into the system.

Figure 6 - Zigzag Grounding Bank

Given the 6 windings and three bushings associated with a zigzag winding configuration, there
are actually 24 different winding configurations connections that could be made. Again, there
are two general types, ZUV (=ZAB) and ZUW (=ZAC), defined by how the two sets of windings are
interconnected and can be seen by inspection of Figure 7. Figure 7 shows what appears to be
the more common variations of ZUV and ZUW but note that a 1800 inversion of each is possible by

9
simply inverting the neutral and bushing connection ends, giving 4 more possible variations of
ZUV , ZUW or inverted ZUV- or ZUW-. See figures for examples. For each R1 and R2 swap of the U,
V, and W bushing connections one should be able to see 8 more similar variations of the zigzag
winding.

To fully define a zigzag connection one needs to specify both the phase shift and whether it is a
based on ZUV or ZUW . For example, an inspection of Figure 7 will show it is possible to get a Z1
(300) phase shift by either the Z1UV connection shown in Figure 7 or by starting with a Z5UW
connection and then introducing a 1200 phase shift by appropriate U/V/W swapping.

Figure 7 - Four Ways to Wire a Zigzag Winding

10
Combining the Various Winding Configurations
An exhaustive listing of all the possible permutations of transformer winding configurations is not
worth presenting. Only a small percentage of all the possible winding configurations is
commonly found in practice, and many are effectively redundant views of the same
configuration. More complete wiring and phasor diagrams of some of the common
configurations are given below in figures 8-19 and the rest can be analyzed as needed, using
the material as a starting point. Except for the two cases of zigzag windings, only transformers
with two basic voltage levels are covered. A somewhat extensive list of the more common
configurations is found in [1] IEC60076-1, Power Transformers. Figures 20 and 21 are drawn
from this standard without the details of Figures 8-19 but annotated with some additional
information.

In the figures to follow, the phasor diagrams found more commonly in US practice are shown on
the left, and phasor diagrams that might be more like those found in international practice are
shown on the right. Polarity marks have been added to the US practice phasor drawings. The
winding diagram in the lower left is intended to show most clearly how the windings would be
interconnected, using US practice bushing names. The figure on the lower right shows a
common international method of showing the winding connections, except the method has been
modified by the addition of polarity marks and winding numbers, and in these figures, when
zigzag windings are shown, one winding set is shown with the polarity mark up, and the other
with the polarity mark down, so that it is easier to see the interconnection of the windings,
compared to IEC drawing practices.

To determine the phase relationship of positive sequence voltage and current between primary
and secondary, use the W1 phasor as the common reference for each winding and compare the
resultant angle between the U phasor on each winding. The angular relationship will be the
primary angle minus the secondary angle, on a 12 hour clock. The naming convention given to
each of the windings also can be used to determine the phase shift during normal conditions.
Consider these examples:

Dy11: D11(UW) pri., Y0 sec.

Phase shift: +11-0 = +11, or -1. The sec. lags the pri. by 3300 or leads the pri. by 300
DzUV2: D1(UV) pri., Z11(UW) sec.
Phase shift: +1-11 = -10, or +2. The sec. lags the pri. by 600 or leads the pri. by 600
Dy7: D1(UV) pri., Y6 sec.
Phase shift: +1-6 = -5, or +7. The sec. lags the pri. by 2100 or leads the pri. by 1500

11
Figure 8 - Yy0 (Y0 - Y0)

Figure 9 - Yy6 (Y0 - Y6)

12
 Figure 10 - Dy1 (D1(=AB) - Y0)

Figure 11 - Dy11 (D11(=AC) - Y0)

13
Figure 12 - Dy5 (D11(=AC) - Y6)

Figure 13 - Dy7 (D1(=AB) - Y6)

14
 Figure 14 - Yd7 (Y0 - D5)

Figure 15 - Yd11 (Y0 - D1(=AB))

15
 Figure 16 - Dd0 (D11 - D11)

Figure 17 - Dd2 (D1(=AB) - D11)

16
 Figure 18 - Yz11 (Y0 - Z1(UV/AB))

Figure 19 - Dz2 (D1(=AB) - Z11(UW/AC))

17
 Phase Shift
(Clock Notation) Wye Primary Delta Primary Zig Zag
Yy0 Dd0 Dz0
I II III

Y0
0
D11 D 11
i ii iii

Y0 D11 Z11uw

Yd1 Dy1 Yz1

Y0
Y0
D1

Y0
Z 11uw
D11

Yd5 Dy5 Yz5

Y0 Y0
5 D11

Y6

D7 Z7uv

Yd6 Dd6 Dz6

Y0

6 D11 D11

Y6
D5 Z5uw

Yd11 Dy11 Yz11

Y0 Y0

11 D11

Y0 Z1uv
D1

Note: Polarity mark is to top of winding in these figures.

Figure 20 – Common connections (IEC 60076-1).

18
 Phase Shift
(Clock Notation) Wye Primary Delta Primary Zig Zag
Dd2 Dz2

2 D1 D1

D11 Z11UW

Dd4 Dz4

4
D 11
D 11

D7 Z7uv

Yd7 Dy7 Yz7

Y0 Y0

7
D1

D5 Y6 Z5uw

Dd8 Dz8

D1 D1
8

D5 Z5uw

Dd10 Dz10

D 11 D 11
10

D1 Z1uv

Note: Polarity mark is to top of winding in these figures.

Figure 21 – Additional connections (IEC 60076-1).

19
Differential Relay Current Balance Equations
Modern numeric transformer differential relays need to be able to work with the many
permutations of how a transformer could be configured, as well as work with connecting the CTs
in delta. The practice of connecting CTs in delta is usually unnecessary when a multifunction
microprocessor based differential relay is utilized. However, in retrofit applications there is some
tendency to leave existing wiring in place, which results in continued use of delta CT
connections (e.g., when the 4th wire required to connect CTs in wye was not provided in the
original installation).

