See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224368637

Fundamentals of adaptive protection of large capacitor banks

Conference Paper · April 2007

DOI: 10.1109/PSAMP.2007.4740911 · Source: IEEE Xplore

CITATIONS READS

13 893

3 authors, including:

Bogdan Kasztenny
Schweitzer Engineering Laboratories, Inc.
281 PUBLICATIONS   2,787 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Travelling Wave Line Protection and Fault Locating View project

Generator Protection View project

All content following this page was uploaded by Bogdan Kasztenny on 11 February 2015.

The user has requested enhancement of the downloaded file.

 Fundamentals of Adaptive Protection of
Large Capacitor Banks
Bogdan Kasztenny Joe Schaefer Ed Clark
GE Multilin Florida Power & Light Company
1. Introduction
Shunt Capacitor Banks (SCB) are installed to provide capacitive
reactive compensation and power factor correction. The use of
SCBs has increased because they are relatively inexpensive, easy
and quick to install, and can be deployed virtually anywhere in
the grid. SCB installations have other beneficial effects on the
system such as improvement of the voltage profile, better voltage
regulation (if they were adequately designed), reduction of losses
and reduction or postponement of investments in the transmission
and generation capacity.
The role of SCBs increased recently in the light of blackout
prevention activities, and increasing penetration of distributed
generation, wind farms in particular, which add generation without
addressing the problem of reactive power support. Moreover,
capacitor banks are valuable assets that must be available for
the daily demands of system operation and must provide reliable
operation through abnormal power system scenarios.
From the protective relaying perspective, however, capacitor banks
are historically considered a relatively low-volume market, and
therefore, did not encourage development of advanced protective
relays dedicated to capacitor banks. Quite often protection
for SCBs is provided by general-purpose multi-function relays,
with only a very few products on the market offering protection
functions specifically tailored to capacitor bank protection.
The utility relay engineer has generally needed to combine the
functionality of multiple relays and customize their programming
to provide the necessary protective system that will avoid false
tripping for system disturbances and obtain the sensitivity for
detecting capacitor CAN faults and minimizing damage.
The SCBs are assembled out of individual cans that are highly
repairable. The need for advanced protection functions is
particularly visible when SCBs are operated under conditions
where one or more capacitor cans are temporarily removed but
the bank is returned to service pending completion of repairs.
However, continuous operation and repairs if needed can be done
only if the bank is protected by a reliable and sensitive relay. This in
turn, can be accomplished by deploying protection principles that
are developed assuming an inherent unbalance in the protected
bank.
Fundamentals of Adaptive Protection of Large Capacitor Banks
Florida Power & Light Company
Presently, in many custom applications or even dedicated
capacitor bank protection products, compensation for inherent
unbalance is based on subtracting historical values from the
operating quantities, and thus making the relay respond to
incremental, “delta” signals.
This paper will show that such simplified approaches are not
optimal. Instead this paper derives technically accurate operating
equations for capacitor bank protection that are derived assuming
both inherent capacitor bank and system unbalance.
It is important that the relay is capable of dynamically compensating
for unbalances between the power system phase voltages. These
differences are constantly changing and may be on the order of
2 percent or more under normal conditions, and tens of percent
during major system events such as close-in faults. The presented
protection methods allow compensating simultaneously for the
bank inherent unbalance and system unbalance increasing both
sensitivity and security of protection.
The presented methods also facilitate auto-setting and self-
tuning applications. Auto-setting is an operation of calculating
new accurate relay constants to account for the inherent bank
unbalances following the bank repair, and is performed in response
to the user’s request and under user supervision. Self-tuning is an
operation of constantly self-adjusting the balancing constants in
order to maintain optimum sensitivity of protection when the bank
reactances change slowly in response to seasonal temperature
variations and other conditions. The self-tuning applications
require monitoring the total changes in the balancing constants
in order to detect slow failure modes, and account for a series of
small changes that do not trigger alarms on their own.
2. Capacitors
Protection engineering for shunt capacitor banks requires
knowledge of the capabilities and limitations of the capacitor unit
and associated electrical equipment including individual capacitor
unit, bank switching devices, fuses, location and type of voltage
and current instrument transformers.
19
 A capacitor unit, Figure 1, is the building block of any SCB. The
capacitor unit is made up of individual capacitor elements,
arranged in parallel/series connected groups, within a steel
enclosure. The internal discharge device is a resistor that reduces
the unit residual voltage allowing switching the banks back after
removing it from service. Capacitor units are available in a variety
of voltage ratings (240V to 25kV) and sizes (2.5kVAr to about
1000kVAr).
The capacitor unit protection is based on the capacitor element
failing in a shorted mode. A failure in the capacitor element
dielectric causes the foils to weld together and short circuits the
other capacitor elements connected in parallel in the same group,
refer to Figure 1. The remaining series capacitor elements in the
unit remain in service with a higher voltage across each of them
and an increased capacitor can current. If a second element fails
the process repeats itself resulting in an even higher voltage for
the remaining elements.
There are generally four types of the capacitor unit designs to
consider.
2.1 Externally Fused Capacitors
An individual fuse, externally mounted between the capacitor unit
and the capacitor bank fuse bus, protects each capacitor unit.
The capacitor unit can be designed for a relatively high voltage
because the external fuse is capable of interrupting a high-voltage
fault. However, the kilovar rating of the individual capacitor unit is
usually smaller because a minimum number of parallel units are
required to allow the bank to remain in service with a capacitor
can out of service. A SCB using fused capacitors is configured
using one or more series groups of parallel-connected capacitor
units per phase, as shown in Figure 2.
2.2 Internally Fused Capacitors
Each capacitor element is fused inside the capacitor unit. A
“simplified” fuse is a piece of wire sized to melt under the fault
current, and encapsulated in a wrapper able to withstand the
heat produced by the arc during the current interruption. Upon the
capacitor failure, the fuse removes the affected element only. The
other elements, connected in parallel in the same group, remain in
service but with a slightly higher voltage across them.
Figure 3 illustrates a typical capacitor bank utilizing internally
fused capacitor units. In general, banks employing internally
Figure 1.
Capacitor unit.
20 Fundamentals of Adaptive Protection of Large Capacitor Banks
fused capacitor units are configured with fewer capacitor units
in parallel, and more series groups of units than are used in banks
employing externally fused capacitor units. The capacitor units
are built larger because the entire unit is not expected to fail.
2.3 Fuseless Capacitors
Fuseless Capacitor Bank designs are typically the most prevalent
designs in modern day. The capacitor units for fuseless capacitor
banks are connected in series strings between phase and neutral,
as shown in Figure 4. The higher the voltage for the bank, the more
capacitor elements in series.
The expected failure of the capacitor unit element is a short
circuit, where the remaining capacitor elements will absorb the
additional voltage. For example, if there are 6 capacitor units in
series and each unit has 8 element groups in series there is a
total of 48 element groups in the string. If one capacitor element
fails, this element is shorted and the voltage across the remaining
