ABSTRACT

Accepted voltage quality standards focus on two areas:

! voltage regulation the magnitude of the voltage
! waveform quality the shape of the voltage
waveform
The growing application of diverse embedded
generation worldwide is creating new problems with
voltage regulation. Increasing amounts of non-linear
loads, in such form as variable-speed drives, are causing
waveform quality to deteriorate.
Traditionally automatic voltage control (AVC) through
on-load tap-changers (OLTCs) has been used to regulate
busbar voltage, although its success has been limited.
This paper highlights the deficiencies in these complex
and restrictive schemes and outline the Transformer
Automatic Paralleling Package (TAPP) scheme, which
has gained acceptance worldwide.
The TAPP scheme has been available for some years in
an analogue form, and is well liked for the simplicity
and effectiveness of its operation. It has now been
developed as a numeric AVC relay. This paper will
show how further development of the control algorithm
allows improved control of customers voltage, even in
areas with low, or varying, system power factors, which
in the past have been difficult to manage.
The numeric relay is now able to tackle the other area of
concern with voltage quality waveform quality.
Although unable to greatly influence the waveform
quality, it can provide information allowing the power
utility to pinpoint areas of the network for further
investigation.
The paper ends by proposing how voltage control
should develop in the future to cater for increasing
diversity of users and generators of electricity, making
use of new technologies.

INTRODUCTION
It is incumbent upon power utilities to maintain
customers supplies within specified parameters under
varying conditions of load. In the vast majority of cases,
the last point at which the power utility can influence
the voltage is the on-load tap-changer (OLTC). A
number of factors influence the optimum tap change
operating point:
! the power system voltage level,
! loading,
! power factor, and,
! if transformers are operating in parallel, the
magnitude of any inter-transformer circulating
current.
As power systems have developed and become more
complex, the potential efficiency savings of a
comprehensive voltage control system cannot be
realised by the majority of existing AVC relays. Indeed
their complexity has often resulted in voltage control
not being implemented.
After examining how voltage control has developed to
meet this need, the paper will describe a simple solution
to remove the unneeded complexity, based upon the
fundamental requirements of the power system.

THE BASICS OF VOLTAGE CONTROL
Simple AVC scheme
AVC
Relay
V
VT
I
CT
R
line
+ jX
line
V
eff

FIGURE 1: Transformer connected to single load
The simplest form of AVC can be used where a single
transformer supplies a single load (FIGURE 1). If the
load is some distance from the transformer, there will be
a voltage drop in the line. The AVC relay makes an
estimate of the voltage at the load using a model of the
IMPROVED VOLTAGE QUALITY THROUGH ADVANCES IN VOLTAGE CONTROL TECHNIQUES
VP Thornley and NJ Hiscock
VA TECH Reyrolle A,C&P Fundamentals Limited
line and applies Line Drop Compensation (LDC) based
on the conditions seen at the transformer:
# $
line line CT VT eff
X R I V V . j . % & ' (1)
target eff dev
V V V & ' (2)
where V
eff
is the effective voltage at the single load, V
VT

and I
CT
are the measured voltage and current
respectively, R
line
+ j.X
line
is the model for the
impedance of the line, V
dev
is the voltage deviation from
target, and, V
target
is the target voltage.
The above represents the ideal situation: in reality there
are usually a number of loads on a transformer
distributed at different distances (electrically) from the
transformer, so the model of the line will always be a
compromise. It can be shown an optimum voltage
control will establish a constant voltage point at the
electrical mid-point of the network, thus achieving a
minimum overall variation between the no-load and
full-load conditions.

Parallel Transformers
It is normal practice for power utilities to parallel
transformers to obtain a higher security of supply. In
FIGURE 2, which shows an example with two
transformers, the load on each transformer (discounting
any circulating current) is half of the total load, so the
model will produce half the required voltage boost.
AVC
Relay
V
VT
I
CT,1
R
line
+ jX
line
AVC
Relay
V
VT
I
CT,2
I
load
Transformer 1 Transformer 2
V
eff

FIGURE 2: Parallel transformers connected to
single load
If the effective terminal voltages of the paralleled
transformers are not identical, a circulating current will
flow around them. This will be highly reactive since the
transformers are highly inductive. If two paralleled
transformers operate the simple AVC scheme described
above, eventually one transformer will be on the highest
tap and the other on the lowest tap. The busbar voltage
will be an average of their terminal voltages and a high
amount of circulating current will flow between them.
This will cause an unnecessary power loss within the
transformers and the network, reducing their useful
capacity and efficiency, and can result in the loss of one
transformer due to overload, and a consequent severe
overvoltage.