To see how relays work with any form of transformer or CTs, the paper will work toward viewing
a transformer differential relay as a device that monitors flux balance on each transformer leg,
looking only at the flux due to load level, or higher, currents. Except for inrush conditions,
excitation current is too small (<3% of self cooled current ratings) to be seen by the differential
relay, so the relay sees only the load or fault level currents and effectively monitors that the flux
this load current produces balances to zero. In normal operation, on a per phase basis, the flux
created by the load current on a given winding will be countered by an equal and opposite flux
created by the source winding on that same transformer leg, for a net of no change in flux levels
in the leg relative to normal excitation levels. The matter of flux not summing to zero on a leg
due to power transfer between phases due to magnetic coupling between phases (and in the
case of zigzag windings, due to winding interconnections) was discussed at the start of the
paper, under “Basic Transformer Design Concepts,” and also under the description of zigzag
windings.

The flux that would be associated with load level currents on any one winding, and calculated
via the equations to follow, would be on the order of a hundred times the current required to
saturate the transformer. Fault currents would be even a higher multiple. However, since for any
current/flux on one phase an opposite current/flux arises on another phase so that transformer
net flux remains at normal excitation levels, saturation does not occur.

Let us apply the flux balance equation to a transformer with 3 phases, 6 windings, and two
winding sets, primary and secondary. Each winding will produce a flux that is proportionate to its
load current. The flux levels will be:

ϕ1 = NPr ik CI1 ϕ4 = NSec k CI4

ϕ2 = NPr ik CI2 ϕ5 = NSec k CI5 (1)
ϕ3 = NPr ik CI3 ϕ6 = NSec k CI6
where N1 and N2 are the number of turns in winding set 1 and 2, and kC is a proportionality
constant that would be specific to the core, most notably the cross sectional dimensions and
permeability. This kC factor will cancel out of the current differential equations so we do not need
to know its specific value. We can invert the concepts of (1) to state the current in terms of the
core flux in each leg of the transformer, and we can start to introduce matrix representations of
the equations. For winding set W123:

 IW1   k ϕ1 0 0   ϕ1 
I  =  0 k 
0   ϕ2  (2)
 W2   ϕ1

 IW 3   0 0 k ϕ1   ϕ3 

where:

20
 1
k ϕ1 = (3)
NPr ik C

The equations for W456 are similar. In a condensed matrix math form, (2) becomes more
compact:

I123 = k φ,Pri × φ123 (4)

This format will be used extensively in the coming material. It might be noted that when an
equation is bold it indicates that the elements in the equation represent matrices, not simply
constants. Also, a bold multiplication symbol, x, indicates the multiplication of matrices using
appropriate matrix math.

Recall the above equations are referring to load flux, not excitation flux. For load induced flux,
we can assume that in each leg of the transformer the flux that is caused by load current sums
to 0 in normal operation. The equations are:

Phase Leg 1: φW1 + φ W 4 = 0

Phase Leg 2: φ W 2 + φW 5 = 0 (5)
Phase Leg 3: φW 3 + φW 6 = 0

The equation assumes the summation (not the difference) of the flux is zero. There is a parallel
with how currents will be handled. In the equations to come, we will assume a positive current
generates positive flux, so in the current differential function, the error current is the summation
of currents coming into the relay.

Using the abbreviated matrix form of (3) this equality can be stated as:

φ123 + φ 456 = 0 (6)

or alternatively, we can say that on each leg that the ratio of flux from the two core flux sources
(e.g., W1 and W4) is -1.

The current that is delivered by the transformer is the flux times the appropriate kφ factor (2).
This current on the two sides of the transformer will differ in the inverse of the transformer turns
ratio (3) so we need to multiply across by the turns ratio in the manner seen below. Using the
abbreviated form of (4) and (6) the equation that represents normal operation is:

 NPri × k φ1 × φ123  +  NSec × k φ2 × φ 456  = 0 (7)

Putting this equation into a more recognizable form, if we expand (7) and cancel the common kC
factor from (3) we obtain the current balance equation:

NPr iIW1 + NSecIW 4 = 0

NPr iIW 2 + NSecIW 5 = 0 (8)
NPr iIW 3 + NSecIW 6 = 0

21
In a wye/wye transformer with no phase shift of positive sequence voltages and currents, Yy0,
the bushing currents are equal to the winding currents (I1U = IW1, I1V = IW2, I1W = IW3, I2U = IW4, I2V
= IW5, and I2W = IW6), equation (8) tells us how to sense flux balance. The balanced condition can
be directly measured by a differential relay that only monitors by a simple ratio difference. (A
real world transformer differential relay would also need to have a “block for inrush detection”
algorithm as well.)

For any other winding or CT configuration than Yy0 we need a systematic method of creating
“phase compensation” equations so that various combined or inverted winding currents on the
relay inputs will be recognized as a balanced condition. To proceed, we need to determine the
current seen at the transformer bushings and the relay for any given winding and CT
configuration.

Negative Sequence Transformation

Before proceeding, some acknowledgement should be given to the fact that there are even
more possible wye and delta connections than listed, if one opens up the thought process of
swapping two phases at a time rather than three, on only one side of the transformer, in which
case one transforms positive sequence phase rotation to negative sequence phase rotation
across the transformer. Since this transformation would not be done in normal practice, it was
not considered in development of this paper.

Wye Winding Transformations

There are 6 ways to connect a wye winding, as described earlier. Each modifies the W1, W2,
and W3 (or W4,5,6,etc.) current by some transformation factor. For instance, if we connect the U, V,
and W bushings to the non-polarity side of windings W1, W2, and W3, respectively, as is done in
the Y6 connection, we have effectively multiplied all currents by -1. Each connection method has
its own transformation effect.

The equations below state the effective transformation in a mathematical format. This format will
be useful when there are several transformations in series and we need to multiply them
together to find the total effect. The equations give the current that will be seen on bushings U,
V, and W for a given current in W1, W2, and W3, but where current is stated in terms of IW =kφφ.
and kφ is as given in (3). All equations relate IUVW to φ123 because the flux generated by the load
current on each leg is the common element for all winding sets.