elements is 48/47 of the previous value, or about 2% higher. The
capacitor bank remains in service; however, successive failures of
elements would aggravate the problem and eventually lead to the
removal of the bank.
The fuseless design is usually applied for applications at or above
34.5kV where each string has more than 10 elements in series to
ensure the remaining elements do not exceed 110% rating if an
element in the string shorts.
2.4 Unfused Capacitors
Contrary to the fuseless configuration, where the units are
connected in series, the unfused shunt capacitor bank uses a
series/parallel connection of the capacitor units. The unfused
approach would normally be used on banks below 34.5kV, where
series strings of capacitor units are not practical, or on higher
voltage banks with modest parallel energy. This design does not
require as many capacitor units in parallel as an externally fused
bank.
3. Configurations of Shunt Capacitor
Banks
Protection of shunt capacitor banks requires an understanding
of the basics of capacitor bank design and capacitor unit
Figure 2.
Externally fused shunt capacitor bank and capacitor unit.
connections. As a general rule, the minimum number of units
connected in parallel is such that isolation of one capacitor unit in
a group should not cause a voltage unbalance sufficient to place
more than 110% of rated voltage on the remaining capacitors
of the group. Equally, the minimum number of series connected
groups is that in which the complete bypass of the group does not
subject the other capacitors remaining in service to a permanent
overvoltage of more than 110%. The value of 110% is the
maximum continuous overvoltage capability of capacitor units as
per IEEE Std 18-1992.
The maximum number of capacitor units that may be placed in
parallel per group is governed by a different consideration. When
a capacitor bank unit fails, other capacitors in the same parallel
group contain some amount of charge. This charge will drain off
as a high frequency transient current that flows through the failed
capacitor unit. The capacitor can fuse holder, when used, and the
failed capacitor unit must withstand this discharge transient.
The discharge transient from a large number of paralleled
capacitors can be severe enough to rupture the failed capacitor
unit or explode a fuse holder, which may damage adjacent units
and even cause a major bus fault within the bank. To minimize the
probability of failure of the explosion of the fuse holder, or rupture
of the capacitor case, or both, the standards impose a limit to the
total maximum energy stored in a parallel-connected group to
4650 kVAr. In order not to violate this limit, more capacitor groups
of a lower voltage rating connected in series (with fewer units
in parallel per group) may be a suitable solution. However, this
may reduce sensitivity of applied unbalance detection schemes.
Splitting the bank into two sections as a double wye may be the
preferred solution, and may allow for better unbalance detection
scheme.
Two prevalent designs of SCBs are the externally fused bank and
the fuseless bank. There are advantages to each design.
Externally fused banks typically have a higher unbalance current
when a unit fails which is used to operate a fused disconnect
device. This design typically results in a simpler bank configuration
and provides an easy method for field identification of a failed
unit. A fused design also requires less sensitive unbalance
protection since the fuse is the principal method used for isolating
a can failure. However, this style of bank has a higher initial cost
and usually higher maintenance costs. Since the fused element
is exposed to the environment, the fuses become less reliable
Figure 3.
Internally fused shunt capacitor bank and capacitor unit.
Fundamentals of Adaptive Protection of Large Capacitor Banks
and require more maintenance to ensure correct operation. As a
result, fuseless capacitor banks have become increasingly popular.
Elimination of the fused connection results in a lower initial cost,
reduced maintenance costs, smaller bank footprint, and fewer
losses. Also, this bank design typically makes catastrophic can
rupture less likely since the discharge energy of a failed element
will be small.
However, the fuseless bank design has two main disadvantages
that increase the emphasis on requiring sensitive relaying
protection. One, the elimination of the external fuse means
that visual indication of the failed capacitor has been lost. In
addition, an element failure results in an overvoltage condition
of the remaining elements, stressing them. Without a fuse as a
means of isolating the failed can, the protective relay must now
be sensitive enough to detect a failed element and alarm before
additional elements fail causing a higher overvoltage condition on
the remaining units. Because of these two factors, it is especially
important to utilize a sensitive protective relay which can correctly
isolate a bank for a failed element. Also, the use of faulted phase
identification assists field personnel in locating a failed capacitor
can without having to test the entire bank.
The optimum connection for a SCB depends on the best utilization
of the available voltage ratings of capacitor units, fusing, and
protective relaying. Virtually all HV and EHV banks are connected
in one of the two wye configurations listed below [1,2]. Distribution
capacitor banks, however, may be connected in wye or delta.
Some banks may use an H configuration on each of the phases
with a current transformer in the connecting branch to detect the
unbalance.
3.1 Grounded Wye-Connected Banks
Grounded wye capacitor banks are composed of series and
parallel-connected capacitor units per phase and provide a low
impedance path to ground. This offers some protection from surge
overvoltages and transient overcurrent conditions.
When a capacitor bank becomes too large, making the parallel
energy of a series group too high for the capacitor units or fuses
(above 4650kVAr), the bank may be split into two wye sections.
The characteristics of the grounded double wye are similar to a
grounded single wye bank. The two neutrals should be directly
connected with a single path to ground.
Figure 4.
Fuseless shunt capacitor bank and series string.
21
 The double wye design facilitates better protection methods. Even 4. Sensitive Capacitor Bank Protection
with inherent unbalances the two banks will respond similarly to
system events, and therefore, methods based on comparing one Methods
split-phase versus the other are more sensitive and less prone to
system events (phase current balance technique, for example). 4.1 Voltage Differential (87V)
With reference to Figure 5, this function is based on a voltage
3.2 Ungrounded Wye-Connected Banks
divider principle – a healthy capacitor string has a constant and
Ungrounded wye banks do not permit zero sequence currents, known division ratio between its full tap (typically the bus voltage)
third harmonic currents, or large capacitor discharge currents and an auxiliary tap used by the protection. The principle could
during system ground faults (phase-to-phase faults may still occur be used on both grounded (Figure 5a) and ungrounded (Figure
and will result in large discharge currents). Another advantage is 5b) banks. In the latter case the neutral point voltage (VX) must
that overvoltages appearing at the CT secondaries are not as high be measured by the relay, and used to derive the voltage across
as in the case of grounded banks. However, the neutral should be the string.
insulated for full line voltage because it is momentarily at phase
The function uses the following operating signal:
potential when the bank is switched or when one capacitor unit
fails in a bank configured with a single group of units.
V11 AA 
 kk AA ˜˜ V
V V for grounded banks (1a)
VOP ( A)
OP ( A ) V22 AA
3.3 Delta-Connected Banks
Delta-connected banks are generally used only at distribution V
VOP ( A)
OP ( A ) V11 AA 
V  kk AA ˜˜ V
V22 AA 
VVXX ˜˜ kk AA 
 11
voltages and are configured with a single series group of capacitors
rated at line-to-line voltage. With only one series group of units no for ungrounded banks (1b)
overvoltage occurs across the remaining capacitor units from the n 2
isolation of a faulted capacitor unit. V
VOP ( A ) V
V 1A 
 kk A ˜˜ nVT
VT 2 ˜ V2 A
˜ V2 A
OP ( A )k is a1division
Where A
A A ratio
nnVTfor
1 the A-phase of the bank.
VT 1
3.4 H-Configuration Identical relations apply to phases B and C.
Some larger banks use an H configuration in each phase with a
current transformer connected between the two legs to compare
Note that equations (1) n
V
or magnitudes. V 