EXISTING VOLTAGE CONTROL SCHEMES
One of the main aims of all the schemes discussed
below is to keep the transformers taps together, to
remove circulating current. To a greater or lesser extent
they all achieve this, however as will be seen some
schemes create other problems in doing so.

Master-Follower
In this scheme, one AVC relay in each paralleled group
of transformers is nominated as the master; all other
AVC relays in the group are set as followers, or slaves.
When the master issues a tap instruction to its tap-
changer, the followers all issue identical instructions to
their tap-changers. All the transformers keep in step.
However, if the incoming voltages are different, a
permanent circulating current will flow.

Negative-Reactance Scheme
This scheme makes use of the fact that circulating
current is reactive. The reactive element of the
measured current is used to bias the measured voltage in
order cause the transformers to tend to a tap position
where any circulating current is minimal.

Transformer Automatic Paralleling Package (TAPP)
TAPP is a modified negative-reactance AVC scheme
that eliminates the main disadvantages of the negative-
reactance scheme. The circuit (FIGURE 3) includes a
negative-reactance control (or coupling element). The
modification is in the three buswires, which enable the
currents of all paralleled transformers in the substation
to be summed.
FIGURE 3 (a) shows circulating currents only, that flow
in opposite directions in the two transformers. As the
circulating current does not flow through the LDC
element, it will not affect the LDC control. This cures
the first disadvantage of the negative-reactance scheme.
FIGURE 3 (b) shows load currents only, that flow in the
same direction in both transformers. It can be seen that
the load current presented to each of the LDC elements
is the total load of the substation. If one transformer is
switched out, the load seen by the LDC element remains
unchanged, so no change to LDC settings is required.
As a further advantage, the switched out transformer
continues to respond to changes in load so when it is
switched back into service it is at the correct tap to
ensure that the busbar voltage remains constant.
A crucial point to note is that the buswires are used only
for the purposes of LDC. The coupling control will still
be effective when the buswires are not used. A corollary
of this is that the scheme will eliminate circulating
currents not only on a common busbar, but also across a
network.
The control algorithm for this scheme is:
( j
e . .
line load VT eff
Z I V V & ' (3)
tr CT ccv
X I V j . ' (4)
# $
ccv target eff dev
V V V V & & ' (5)
where the model for the line R
line
+ j.X
line
is now
represented in mag-arg form by Z
line
.e
j(
, I
load
is the
summated load, X
tr
is the reactance of the transformer
and V
ccv
is the corrective coupling voltage.

Scheme Comparisons
The three schemes outlined above are compared in
TABLE 1.
The ideal of voltage control is to achieve the correct
system voltage with minimal circulating current under
all system conditions. Master-Follower is cumbersome
and complex, difficult to operate and rarely produces
acceptable results. The negative-reactance scheme
eliminates most of the failings of master-follower but
still fails to provide an effective LDC.
While TAPP can achieve superior performance to both
master-follower and standard negative-reactance
schemes in nearly all circumstances, it suffers from
some minor disadvantages itself, which in an ideal
world would be eliminated.
The area that needs most attention is the LDC
performance under varying power factors. All the
existing AVC schemes do suffer from droop when the
power factor is less than unity.
Coupling
element
LDC
element
Coupling
element
LDC
element
Coupling
element
LDC
element
Coupling
element
LDC
element
(a) circulating current
(b) load current
Transformer 1 Transformer 2
Transformer 1 Transformer 2
I
load
circulating
current

FIGURE 3: TAPP scheme showing circulating and load currents
TABLE 1: Comparison of AVC Schemes

M
a
s
t
e
r
-
F
o
l
l
o
w
e
r

N
e
g
a
t
i
v
e

R
e
a
c
t
a
n
c
e

T
A
P
P

Differing incoming voltages ! " "
Differing transformers ! " "
Differing tap spacings ! " "
Paralleling across the network ! " "
No re-configuration required for
transformer switch-out
! ! "
Simple scheme to implement ! " "
Linear LDC at normal system
power factors
" ! "
Linear LDC at all system power
factors
! ! !
Differing CTs and CT phases " " !