There are six 3x3 matrices that represent how the winding currents could be transformed by the
various wye connection. In the condensed matrix equations on the right side, it can be seen that
only four matrix transformations, 1, -1, R1, and R2, are needed to represent the 6 wye
transformation matrices. The value of this reduction in possible transformations will become
clearer when delta and zigzag windings are analyzed further below. The R1 and R2
transformations “rotate” the bushing and winding connections. Given currents I1, I2, and I3, the
effect of each rotation is to say:

R1 (1200 CCW shift): I1new = I3old Inew

2 = I1old Inew
3 = Iold
2
(9)
R 2 (1200 CW shift) : I1new = Iold
2 Inew
2 = Iold
3 Inew
3 = I1old

The matrix form of (9), R1 and R2, will be seen in the equations to follow. In the equations to
follow, note a positive flux generates a positive current.

22
Wye Transformations

 IU   1 0 0   kϕ 0 0  ϕ1 
     
Y0 (Direct connect):  IV  = 0 1 0  0 kϕ 0  ϕ2  IUVW = 1× k φ × φ123 (10)
 IW   0 0 1   0 0 k ϕ   ϕ3 

 IU   0 0 1   kϕ 0 0  ϕ1 
     
Y4 (R2 phase swap):  IV  = 1 0 0  0 kϕ 0  ϕ2  IUVW = R 2 × k φ × φ123 (11)
 IW   0 1 0   0 0 k ϕ   ϕ3 

 IU   0 1 0   kϕ 0 0  ϕ1 
     
Y8 (R1 phase swap):  IV  = 0 0 1  0 kϕ 0  ϕ2  IUVW = R 1 × k φ × φ123 (12)
 IW   1 0 0   0 0 k ϕ   ϕ3 

 IU   −1 0 0   k ϕ 0 0   ϕ1 
I  =  0 −1 0   0  
Y6 (Negated Y0):  V    kϕ 0   ϕ2  IUVW = -1× k φ × φ123 (13)
 IW   0 0 −1  0 0 k ϕ   ϕ3 

 IU   0 −1 0   k ϕ 0 0   ϕ1 
I  =  0 0 −1  0  
Y2 (Negated Y8):  V    kϕ 0   ϕ2  IUVW = -1× R 1 × k φ × φ123 (14)
 IW   −1 0 0   0 0 k ϕ   ϕ3 

 IU   0 0 −1  k ϕ 0 0   ϕ1 
I  =  −1 0 0   0  
Y10 (Negated Y4):  V    kϕ 0   ϕ2  IUVW = -1× R 2 × k φ × φ123 (15)
 IW   0 −1 0   0 0 k ϕ   ϕ3 

where kφ is as defined by (3) for the particular winding.

Note that every matrix can be written of as the negative of one of the other matrices in the list,
as noted in the left column. If one says a negated transformation matrix is not unique, there are
really only 3 unique matrices above.

Delta Winding Transformations

If we connect a winding in delta, the current seen by the relay is a summation of two phase
currents. There are two basic forms of a delta, D1 connection (i.e., a DAB connection), and a D11
connection (i.e., a DAC connection). Given currents I1, I2, and I3:

D1( AB) : I1new = I1old − I2old Inew

2 = I2old − I3old Inew
3 = I3old − I1old
(16)
D11( AC) : I1new = I1old − Iold
3 Inew
2 = I2old − I1old Inew
3 = I3old − I2old

Phase swaps (R1 and R2, above) combine with the above connections to create 6 different ways
to connect a delta. A mathematical way to say the effect of the 6 possible delta connections on
net phase current is given below. As with the wye connections, there are six 3x3 matrices that

23
represent how the winding currents could be transformed by the various delta connections. To
perform all transformations listed below, two additional matrix transformation operations, D1 and
D2, are needed over what was introduced for the wye transformations. Actually, only one of
these is required. Any D1 (or D11) transformation can be restated in terms of a D11 (or D1)
operation. For instance, D1 = -1 x R1 x D11 (compare (17) and (22), below).

Delta Transformations

 IU   1 −1 0   k ϕ 0 0   ϕ1 
I  =  0 1 −1  0  
D1 (DAB):  V    kϕ 0   ϕ2  IUVW = D1 × k φ × φ123 (17)
 IW   −1 0 1   0 0 k ϕ   ϕ3 

 IU   −1 0 1   k ϕ 0 0   ϕ1 
I  =  1 −1 0   0  
D5 (D1 x R2 swap):  V    kϕ 0   ϕ2  IUVW = R 2 × D1 × k φ × φ123 (18)
 IW   0 1 −1  0 0 k ϕ   ϕ3 

 IU   0 1 −1  k ϕ 0 0   ϕ1 
I  =  −1 0 1   0  
D9 (D1 x R1 swap):  V    kϕ 0   ϕ2  IUVW = R 1 × D1 × k φ × φ123 (19)
 IW   1 −1 0   0 0 k ϕ   ϕ3 

 IU   1 0 −1  k ϕ 0 0   ϕ1 
I  =  −1 1 0   0  
D11 (DAC):  V    kϕ 0   ϕ2  IUVW = D11 × k φ × φ123 (20)
 IW   0 −1 1   0 0 k ϕ   ϕ3 

 IU   0 −1 1   k ϕ 0 0   ϕ1 
D3 (D11 x R2 swap): I  =  1 0 −1  0 kϕ

0   ϕ2  IUVW = R 2 × D11 × k φ × φ123 (21)
 V   
 IW   −1 1 0   0 0 k ϕ   ϕ3 

 IU   −1 1 0   k ϕ 0 0   ϕ1 
I  =  0 −1 1   0  
D7 (D11 x R1 swap):  V    kϕ 0   ϕ2  IUVW = R 1 × D11 × k φ × φ123 (22)
 IW   1 0 −1  0 0 k ϕ   ϕ3 

where kφ is as defined by (3) for the particular winding.

Similar to what was seen for the wye connections, one might say that there are only 3 unique
matrices above. Every one can be written of as the negative of one of the other matrices in the
list.