1 A During
can
VT 2be implemented
kk A ˜˜ nno-fault
˜  ˜
conditions
nVTX
and 
using either phasors
VT 2 ˜ V2 A  V X ˜ nVTX ˜ k A  1
˜ k A small
 1 bank 
VOP V V V nnVT 1under
( A)
OP ( A ) 1A A n 2A X
the current down each leg. As long as all capacitors are balanced, unbalances caused by ninternal VT 1
VT 1 bank failures,
VT 1the two voltages
no current will flow through the current transformer. If a capacitor will be almost in phase, suggesting the phasors and magnitude
fuse operates, some current will flow through the current versions would yield similar results. However, the function is set
transformer. This bridge connection facilitates very sensitive nnVT 2 and given possible angular errors of the used VTs,
kk A ˜˜sensitive
very VT 2 | 1
|1
protection. The H arrangement is used on large banks with many there
A will nnVTbe
1 differences in performance between the two possible
capacitor units in parallel. versions. The performance depends on the type of security
VT 1
measures used to deal with errors of instrument transformers.
More V
ˆ information
V1 A is provided in one of the following sections.
kkˆA 1A
A V
Typically,
V22theA method is used on grounded banks and equation (1a)
A
is used. In theory, the algorithm could be applied on ungrounded
banks using equation (1b), but it requires both the neutral voltage
VAA 
V
and 
theV VBB 
V
VXXtapvoltagesVVXXtobeV C  VX Such arrangements may
C  VX
Vmeasured. 00
not Z 
be practical (the
Z 
tap voltages Z not measured on ungrounded
Z A If the tap voltages
banks).
A ZB B ZCC
are measured, one could apply multiple
overlapping protection zones to the ungrounded bank as long as
§§ 11 relay(s)
the applied 11 ··theV
11 support required
V number
V of inputs and
 V ˜
associated ¨
 VX ˜ ¨voltage 
protection
 differential;
 ¸
functions.  V  V
Specifically,
A
¸  two neutral
A
B
B  V
 voltage00(1b)
equation
C
C can be
usedXfor¨ ©© Z
Z AA Z Z BB Z Z CC ¸¹¹ andZ
Z AA Z Z BB Z Z CC unbalance
protection elements can be used – one balancing the bus voltages
with the neutral voltage, and another balancing the tap voltages
against the § 1neutral1voltage. 1 · V
§¨ 1  1  1 ·¸  VAA  VBB  VCC  VBB  VBB  V
V V V V
VCC  V