With regard to the TAPP scheme the reason for the
droop is seen by reference to FIGURE 3(b). Although
circulating currents do not flow in the LDC element,
transformer load currents do flow in the coupling
circuit. Resistive currents have little effect on the
coupling control. However, as the power factor of a load
decreases, the increasing reactive load causes the
coupling control to respond. It is normal with the TAPP
method to take account of some reactive component in a
load current and to shift the angle by 15) before
applying it to the coupling circuit. FIGURE 4 shows a
typical AVC response of the TAPP scheme at different
power factors for an application with transformer
impedance of 10% and required LDC boost at full load
of 2.5%.
100
101
102
103
0 50 100
Load Magnitude/%
pf = 1.00
pf = 0.97
ideal
pf = 0.95
pf = 0.90
pf = 0.80

FIGURE 4: Busbar voltage for theTAPP scheme

DEVELOPMENT OF NUMERIC VOLTAGE
CONTROL
There are many opportunities afforded by numeric AVC
relays, however, the potential has not been exploited.
Numeric AVC relays to date have implemented
traditional analogue techniques together with a
sophisticated user interface and the ability for remote
communications. This section looks at ways in which
numerical methods are now employed to bring further
improvements to voltage control than those already
provided by TAPP.

Voltage Regulation
Numerical techniques provide the method of removing
poor performance for low power factors. Because it is
possible to manipulate magnitudes and angles as the
designer chooses, the angle shift required in the
coupling circuit can be tailored to the system power
factor. On most circuits, the system power factor will
not vary widely. The user applies the normal system
power factor to the AVC relay as a set point.
Embedded generation is becoming increasingly
common, both synchronous and asynchronous. One
feature of asynchronous machines is their ability to
export real power while absorbing reactive power. Their
characteristics can cause the power factor to swing
substantially and a fixed power factor set point is not a
viable option. An alternative scheme, known as the true
circulating current scheme, is described below and can
be used in these circumstances.
V
VT
I
CT,1
V
VT
Transformer 1
Transformer 2
I
CT,2
I
circ,1
I
circ,2
I
load,1
I
load,2

FIGURE 5: Multiple loads applied to numeric TAPP
With numerical techniques, the various components that
make up the measured current I
CT
can be completely
isolated, giving improved performance. FIGURE 5
shows the current seen by two AVC relays I
CT,1
and
I
CT,2
, with respect to their phase voltages V
VT
(when the
transformer LV circuit breakers are closed the measured
voltages will be identical). The load currents, I
load,1
and
I
load,2
, have the same power factor. Transformer 1 is on a
higher tap position than transformer 2, hence a
circulating current will flow represented by I
circ,1
and
I
circ,2
in the diagram. If the measured currents, I
CT,1
and
I
CT,2
, are summated, the network power factor can be
found. The true load on each transformer and its
contribution to circulating current can be established;
therefore, compensation is always correct resulting in
the complete elimination of droop in the AVC response.
However, as the currents must be summed for this
technique to work, the ability to parallel across
networks is lost. A technique that will eliminate droop
and allow paralleling across networks is to use a
combination of the fixed power factor set point and the
true circulating current method.
Consider the network of FIGURE 6. Transformers 1 and
2 are at the same site and so have strong coupling;
transformer 3 is some distance away and so the coupling
will be weaker. Any mismatch in transformer open-
circuit terminal voltages between transformers 1 and 2
will result in a high circulating current, while a
difference between transformers 1 and 3 will result in a
much lower circulating current because of the network
impedance.
The circulating current method fully eliminates
circulating currents between transformers 1 and 2 and
allows them to provide correct LDC under any power
factor. A weaker compensation is also applied by
measuring the difference between the transformer
current and the power factor set point, I
res
in FIGURE 7.
Although the system power factor might fluctuate, the
set point applied should be the average power factor.
This prevents transformers 1 and 2 drifting from
transformer 3 any tapping action will naturally keep
them together. If there is circulating current flowing
between the two sites it will be small so the LDC should
not exhibit droop. The control algorithm for this scheme
is:
# $
tr res tr circ target eff dev
Z k I Z I V V V . . . & & & ' (6)
where Z
tr
is the reactance of the transformer. Note that
all quantities in this equation are scalar.
At any time, if the system power factor is known it is
possible to calculate the circulating current so the
AVC
Relay
V
VT,a
I
CT,1
R
line,1
+ jX
line,1
AVC
Relay
I
CT,2
I
load,a
Transformer 1 Transformer 2
AVC
Relay
I
CT,3
Transformer 3
R
line,2
+ jX
line,2
V
VT,a
V
VT,b

FIGURE 6: Coupling across a network
V
VT
I
CT,1
V
VT
Transformer 1 Transformer 2
I
CT,2
I
circ,1
I
circ,2
I
load,a
I
load,a
I
res,1
I
res,2
V
VT
Transformer 3
I
CT,3 I
res,3
sys pf
sys pf
sys pf

FIGURE 7: Eliminating circulating currents across a network
numeric TAPP method will be very accurate. Therefore,
an adaptive algorithm is needed that can:
(a) establish the optimum value k, and,
(b) progressively estimate the true system power
factor over time.
Work is currently being undertaken to achieve this.