For eventual calculation of the overall transformation equations we will need to know the
transformer turns ratio. Referring to figures 10 and 16 for winding numbers, in a delta/wye and
delta/delta transformer, the turns ratios are:

24
 NW1 V
Dy Turns Ratio: RDy = = 3 D,LL
NW 4 Vy,LL
(23)
N V
Dd Turns Ratio: RDd = W1 = D,LL
NW 4 Vd,LL

Double Delta Transformations

A double delta refers to the case where there are two delta transformations in series. We can
simplify the equations somewhat by pre-multiplying the two matrices together. Some examples
of double delta transformation matrices include:

 1 −1 0   1 −1 0   1 −2 1 
DD1x1 = D1 × D1 =  0 1 −1 ⋅  0 1 −1 =  1 1 −2  (24)
     
 −1 0 1   −1 0 1   −2 1 1 

 1 −1 0   1 0 −1  2 −1 −1
DD1x11 = D1 × D11 =  0 1 −1 ⋅  −1 1 0  =  −1 2 −1 = -1× R × DD (25)
      2 1x1
 −1 0 1   0 −1 1   −1 −1 2 

 1 0 −1  1 0 −1  1 1 −2 
DD11x11 = D11 × D11 =  −1 1 0  ⋅  −1 1 0  =  −2 1 1  = R 1 × DD1x1 (26)
 0 −1 1   0 −1 1   1 −2 1 

Note the double deltas found in normal practice can be redefined as a combination of DD1x1, R1,
R2, and -1 transformation factors.

The double delta transformation does not arise too often, but it is physically possible, does
occur, and is easy enough for relay compensation. Targeting and relay testing will be especially
problematic whenever a double delta is utilized because current on any phase is seen in all the
differential comparators.

Zigzag Winding Transformations

To analyze current across a zigzag winding, let us continue to assume that, in balanced
conditions, all load flux in each leg sums to zero.

Use the Z1UV transformer of Figure 7 for our discussion (one can also refer to Figure 18) and
assume the flux on the primary winding set will be named φU,V,W . Taking into account that φW4,5,6
can be restated in terms of their equivalent φW1,2,3 the flux balance says that:

ϕU,Pri + ϕW1 − ϕW 3 = 0
φV,Pri + ϕW 2 − ϕW1 = 0 (27)
φW,Pri + ϕW 3 − ϕW 2 = 0

Let us convert the flux to the equivalent current equations, using (1). We have:

25
 NPr ik CIU,Pr i + NZk C ( IU,Z − IW,Z ) = 0
NPr ik CIV,Pr i + NZk C ( IV,Z − IU,Z ) = 0 (28)
NPr ik CIW,Pr i + NZk C ( IW,Z − IV,Z ) = 0

Solving (25) for IUVW, primary:

( IU,Z − IW,Z )
NZ
IU,Pr i = −
NPr i

( IV,Z − IU,Z )
NZ
IV,Pr i = − (29)
NPr i

( IW,Z − IV,Z )
NZ
IW,Pr i = −
NPr i

Expressing this in matrix form:

 IU.Pri   1 0 −1  IU,Z 

  NZ   
 IV.Pri  = − N  −1 1 0   IV,Z  (30)
 IW,Pri  Pr i
 0 −1 1   IW,Z 
 

The analysis will start by viewing the Z winding set side as an effective wye connection; terminal
current is equal to winding current. We wish to see how the current will appear to the lines on
the other side of the transformer. An examination of (29) and (30) will show that the primary will
see the secondary in a fashion that is similar to a delta transformation.

Using equations (1) and (2), let us convert (30) to an equation closer to the format used for the
wye and delta transformations earlier in the paper. Also note that when solving (28) for IPri we
introduced a negative correlation between primary current and secondary current. We do not
want the negative correlation. We want to have the concept of a positive flux creates a positive
current on both windings, so we will introduce an “x -1” factor.

 IU,Pri   1 0 −1  k ϕ,z 0 0   ϕ1(U,Z) 

  NZ  −1 1 0   0  
 IV,Pri  = N   k ϕ,z 0   ϕ2( V,Z)  (31)
 
 IW,Pri 
Pr i
 0 −1 1   0 0 k ϕ,z   ϕ3(W,Z) 

where kφ,Z=1/(NZkC). Note that the matrix is the same as for a D11 connection, typical of all the
wye-zigzag connection transformations.

We can reorganize this to the form of the delta and wye winding equations. We need two
different equations: one for the transformer as seen from the zigzag side bushings and one as
seen from the primary side bushings. First, from the zigzag side, the current in the zigzag
windings has the same transformation as a Y0 connection (though other Y# transformations are
possible via phase swaps and winding connection reversals). From the high side we will have a
variety of transformation equations that vary with how equations (27) through (30) work out for
the particular configuration. If the primary is connected in delta, the equations have additional
transformation factors, described below.

26
Zigzag Transformations, Seen From Zigzag Side

 IU,Z   1 0 0   k ϕ,z 0 0   ϕ1 
    
0   ϕ2 
ZZ Side  IV.Z  =  0 1 0   0 k ϕ,z IUVW = 1× k φ,Z × φ123 (32)
   k ϕ,z   ϕ3 
 IW,Z   0 0 1   0 0

1
where k ϕZ = .
NZk C

Phase swaps and winding connection reversal will affect the transformation equations by adding
-1, R1, and R2 transformation factors.

Zigzag Transformations, Seen From Wye Primary

 IU,Pr i   1 0 −1  k ϕ,Pr i 0 0   ϕ1 

     
Z1 (ZAB/UV):  IV,Pr i  =  −1 1 0   0 k ϕ,Pr i 0   ϕ2  IUVW = D11 × k φ,Pri × φ123 (33)
 IW,Pr i   0 −1 1   0 0 k ϕ,Pr i   ϕ3 
  

 IU,Pr i   −1 0 1   k ϕ,Pr i 0 0   ϕ1 
     
Z7 (=-Z1):  IV,Pr i  =  1 −1 0   0 k ϕ,Pr i 0   ϕ2  IUVW = -1× D11 × k φ,Pri × φ123 (34)
 IW,Pr i   0 1 −1 0 0 k ϕ,Pr i   ϕ3 
  

 IU,Pr i   1 −1 0   k ϕ,Pr i 0 0   ϕ1 
     
Z11 (ZAC/UW):  IV,Pr i  =  0 1 −1  0 k ϕ,Pr i 0   ϕ2  IUVW = D1 × k φ,Pri × φ123 (35)
 IW,Pr i   −1 0 1   0 0 k ϕ,Pr i   ϕ3 
  

 IU,Pr i   −1 1 0   k ϕ,Pr i 0 0   ϕ1 
     
Z5 (=-Z11):  IV,Pr i  =  0 −1 1   0 k ϕ,Pr i 0   ϕ2  IUVW = -1× D1 × k φ,Pri × φ123 (36)
 IW,Pr i   1 0 −1  0 0 k ϕ,Pr i   ϕ3 
  

1
where k ϕ,Pr i = .
NPr ik C

Phase swaps and winding connection reversal will affect the transformation equations by adding
-1, R1, and R2 transformations.