 V
V X ˜˜ ¨   ¸      
VC
 ZC 00
Figure 5.
X ¨ Z(1) apply
Equations to ¸
primary voltages, and as such
©© Z AA Z BB Z CC ¹¹ Z AA Z AA Z AA Z BB Z AA Z CC Z AA
Z Z Z Z Z Z they
Z Z
Voltage differential application to grounded (a) incorporate the voltage-dividing ratio of the capacitor, but ignore
and ungrounded (b) banks. the ratios of applied instrument transformers. In secondary
voltages,§ the
1 operating
§ 1 1 · is:1
1 voltage ·¸  1 ˜ V  V  V   V ˜ §¨ 1  11 ··¸  V ˜
§ 1
 VXX ˜˜ ¨¨¨ Z  Z  VAAgrounded C   VB ˜ ¨
1 1
V Z ¸¸  Z ˜ for  VBB  Vbanks ¨© Z  Z ¸¸  VCC ˜
A1 A ZkBA ˜ VZ
B (1c)
( A ) © ZV ¹¹ Z AA © B Z AA ¹¹
C
V A1 A  kBA ˜ V22 A
OP ( A ) © V
VOP C
AC
Z B

V (§
§A) V
¨A11)  Z11AAAA 
Z ZAAAA ˜˜··¸V
 kkZ V22 AA3˜ VV kk AA˜ §§¨111 ZZ AA ··¸  V ˜ §§¨1  ZZ AA ··¸ 0
V0XX˜˜ V
X ˜˜( ¨
VOP V
VOP   ¸  3 ˜ V  V B ˜ ¨1  ¸  VCC ˜ ¨¨1  Z ¸¸ 0
B ¸
B ¨
Z CC ¸¹¹
X ¨
©© Z Z BB Z
0
©© Z Z ¹
B ¹ banks © (1d)
© Z CC ¹¹
for ungrounded
n 2
V
VOP ( A) Z V V 1A  
X kk A ˜˜ nVT VT 2 ˜ V2 A
˜ VZ
kk OP |
( A) Z A 1 A X A A n
, k Z2 AA | X XA
AB
AB Z
A
| X
A
, kn VT 1
AC
VT
AC 1
Z | XA
Z BB Capacitor
X BB Banks Z CC X CC
22 Fundamentals of Adaptive Protection of Large