Transformer Switching
As numeric voltage control develops, functions will be
added to further improve voltage regulation. The
master-follower scheme suffers from inherent
complexity in its operating method. In order not to fall
into a similar trap with what, on the face of it, is a
simple task, the underlying principle that is used when
developing the scheme is to answer the question: What
would you do if you were there? The development
process is assisted by attempting to recreate what an
engineer does if he operates the tap-changers manually.
The result is a scheme that is based upon the
fundamental measurands of the system, not on a
complex arrangement of inter-transformer wiring and
conditions.
This philosophy finds benefits in the area discussed
here: switching transformers out of service. As was
outlined previously, the TAPP method will ensure that
the transformers are on the tap position to result in
minimal voltage change when switching a transformer
into service. Numeric communication between AVC
relays allows the operator to prepare a transformer for
being switched out of service again resulting in
minimal voltage change.
If an operator had to switch out a transformer manually,
with minimal voltage change, he would first put that
transformer onto the tap position where it will cause
minimal effect when switched out transformer current
at unity power factor. This way there is little voltage
drop across the transformer, so switching it out will
cause little change to busbar voltage. Of course, it is
necessary to tap the other transformer to keep the
voltage at the correct level. During this time, there will
be circulating current flowing, but this is acceptable for
a limited period until the transformer is switched out.
Implementation of this procedure automatically,
consists of the following steps:
1. On the transformer to be switched out, set the
target power factor to unity.
2. Instruct the other transformers in the busbar
group to relax their coupling control.
3. When all voltages balance again, inform the
operator that the transformer is ready to be
switched out.
4. Switch out the transformer.
5. Return the other transformers to their normal
operating mode.
FIGURE 8 (the dotted line) shows the result of a
simulation for a two-transformer substation (the worst
case) with 30MVA 20% transformers when the total
load on the substation is 30 MVA at power factor 0.97.
This is compared to a normal switching operation,
represented by the solid line.
10.8
10.9
11.0
11.1
11.2

FIGURE 8: Comparison of switching methods for transformers
Operator issues Prepare
for Switch-out
Operator opens CB
*1.25% band
AVC responds
after normal delay
Ready for Switch-
out. Operator
opens CB
AVC responds
quickly to
correct voltage
Trending and Prediction
Numeric relays are able to store and analyse vast
amounts of data. Use of historical data enables the
operators to make predictions of the likely system
voltage and load at a given time on a given day. This
can be used to provide a bias to the voltage control
algorithm with the intention of reducing unnecessary
tapping operations if the relay predicts that the load is
about to increase there is little point in tapping down.
The algorithm used for this prediction might range from
a simple What is the load likely to be in 10 minutes,
based on yesterdays levels? to an adaptive technique
that makes it decision based upon the predicted load and
predicted incoming voltage for the time of day, day of
week and season.
The advantages gained from applying these techniques
are both to the customer better regulation; and the
network operator less tap-changes mean less
maintenance.

WAVEFORM QUALITY
There is now increasing interest in the quality of the
waveform, not just in its RMS value.
A number of bodies have a stake in the power system,
and they tend to have differing requirements:
1. users of power require the voltage waveform to
be a sinewave,
2. suppliers of power require the current
waveform to be a sinewave,
3. suppliers of power and the network operator
prefer the power factor to be close to unity, and,
4. the network operator prefers a balanced three-
phase voltage and current.
The second requirement is a direct result of the first a
current waveform of low quality makes it difficult to
produce a voltage waveform that is a sinewave. The
third and fourth requirements are in the interests of
system efficiency.
With numeric relays, it is possible to record various
measurements of waveform quality. Simple measure-
ments that can provide indication of how the network is
performing include form factor, crest factor and the
ratio of negative phase sequence to positive phase
sequence. The AVC relays throughout the network
provide ample opportunities to record this information.
This section of the paper describes the work currently
being undertaken to turn these easily obtainable
measurands into a tool to help to diagnose the source of
waveform quality problems.
Form factor and crest factor are two measurements that
vary with the form of a wave. Form factor is the RMS
value of the waveform divided by the mean absolute
deviation (MAD) that is the mean of the rectified
waveform. Crest factor is the peak value of the
waveform divided by the RMS value.
For a pure sinewave the form factor is 1.111 and the
crest factor is 1.414. As harmonics are superimposed,
the values will vary. Although it is not possible to
completely identify the nature of the harmonics from
1.25
1.30
1.35
1.40
1.45
1.50
1.55
1.05 1.10 1.15
Form factor
3rd @ 0 degs
3rd @ 90 degs
3rd @ 180 degs
5th @ 0 degs
5th @ 90 degs
5th @ 180 degs
7th @ 0 degs
7th @180 degs
7th @ 90 degs