Zigzag Transformations, Seen From D1 Primary

The zigzag with a delta winding introduces the issue of back to back delta transformations.
Since a zigzag is effectively a delta as seen from the primary, we will have a “double delta”
transformation. The approach is to simply multiply the two delta effects together as was seen in
the double delta discussions above. i.e., if a primary is connected in delta, we simply add the

27
appropriate delta transformation to the overall transformation equation. The equation can be
simplified a bit using the D1xD11 transformation previously mentioned (25).

 IU,Pr i   2 −1 −1  k ϕ,Pr i 0 0   ϕ1 

     
D1xZ1 (ZUV):  IV,Pr i  =  −1 2 −1  0 k ϕ,Pr i 0   ϕ2  IUVW = D1 × D11 × k φ × φ123 (37)
 IW,Pr i   −1 −1 2   0 0 k ϕ,Pr i   ϕ3 
  

 IU,Pr i   1 −2 1   k ϕ,Pr i 0 0   ϕ1 
     
D1xZ11 (ZUW):  IV,Pr i  =  1 1 −2   0 k ϕ,Pr i 0   ϕ2  IUVW = D1 × D1 × k φ × φ123 (38)
 IW,Pr i   −2 1 1   0 0 k ϕ,Pr i   ϕ3 
  

The reader can use this example to determine how a D11 connection changes the equations.

For eventual calculation of overall transformation equations we will need to know the
transformer turns ratio. Referring to Figure 18 and 19 for winding numbers and looking at the
windings on one core leg, in a classical zigzag transformer, the turns ratios are:

NW 4 = NW 7
NW1 V
Yz Ratio: R Yz = = 3 Y,LL (39)
NW 4 or NW7 VZ,LL
NW1 V
Dz Ratio: RDz = = 3 D,LL
NW 4 or NW7 VZ,LL

As a side note, setting N4 and N7 different from one another allows a continuous variety of
phase shifts across the transformer. Such transformers are used in harmonic cancellation
techniques in VFD applications and in large phase shifting transformers used for power flow
control in important transmission lines.

CT Connections Transformations
A wye CT is fairly straightforward and simply changes the transformer current by a specific ratio.
A complication can occur if a CT connection is made backwards such that it adds a -1 factor to
the current balance equation. This -1 factor might arise if one of the CT sets is connected
backwards relative to the relay manufacturer’s instruction manual.

 IU,Relay   NCT 0 0   IU 
  
YCT0 :  IV,Relay  =  0 NCT 0   IV 
 IUVW,Relay = NCT × IUVW (40)
   0 NCT   IW 
 IW,Relay  
0

 IU,Relay   −NCT 0 0   IU 
  
YCT6 :  IV,Relay  =  0 −NCT 0   IV 
 IUVW,Relay = -1× NCT × IUVW (41)
   0 −NCT   IW 
 IW,Relay  
0

28
where NCT is a step-down factor of:

ISec
NCT = (42)
IPr i

A delta CT complicates the transformer differential relay compensation by adding a second

delta transformation effect. A delta CT connection will virtually always be a D1 or a D11 form
transformation, but other connections are not impossible.

 IU,Relay   NCT 0 0   1 −1 0   IU 
  
DCT1 (DAB):  IV,Relay  =  0 NCT 0   0 1 −1  IV  IUVW,Relay = NCT × D1 × IUVW (43)
   0 NCT   −1 0 1   IW 
 IW,Relay  
0

 IU,Relay   NCT 0 0   1 0 −1  IU 

  
DCT11 (DAC):  IV,Relay  =  0 NCT 0   −1 1 0   IV  IUVW,Relay = NCT × D11 × IUVW (44)
   0 NCT   0 −1 1   IW 
 IW,Relay  
0

It might be possible also to have reverse connected CTs that are then connected in delta,
resulting in negated versions of DCT1 and DCT11, creating DCT7 and DCT5, respectively.

Net Compensation Equations

Relay Differential Comparator Equations

As previously discussed, the assumption is that when the transformer is operating correctly, the
flux on each leg due to load current (or external fault current) sums to zero. A differential relay
has three comparators. The three comparators compare through current (current sensed at
each input) to error current (difference between the two through currents), ideally on each phase
of the transformer, but when delta connections are between the relay and the transformer, delta
summations are monitored instead. The details of what the relay does with the current it senses
in the comparator is beyond the scope of the paper and varies with each relay manufacturer, but
the current the comparator is monitoring is about the same in most current differential relays.

The error current equation below assumes the relay is configured in such a way that current into
one winding set and out the other winding set results in positive current into one input of the
relay and negative current (i.e., 180° phase shift) into the other input. The summation (not the
difference) of the two currents would therefore normally be 0. This approach is in accordance
with the previously described flux balance equation, (5). Hence, the three differential
comparators in the relay monitor:

29
 Through Current: I1Compensated and ICompensated
4
U Comparator: (45)
Error Current: I1Compensated + ICompensated
4

Through Current: ICompensated

2 and I5Compensated
V Comparator: (46)
Error Current: ICompensated
2 + I5Compensated

Through Current: ICompensated

3 and ICompensated
6
W Comparator: (47)
Error Current: ICompensated
3 + ICompensated
6

The comparators were named U, V, and W above, but be aware that this can lead to some
confusion on faulted phase identification when delta connections are involved due to the
associated inter-connection of phases.

Example Compensation Equations

Each transformer or CT connection has its own effect on the current that eventually makes its
way to the relay. Let’s look at a few examples to see how this works out.