VT 2 ˜ V2 A  V X ˜ nVTX ˜ k A  1
n 2 nVTX
1A  kk A ˜˜ nVT
 ˜V2xA 3 ˜VX 0 nnVTVVBB1 ˜˜˜11kA kkAB 1  V ˜ 1  k 
V V  ˜  ˜
V OP ( A ) 1
11A k ABA nk AC
V V V
3 1  k AB  nkVT AC1  ˜ V x  3 ˜ V0  AB   VC ˜ 1  k AC 
V 1
VOP OP
OP
( A)
VT
1
VT 1
C AC
 ˜ V  VX ˜ VTX ˜ k A  1
nVT 2 n
V V k ˜ VnX ˜ k A 21A
V1 A 1 Ak A ˜ V2AA  n
V1 A  k A ˜ V2 A  VX ˜ k A  1
OP ( A )
VOP nnVT 1
˜ k A  1
( A)
VOP ( A) VOP ( A) V1 A  k A ˜ VT21
VT
˜ V2 A  VX ˜ VTX
nVT 1 nVT 1
nVT 2
VOP ( A)n V1 A  k A ˜ ˜ V2 A
A˜ V1 A|1k A ˜ V2 A Bank Neutral Voltage
VT 2 nVT 1
Where V1 A  k Asignal
VOP ( A)the operating
n
˜ VT 2is ˜inVsecondary volts of the bus VT, and VkOP
4.2 ( ACompensated
n)nVT
kA ˜ |1
2A VT21
the nVT1, nVT2 and nVTX stand
nVT 1for ratios of the bus, tap, and neutral Unbalance  k ˜ Vk (59NU)
X ˜ 1 ˜ k A  1
voltage transformers, respectively. n nVT 2 nVTX
V11AA  kA A˜ ˜nV2 A ˜VV X k A 
2 A ˜ V
VV V
VT V1
OP ( A ) 1A A 2A
VOP OP ( A )
( A) nVT 1 is applicable to ungrounded
With reference V1 A to Figure VT 1 6 this function
Normally the VT ratios are selected so that the secondary Vkˆ A V  k ˜ V  V ˜ k  1
banks, and is based on the Kirchhoff’s currents law for the neutral
˜ under 
voltages nVTtap
2 voltages are n
OP ( A ) 1A A 2A X A

VOP ( A) forVthe bus and

1A  k A ˜ ˜ V2 A  VX ˜ similar
VTX
k A  1nominal ˆ nof
node
V12AA
VT 2the bank: nVT 2
system voltage. This leads nVTto1 the effective matchingnVT 1 factor for the VVkkOP A ˜( A )
A
nVTV V1 2 
|11A nk A ˜
V ˜ V2 A
A k ˜ n ˜ VnVT 1
VT 2
secondary voltages being close to unity: OP ( A ) 1A A 2A

VA  VX VB  VX VC  VX VT 1

(3a)
nVT 2 V1 A  n  0
kA ˜ |1 V ˆ Z  V V Z 
V AA AV X  k ˜ B ˜ VBnVTX2V ˜ C ˜ kC X1n V V Z  V
˜ V2nA  VX ˜  VTX0˜ k A  1
k n
) 2 A V1
(1e)
VT 2 VTX
VOP ( AV A nk A ˜
OP ( A ) 1A A 2A X A

nVT 1 ZA n
Z B VT 1
VT 1
Z C nVT 1
VT 1

Thenabove expression can be rearranged as follows:

kV A˜  V X|§1 1 VB  VX1 VC 1VX· V
VT 2
V V
Voltage-based capacitor protection functions are set sensitive. 
 nn ˜ ¨
ZB  Z C ¸¸ 0 A  B  C 0
A

V VZ ¨ VT 1
Given
ˆ the 1 Aformat of equations (1) both the bus and tap voltages ˜ X
AVT 2
§ Z1 Z1 Z1 · VZ
|
 VVXnVT˜ ¨¨1© A  B  C ¸¸¹  AA 
kA k 1 VZBB VZCC

A
shall be
V2measured accurately in order to gain sensitivity of 0 (3b)
kˆ 1A

¸¸  A ¹ B  C 0
V § ©1 A 1
A
protection. As a result the VT ratios shall be selected so that the
A
Z Z B1 · V Z C Z
V AV B Z Z C
 VX ˜ ¨¨  
2A

resultant secondary voltages fall in the region of maximum relay V § ·

VkˆA Vfurther ©1 A A 1V B V 1CV ¹ 1 A
Z Z Z Z Z BV Z
A C VB V V V V V
VA  VX andVBthe Vtwo
accuracy, VTs
VCwork  VXwithin their maximum class and ZVXV˜ ¨¨V Zto 
an equivalent  0 ¸¸form  of:   C  B  B  C  C 0
 