FIGURE 9: Form factor vs. crest factor for different harmonic content
these two figures, they can give some indication of the
type of harmonic.
FIGURE 9 compares the plots of form factor versus
crest factor for 3
rd
5
th
and 7
th
harmonic as their
magnitude increases from zero to 10% of fundamental.
Points to note from this plot are:
! All harmonics at 90) to fundamental will exhibit
unchanged form factor. The crest factor will
increase as the harmonic content increases.
! The plots of 5
th
at 180) and 7
th
at 0) show knee
points. These coincide with the peaks of the
waveforms being reversed, e.g. as shown in
FIGURE 10.
! Where the peak of a harmonic coincides with the
peak of the fundamental (e.g. 3
rd
at 180) or 5
th
at
0)) the crest factor will be higher than that for a
sinewave; and vice-versa.

FIGURE 10: Composite waveform comprising
fundamental plus 10% 5th harmonic at 180)
Given analysis of the information, steps can be taken to
improve the quality devices are now available that can
actively improve waveform quality, e.g. dynamic
voltage restorers (DVRs).

TRANSFORMER MANAGEMENT
It was once the case that power utilities were extremely
reluctant to mix control and protection. That philosophy
is now being reconsidered with the arrival of feeder
managers devices that handle the protection and
control of feeders together with providing SCADA
access. A similar trend can be expected with
transformers as power utilities become more confident
with the operation of numeric relays and wish to make
greater use of the benefits this affords.
With transformer managers there are 3 elements which
can be included in the device:
! transformer protection differential, overcurrent,
REF, Bucholz etc.
! automatic voltage control
! transformer condition management
Knowledge of the tap-position by the protection enables
tighter settings on the differential protection. However,
the greatest benefits come in the integration of
automatic voltage control and the control of
transformers pumps and cooling devices. The load on
the transformer can be used to predict temperature for
the control of cooling devices. The load and temperature
can be optimised for most efficient operation.
The control algorithm can be expanded to include this
extra functionality, which is more easily represented in
matrix form, figure (7). The values of the constants will
be dependent on the transformer and, in fact, might not
be constant, or indeed, linear.
# $
+
+
,
-
.
.
/
0 & & &
'
+
+
,
-
.
.
/
0
1 0 0 0
0 0 .
,
,
freqheat circheat loadheat
freq tr net tr LDC v
n dev
n dev
k k k
k Z k Z k k
T
V

# $
# $
# $
+
+
+
+
+
+
+
+
+
+
,
-
.
.
.
.
.
.
.
.
.
.
/
0
&
&
&
!
&
&
1 ,
1 ,
arg
n dev
n dev
sys
res
reqcirc circ
load
et t VT
T
V
f f
I
I I
I
V V
(7)

FUTURE DEVELOPMENT OF VOLTAGE
CONTROL
The means of implementing voltage control may well
change in the future. Currently OLTCs are used
extensively throughout the network and these are
mechanical devices, the design of which has changed
little in the past fifty years. Solid state devices will have
a future in voltage control and it is possible that the
DVR and OLTC functions will merge. From the point
of view of the control system, this provides new
opportunities, as the means of applying control will
probably be more flexible, with a response time of less
than one cycle (rather than 1 second as at the moment).
It will be possible to actively improve waveform
quality.
For these developments to succeed three conditions
must be fulfilled:
1. the technology must be in place to enable power
quality control,
2. there must be a willingness from power utilities
to embrace these new technologies, and
3. the AVC system must have the ability to make
full use of the control device.
We, as developers of AVC systems, are now preparing
for this vision of the future, using our knowledge and
experience of the fundamental requirements of power
systems.