Example 1: Dy1 Transformer

Dy1 Transformer, 115kV (D1(AB)) / 13.8kV (Y0);
20MVA (IFull Load =100A @ 115kV and 837A @ 13.8kV);
Primary CTs 200:5, Y0;
Secondary CTs 1200:5, YCT6 (note the CT has been inverted).

Current @ relay, 115kV:

IUVW.115 = D1 × k φ,115 × φ123 (see Eq. 17)
IUVW,Relay = NCT,115 × IUVW ,115 (see Eq. 40)

The net 115kV equation is:

IUVW,Relay,115 = NCT,115 × D1 × k φ,115 × φ123 (48)

Current @ relay, 13.8kV

IUVW,13.8 = 1× k φ,13.8 × φ 456 (see Eq. 10)
IUVW,Relay,13.8 = -1× NCT,13.8 × IUVW,13.8 (see Eq. 41)

The net 13.8kV equation is:

IUVW,Relay,13.8 = -1× NCT,13.8 × k φ,13.8 × φ 456 (49)

Note that in balanced conditions φ123 + φ456 = 0, and as seen by comparing (48) and (49), what
started out as a balanced condition in the transformer has become two different current values
at the relay. The balanced flux has been modified by:
• Transformer Turns (part of kφ)
• Transformer winding connections
• CT ratios
• CT connections

30
The approach used in numeric relays is to compensate by adding multiplying factors CF115 and
CF13.8 to (48) to create a target multiplication factor that puts the equivalency back into the
sensed current, so that the error current is 0 in the differential comparator equations (43), (44),
and (45). The initial unmodified compensation equation is:

 CF115 × ( NCT,115 × D1 × k φ,115 × φ123 )  +  CF13.8 × ( -1× NCT,13.8 × k φ,13.8 × φ 456 )  = 0 (50)
   

The part in the parenthesis in (50) represents the two sets of current inputs to the relay IRelay.
and CF# represents the math the relay needs to apply to the current to see the currents as
equivalent. Let us simplify (50) by first collecting the magnitude compensation adjustments to
the left. The resultant equation is:

 CF115 ×
 ( ( -1× k φ,115 × k -1 -1
) )
φ,13.8 × NCT,115 × NCT,13.8 × ( D1 × φ 123 )
 + [ CF13.8 × φ 456 ] = 0

(51)

This allows us to state CF115 in terms of only phase compensation adjustments and CF13.8 in
terms of magnitude and phase adjustments. In this case, CF115 has no phase compensation
adjustments, but CF13.8 has both magnitude and phase adjustments. To make (51) true, we
need:

CF115 = 1 (52)

(
CF13.8 = -1× k φ,115 × k -1 -1
φ,13.8 × NCT,115 × NCT,13.8 × D1 ) (53)

CF13.8 Magnitude Compensation Factor

The magnitude effect has to account for the transformer and CT turns ratio. Recall from (1) and
(2) that kφ,# = 1/(N#kC) where N# is the # of transformer coil turns, and kC is a constant for the
core design. Recall also the turns ratio calculations of (23). The overall ratio of kφ,# is:

NTurns,13.8 13.8
k φ,115 × k -1
φ,13.8 = = = 0.069282 (54)
NTurns,115 3 ⋅ 115

The CT ratio factor (42) is:

5
NCT,115 ⋅ NCT,13.8
-1
= 200 = 6 (55)
5
1200

Hence, the relay multiplies the current coming into the 13.8kV side by the factor:

φ,13.8 × NCT,115 × NCT,13.8  = [ -1× 0.069282 × 6 ]

CF13.8,Mag =  -1× k φ,115 × k -1 -1

 -0.415692 0 0  (56)
=  0 -0.415692 0 

 0 0 -0.415692 

31
Some may note a missing “sqrt (3)” factor in this equation that is used to reduce the magnitude
of current that is seen after a delta summation. This matter will be discussed later, under
“Application to Protective Relay Settings.”

Net Compensation Equations

A phase compensation factor is required when different phase transformation factors (e.g., R# or
D#) show up in the two phase compensation factors, such as was found in (52) and (53).

The compensation is on the currents making it to the relay, so let us give some names to the
relay currents:

 I1,Relay   I115,U,Relay   I4,Relay   I13.8,U,Relay 

       
 I2,Relay  =  I115,V,Relay   I5,Relay  =  I13.8,V,Relay  (57)
       
 I3,Relay   I115,W,Relay   I6,Relay   I13.8,W,Relay 

The 115kV phase compensation is the most basic equation possible:

Compensated
I123,Relay = CF115,Phase × I123,Relay
Compensated
I123,Relay = 1× I123,Relay (58)
 I1,Relay
Compensated
  1 0 0   I1,Relay 
 Compensated    
 I2,Relay  =  0 1 0   I2,Relay 
 Compensated   0 0 1   
 I3,Relay     I3,Relay 

The overall (magnitude and phase) 13.8kV compensation equation is:

ICompensated
456,Relay = CF13.8,Mag × CF13.8,Phase × I456,Relay
ICompensated
456,Relay ( -1
= -1× k φ,115 × k φ,13.8 × NCT,115 × N-1 )
CT,13.8 × D1 × I456,Relay (59)
 ICompensated
4,Relay
  −0.415692 0 0   1 −1 0   I4,Relay 
 Compensated     0 1 −1  I 
 I5,Relay = 0 −0.415692 0    5,Relay 
 Compensated   −0.415692   −1 0 1   I6,Relay 
 I6,Relay   0 0
 

This in clearer terms is:

ICompensated
4,Relay = −0.415692 ⋅ ( I4,Relay − I5,Relay )
ICompensated
5,Relay = −0.415692 ⋅ ( I5,Relay − I6,Relay ) (60)
ICompensated
6,Relay = −0.415692 ⋅ ( I6,Relay − I4,Relay )

The results of (58) and (60) are then fed into the comparator equations (45), (46), and (47).

32
Example 2 - Dz2 with Delta CT Configuration
Let us look at a more complicated application that gives a more involved application of the
phase transformation factors. We will use a fairly uncommon transformer and assume someone
selects fairly strange CT connections. Let’s see if and how the relay could compensate.