A X B X C X

§ Z1 Z1 · VZ
X
0
accuracy under nominal system voltage. The latter is ensured for
 VVX˜ §¨˜ ¨¨©1 A 1  B1 ·¸  VCA ¸¸¹VB AA VC  BVA B VCBA VC BB VC BA0 CC  CA 0
2 A A Z1 Z B CVZ VZ VZ VZ VZ VZ
the ZbusA voltage; Z selection
B of the Z C VT for the tap voltage shall be
¨ Z Z1VA Z1 V Z· BV ¸ZVZCV¹V ZZ0 A Z Z AZ ZZA ZZ B Z Z A Z C Z A
˜ ¨¨V©X©
X§ 1
done carefully to minimize VT and relay errors for the tap voltage. VVA  A  B B ¸¸ XZ C ¹V
 ¹ Z Z ·Z
A B C
C A X A
© §
X
0 A B A C A
Relay setting range for the ratio-matching factor is another § 1 1 ·
Z Z Z
§ that 1 · VofA thisVVT 1 1 1 1
 
A B C A B C
Z Z Z
condition
 VX ˜ ¨¨
1 may1 limit selection
¸¸ 
VC
B ratio.  V A
˜ ¨  B
 C
¸  ˜ V  V  V  V ˜ ¨
(3c) 
§ 1 Z1 ·¸·
¸  VC ˜
    0 §˜ §¨1 §¨1 Z ·1 1V ·¸ Z 1 ·¸V Z1 V ˜ §¨ 1  1 ·¸  V§¨ Z
X A B C B
 V X˜ ¨¨ ¨ ¨© A
1 1 1  Z B ¸ C ¹ A A 
 A¸¸Z VZ B˜ ZV   ©C B A ¹
© ZVBC  
 1 V1 ˜ VV V  V  ˜1¨  1 ¸ 0
¸¸ Z Z AV¸¹B ˜ ¨¨ ¨© Z C Z A ¸¸¸¹  VC ˜
V V V V
 VX© Z©§is˜ ¨Zidentical
1AZ Z1ZB ¹with: VCA Z VBZ 
Z B Z C ¹ apply C B ¨ 0
© Z A characteristics Z A toZ Bthe voltage
ZC A B C B B C C
The following differential which X
C1 ¹ · ZZV
Z
 VX ˜ ¨¨ © Z A ZB ¸¸ZC ¹  Z A  © ZB Z A ¹
A B C A A A B A C A
A B
function [3]: 0
© Z Z Z ¹ Z Z Z
0A ·
1 · 1· § 1 §1 ·
V ˜§¨¨ § Z AZAZ A ¸¸·Z ˜ V  V §V  ZV ·˜ ¨§  §Z A¸¸ ·Z
§ 1 A
1 B C A B § 1
C
1 ·
V A ˜·¨¨ ¸¸ Z
• The§element 1 · V VBto VCV  
VXX ˜ ¨¨˜V 1¨ZC1 Z  Z¹¸¸ ZA3 ˜¸V03V˜BV˜ ¨¨01 VBA ¸¸˜¨©¨Z1V
¸Z C ¹ §¨ Z ·¸¸ 0
˜Z¨¨1 ¹ ¸  ZC 0˜ ¨Z1 
1 1shall support individualVB per-phase
VC Vsettings ¸¸V
¨ ¸¸  Z B ¹§¨
X A B C B C

 Vcope
X ˜ ¨ withdifferent  unbalances A

between  the phases
 B

(repairs   ©§¨ Z BZ0 Z C ¹Z ·¸
© A B C
©
A C
© ·
© ¹ B A C A

Z¨˜©§¨1A 1 AB 1  AC1¸¹·¸3V˜ V VB V©1 Z B ¹ VB V©  VC AC ¸¹ 0

and©shorted
Z A Zunits).B Z C ¹ Z A Z A Z A Z B Z A Z CV V X§X˜ ¨
A C  VBA ¸ 
¸ ¸˜ §¨1 ZZ 0·¸VZV˜ §¨B1 ˜ Z¨¨
Z · ¸ VC ˜ ¨¨C1  (3d) ¸0
V ˜ ¨¨1 ©¨
Z Z ·
© Z  Z ¸Z  ˜ Z C C¹ ¹Z¨

Z A ¨ ©Z A ¸
¸ Z B ¹ A Z Z©C ZZAC ¹
A A A A
X B¸ B 3 V V AX¸ 0 B 0 C
Z AZX
©A | ¹A © ¹
B
k AB© ¹A , k Z
• The element is capable of indicating the affected phase, and Z|Aboth
A B C B C

Multiplying Xsides AC
byZ CandXZ X A the sum of the phase
substituting
§ 1 the1 number
potentially 1 ·of faulted 1 capacitor elements, to aid k 1 · by
§ 1 voltages Z X|§V 1Z , kXAC A
1 · A
|
˜ Vbank.
A  VB  VC   VB ˜ ¨
B B C
¨
 Vtroubleshooting
X ˜¨   repairs ¸¸  of the ¨ k AB Z¸¸§¨  |Z 1XV 3, ˜k•¨1X 1| · ¸¸1 Z0AC § 1 1 · § 1 1 ·
and ¨   VXB AC VC   VB ˜ ¨¨
A A A A
0
 V ˜ AB BC
A  | 
BZ
A ¸  ˜
AC
V A|  ¸¸  VC ˜ ¨¨  ¸
¸ Z C Z A ¸¹
X ¨
© A C ¹ © B yields:
k ABA ¹1 Z©X , k ¹3A˜VZ  V ˜X1  k   V ˜ ©1 Z kB Z A ¹
Z Z Z Z Z Z Z X B Z X
ZB C C
B A
• The function shall apply appropriate security measures for
VOP © Z Z 1AB k AB B k Z C ˜ ¹
C
B AC
V
ACA Z
x  0C B C AB C AC
©
13
 ˜ 1 1 k  k  ˜V  3 ˜V  V ˜ 1  k   V ˜ 1  k 
OP§ V Z V 1 Zk AB·  k AC  ˜Vx § 3Z˜V0·  VB ˜§1 Zk AB·   VC ˜ 1  k AC 
VOPsensitive
( A) V but
1A k
secure
A V operation:
2A appropriate restraint signal V
3
OP AB AC x 0 B AB C AC

§ be
could Z A · to accompany§ the operating
Z Adeveloped ZA · § (1).Z A · V
signal
¨
VX ˜Setting   ¸  ˜  ˜ ¨  ¸  ˜ ¨  ¸ VOP ˜ ¨ 1 1 X3 A 
 ¸¸  3 ˜ V0  VB ˜ ¨¨1  A ¸¸  VC ˜ ¨¨1  A ¸¸ 0
¹  k AC  ˜ Vx © 3 ˜ZVB0¹ VB ˜©1 ZkCAB¹   VC ˜ 1  k AC 
A
1 3 V V 1 V 1 V 0 ¨
¨ Zrange Zshall¸ allow disabling the ¨ restraint ¸ ¨ so. ¸ V
0
0 B if desired
C V X V V 1  k
VOP (©A) VB1 A  kCA ¹˜ V2 A  VX ˜ k A © 1 Z B ¹ © Z C ¹ OP© 3 B
OP
Z ZX
CAB
0

• Several independent pickup thresholds shall be provided for VOP VX  V0 (3e)

alarming and tripping. ZA X A ZA X A
ZA X A ZA X A k
VAB OP ZV X XV0
| , k |
| matching | (k) shall be individually
AC
k , knACVT 2coefficients ZC X C
• VAB The ZvoltageV 
X k ˜ ˜ V
Z X
B B
OP ( A ) B 1 A B A 2CA
setable per phase. n C
VT 1

• Both auto-setting and self-tuning applications of this method

VOP
1
1  k AB  k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
    ˜ 1  k AC 
1 3
1  k ABProvision
VOPare possible.  knAC ˜could
Vx  3be˜Vmade
0 nVB ˜ 1  k AB  VC
to calculate the
VOP ( A) V1 A  k A ˜
matching
3 factors k VT 2
the user, either locally
˜ V2 A  VX ˜
automatically under manual
VTX
˜ supervision
k A  1 of   VOP VX  V0
nVTor1 remotely (auto-setting),
nVT 1 or calculate
the factor constantly in a slow adjusting loop (self-tuning).
VOP VX  V0
nVT 2
k ˜
The process
|1
nVT 1 of finding the constant balancing a given phase of
A
protection is based on the following simple equation:

V1 A
kˆA (under no-fault conditions) (2)
V2 A

The
VA voltage
 VX differential
V  V method V  Vcan be used in a number of
 asB longX as theC relayX allows
configurations 0 wide range of ratio
matching
Z A for the compared
ZB voltages:
Z C tap voltage can be compared
with the bus voltage; two taps can be compared on the same bank;
Figure 6.
two taps can be compared between two parallel banks, etc.
Compensated bank neutral overvoltage application.
§ 1 1 1 · VA VB VC
 VX ˜ ¨¨   ¸¸    0
© Z A Z B Z C ¹ Z Fundamentals
A Z B ZC
of Adaptive Protection of Large Capacitor Banks 23