Dz2 Transformer
115kV (D1) / 13.8kV (Z11(AC/UW))
20MVA (for IFull Load =100A @ 115kV and 837A @ 13.8kV)
Primary CTs 200:5, Y0.
Secondary CTs 1200:5, D11

The net 115kV equation is (Note Zigzag with Delta primary created Double Delta):

IUVW,Relay,115 = 1× NCT,115 × D1 × D1 × k φ,115 × φ123 (61)

The net 13.8kV equation is (Note CT is DCT,11):

IUVW,Relay,13.8 = D11 × NCT,13.8 × k φ,13.8 × φ 456 (62)

The overall compensation equation is:

 CF115 × ( 1× NCT,115 × D1 × D1 × k φ,115 × φ123 )  +

 
(63)
 CF13.8 × ( D11 × NCT,13.8 × k φ,13.8 × φ 456 )  = 0
 

Again, we can rearrange (63) to bring the magnitude compensation factors to one side of the
equation to allow us to state CF115 in terms of only phase compensation adjustments, and CF13.8
in terms of magnitude and phase adjustments.

 CF ×
 115 ((k φ,115 × k -1 -1
)
φ,13.8 × NCT,115 × NCT,13.8 × ( D1 × D1 × φ123 ) )
+
 (64)
 CF13.8 × ( D11 × φ 456 )  = 0

and hence to make (64) true, we need:

CF115 = D11 (65)

(
CF13.8 = k φ,115 × k -1 -1
)
φ,13.8 × NCT,115 × NCT,13.8 × ( D1 × D1 ) (66)

The magnitude adjustment factor is nearly the same as the previous example, but the Y6 CT
connection added a negative in the previous example, and since this is a delta/zigzag winding,
the kφ is slightly different. Applying (3) and (39) we will find:

NTurns,13.8 3 ⋅ 13.8
k φ,115 × k -1
φ,13.8 = = = 0.040 (67)
NTurns,115 115

and the CT ratio is 6, as it was in the previous example. The net magnitude adjustment factor is:

33
 φ,13.8 × NCT,115 × NCT,13.8  = [ 0.040 × 6 ]
CF13.8,Mag =  k φ,115 × k -1 -1

 0.240 0 0 
(68)

= 0 0.240 0 
 0 0 0.240 

The 115kV phase compensation equation is:

Compensated
I123,Relay = CF115,Phase × I123,Relay
Compensated
I123,Relay = D11 × I123,Relay (69)
 I1,Relay
Compensated
  1 0 −1  I1,Relay 
 Compensated    
 I2,Relay  =  −1 1 0   I2,Relay 
 Compensated   0 −1 1   
 I3,Relay     I3,Relay 

This in clearer terms is:

Compensated
I1,Relay = ( I1,Relay − I3,Relay )
ICompensated
2,Relay = ( I2,Relay − I1,Relay ) (70)
ICompensated
3,Relay = ( I3,Relay − I2,Relay )

The 13.8kV compensation equation includes a “double delta” D1xD1 factor, introduced
previously (24). The net13.8kV magnitude and phase compensation equation is:

ICompensated
456 ,Relay = CF13.8,Mag × CF13.8,Phase × I456,Relay
ICompensated
456 ,Relay (
= k φ,115 × k -1 -1
)
φ,13.8 × NCT,115 × NCT,13.8 × ( D1 × D1 ) × I456,Relay (71)
 ICompensated
4,Relay
  0.240 0 0   1 −2 1   I4,Relay 
 Compensated    
 I5,Relay = 0 0.240 0   1 1 −2   I5,Relay 
 Compensated   0 0.240   −2 1 1   I6,Relay 
 I6,Relay   0
 

In clearer terms, this is:

ICompensated
4,Relay = 0.240 ⋅ ( I4,Relay + I6,Relay − 2 ⋅ I5,Relay )
ICompensated
5,Relay = 0.240 ⋅ ( I4,Relay + I5,Relay − 2 ⋅ I6,Relay ) (72)
ICompensated
6,Relay = 0.240 ⋅ ( I5,Relay + I6,Relay − 2 ⋅ I4,Relay )

The results of (70) and (73) are then fed into the comparator equations (43), (44), and (45).

34
Example 3 - Dy1 but Phase Transformations Cancel Out
Assume the same system as Example 1 but let us hook the secondary CT in D1. The
compensation equation (50) becomes:

 CF115 × ( NCT,115 × D1 × k φ,115 × φ123 )  +  CF13.8 × ( D1 × NCT,13.8 × k φ,13.8 × φ 456 )  = 0 (73)

   

The relay sees the same D1 phase compensation on both inputs and hence the relay adds no
additional phase compensation transformation to the incoming currents. It will only need to do
magnitude compensation. The rest of the compensation factor analysis is approximately the
same as the previous examples so will not be developed.

Application to Protective Relay Settings

A few notes on how the above material works out in a numeric current differential relay:

1) Relay Magnitude Compensation Factors.

Relays can be configured to adjust for the magnitude of currents via two methods:
a) User selected compensation taps. In this classical method the user takes into account the
transformer ratio and CT ratio that is part of 47, as well as the normal loading that the user
expects, and selects a current tap that tells the relay what constitutes balanced current
magnitudes. The taps may include the sqrt(3) factor described in the previous paragraph,
depending on relay settings instructions from the manufacturer.
b) Enter Xfmr parameters and let relay decide compensation factors. In this approach, the
user enters transformer data, such as voltages, winding configuration, CT ratio, CT
configuration, and full load MVA, and the relay selects the appropriate current magnitude
adjustment factor. This approach is more user friendly and likely results in fewer mistakes, but
there may be an argument that users should know what is going on and not let the relay do their
thinking for them.
c) Some relays allow customers to use either approach a) or b) via a selection in the relay
logic.

2) Additional 1/Sqrt(3) Factor in Magnitude Adjustments for Delta Configurations

When balanced three phase currents are in the transformer, one effect of a D#
transformation is to increase the magnitude of balanced current by sqrt(3). Some relay
manufacturers include a 1/sqrt(3) factor in the magnitude compensation factor to remove the
current increases to balanced current flow that delta transformations cause. Including the sqrt(3)
inside the relay calculations tends to make relay settings more intuitive to some end users.

3) Per-Unitizing Sensed Current

Something a bit more involved is done to current magnitudes than just multiplying the
current on one input by the magnitude adjustment factor. Current on both inputs are multiplied
by an appropriate number to bring the current down to a common per-unitized and normalized
level, commonly referred to as “per unit of tap” (dating back to electromechanical relays). The
trip decisions and instruction manuals refer to things in a per-unitized manner, so a user has to
be aware of the per-unitized math that occurs inside the relay. Even beyond the per-unitization
on a tap basis, in a real world protective relay, if one could monitor the internal machine coding,
the relay also includes additional scaling of currents to levels that give the best use of the relay’s
numerical methods and a wide range to the current levels the relay can work with accurately.

4) Configuring the Phase Compensation Equations

35
 It can be seen that just as there are a very large number of possible ways to configure a
transformer and its CTs, there is a similarly very large number of possible phase compensation
equations that might be needed. However, the relay does not need to have a library in its
memory bank of every possible transformer and CT configuration and the resultant phase
compensation that it must apply. This paper has made some effort to show that many
transformation equations can be seen as the multiplication of 2 or 3 other transformations.
Hence, the relay can have a subset of possible transformation equations that can be applied as
needed to create the overall compensation equation for a given application. Once these
transformation equations are in the relay memory, the next issue is when and how to apply
them. The response from the relay manufacturers on how the various transformations are
selected and used and presented to the end user varies quite a bit.
• Some manufacturers supply a set of transformation equations and ask the user to decide
how to apply them, given instructions in the manual that hopefully are sufficient. Extremely
detailed analysis, such as found in this paper, is not too common though. Documentation
and instruction manuals are always subject to misunderstanding, especially by engineers
who have limited time to study the matter. This approach puts the user in the “think hard
because there is lots of room for messing up” mode.
• Some relays, especially the basic lower budget versions each manufacturer supplies, have
a reduced degree of support of all the possible transformers and CT connections. This
reduces the confusion factor and makes correct setup easier for the mainstream
applications, but it is done at the expense of flexibility.
• Some manufacturers make the setting a bit more user friendly by listing the phase
compensation equations that should be used for a given transformer configuration and
phase shift. They appear to say that if you know phase shift and a few other items on the
transformer, these are the equations you want to use. Without a detailed analysis of where
the equations come from a user might be confused and still make a selection error.
• Some relay manufacturers attempt to remove the chance of user error by having the user
state the specific transformer and CT configurations. It becomes the task of the relay internal
coding to decide which phase compensation equations are proper for the configuration that
was entered. This approach takes the previous bullet one step farther from a good
understanding of what is being calculated by the relay, so has its drawbacks too. Hopefully
this paper supplies the needed information so that a user can reason through what is going
on inside the relay.

Conclusion/Summary
Hopefully the reader has learned from the above material, 1) how the myriad of “round the
clock” power transformers are internally configured, as well as some understanding of the
nomenclature and figures that are used, 2) the various ways delta and zigzag transformers can
be configured, 3) the ways that the myriad of possible transformer and CTs connections affect
the current seen by a transformer differential relay, and 4) how a transformer differential relay
can numerically compensate for the differences in any transformer configuration to still
determine balanced current conditions.

Author Biographies
Larry Lawhead received a BS degree in Engineering from Duke University in 1982. He worked
at Potomac Electric Power Company in the Substation Test and System Protection groups until
joining Basler Electric in 1989. Mr. Lawhead has held a variety of Sales and Marketing positions

36
with Basler and is currently Principal Application Engineer, based at Basler's Highland, Illinois,
headquarters. Mr. Lawhead is a member of the IEEE-PES and is active on several working
groups of the Power Systems Relay Committee.

Randy Hamilton is a Principal Electrical Design Engineer for Basler Electric Company. He has
worked for Basler for more than 25 years, with experience in test and design in the areas of
magnetic transformers, power supplies, voltage regulators, and, for the last ten years, protective
relays. Randy graduated Magna Cum Laude from Southern Illinois University.

John Horak received his BSEE degree from the University of Houston in 1988 and his MSEE
degree, specializing in power system analysis, from the University of Colorado, Denver, in 1995.
John joined Basler Electric in 1997 and is a Senior Application Engineer. Previous work
included ten years with Stone and Webster Engineering where he was on assignment for six
years in the System Protection Engineering offices of Public Service Company of Colorado. His
work has included fault calculations, relay coordination settings, control design, and equipment
troubleshooting. Other previous employers include Houston Lighting and Power and Chevron.
He has authored, coauthored, and presented technical papers at a variety of industry events
and conferences. John is a member of IEEE-IAS and -PES and has P.E. license in Colorado.

References
1) ANSI/IEEE C57.12.00 IEEE Standard General Requirements for Liquid Immersed
Distribution, Power, and Regulating Transformers

2) IEEE C57.12.70 IEEE Standard Terminal Markings and Connections for Distribution and
Power Transformers

3) IEC 60076-1 Power Transformers – Part 1: General

4) J&P Transformer Book, A.C. Franklin, D.P. Franklin (12th Edition, 1998, changes author
to Martin Heathcote); Butterworth-Heinemann Division, Reed Educational and
Professional Publishing Ltd.

37
 Route 143, Box 269, Highland, Illinois U.S.A. 62249
If you have any questions or need Tel +1 618.654.2341 Fax +1 618.654.2351
No. 1300 North Zhongshan Road
Wujiang Economic Development Zone
additional information, please contact e-mail: [email protected] Suzhou, Jiangsu Province - P.R. China 215200
Tel +86(0)512 6346 1730 Fax +86(0)512 6346 1760
Basler Electric Company. e-mail: [email protected]
Our web site is located at: P.A.E. Les Pins, 67319 Wasselonne Cedex FRANCE
Tel +33 3.88.87.1010 Fax +33 3.88.87.0808
http://www.basler.com e-mail: [email protected]
e-mail: [email protected]