§ 1 1 1 · VA VB VC VB VB VC VC
 VX ˜ ¨¨   ¸¸        0
 § 1 1 · 1 § 1 1 · § 1 1 ·
˜ VA  VB  VC   VB ˜ ¨¨
1
 VX ˜V¨¨1 A   ¸¸   ¸¸  VC ˜ ¨¨  ¸¸ 0
k VA ˜ §¨ ©1Z A 1 ZB 1 ·¸ZCV¹A  ZVBA  VC  VB  VB  VC  V©C Z B0 Z A ¹
ˆ
© C
Z Z A ¹
X ¨V ¸
© Z2 AA Z B Z C ¹ Z A Z A Z A Z B Z A Z C Z A
§ Z Z · § Z · § Z ·
VVXVA ˜¨¨˜ 1§¨VX1 AV1B  VA1X¸¸·¸3V1˜CV˜0VVVXBV˜ ¨¨1V  AV¸¸ ˜ §¨V1C ˜¨¨11·¸ AV ¸¸ ˜ §¨ 01  1 ·¸ 0
Introducing 
Zthe following 
matchingA k-values 0
Z©A© Z A BZ B Z BZC C¹¸¹ Z A Z C
X ¨ Z B
© CtoZreflect
B
©
the inherent C saturation ¸of the relay input – the function shall be blocked in this
B ¹ ¨ Z B © Z A ¸Z C ¹ ¨ Z C
¹ © ZA ¹
bank unbalance: case under external faults either by time delay or explicit logic in
order to cope with the spurious unbalance caused by saturation
§ ZZ ZX · Z § ZXA · § Z · of the VX measurement. In any case, one shall observe the thermal
k XAB˜ ¨¨1 § A1A |
V 10 ·VB ˜V¨¨A1A|
 A1A¸¸ , 3k˜ VAC  VA¸¸B VCV˜ C¨¨1  A ¸¸ 0 (4) withstand rating of the relay input when selecting relatively low-
 V©X ˜ ¨¨ZZBB ZXC B¹  ¸¸ Z C© ZXB C¹  © Z0C ¹ ratio VT for the measurement of the VX signal.
© ZA ZB ZC ¹ ZA ZB ZC
allows Zre-writing
X the balance
Z A Xequation (3e) into the following When written for secondary voltages the key operating equation
| A , k AC | A
§B 1 B AB 1 C· xVCA V0B BVC VBAB  VB C ˜ V 1  k ACV
k AB 1A signal: becomes:
VOP Z 1X k  k Z ˜ VX  3 ˜ V  V ˜ 1  k
operating
1 AC V
 VX ˜ ¨¨3   ¸¸       C•  When
C
0
measuring the 3 • V0 internally and expressing the
1© A Z Z Z C ¹ Z Z Z Z Z Z Z
operating signal in secondary volts of the bus voltage:
1  k AB  k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
B A A A B A C A

OP 3V X  V0
VOP
V
§ 1 1 · 1 §(5) 1 V 1 ·1 nVTX 1§ k1  k 1 ˜ V·  3 ˜ V  V ˜ 1  k   V ˜ 1  k 
˜ ¨X  V0
VOPV V
1
 ¸  ˜ V
¸not magnitudes, V  V   V ˜ ¨¨  OP ¸¸3 nV ˜ ¨ AB  AC ¸¸x 0 0 B
VTC ¨
Equation¨
AB C AC
X A B C B
(5)
© A
Z involves
Z B
phasors,
Z C ¹ Z A
i.e. the vectorial sum © B
Z Z A ¹ © C
Z Z A ¹
of the voltages is created by the protection function implementing (7a)
the method. VOP
1 nVTX
1  k AB  k AC  ˜ Vx  3 ˜ nVT 0 V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
§ Z ·
Z A the capacitor impedancesZ § · Z § · • When3 nVTmeasuring the 3 • V0 nfrom an open-delta VT and
Note ¨¨1 theAratios
VX ˜that  of  VC ˜ ¨¨phase
¸¸  3 ˜ V0  VB ˜ ¨¨1  A ¸¸between ¸¸
VT
1 A 0 expressing
1 nVTX the operating signal in secondary volts of the bus
1  k AB  k AC  ˜ Vx  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
A and the Z
two other phases are close to unity, and therefore VOP
© B ZC ¹ © ZB ¹ © ZC ¹ Re ^1voltage:
 VT  k AC  ˜ V X  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  k AC `
3 knAB 0
the correcting factors for the B and C-phase voltages are small
numbers, while the coefficient in front of the VX voltage is close
Im ^1 
1 knVTX  k  ˜ V  3 ˜ V  V n˜VT10 k   V ˜ 1  k ` 0
AB 1 ACk
AB  k AC  ˜ Vx  3 ˜  VB ˜ 1C k AB  AC
VC ˜ 1  k AC 
to 3. Z X Z X V X 0 B V0 AB
| |
A A A A OP
k AB , k AC 3 nVT nVT
EquationZ(5) while
X B following Zrelations
X C(4) is a proper neutral Z1 A  Z 2 A
B C (7b)
overvoltage function compensated for both the system unbalance IReDIF^(1
A ) k AB  k AC
VBANK ( A ) ˜ V X  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  k AC ` 0
Z1 A ˜ Z 2 A
(V0), and the bank unbalance (kAB, kAC). To understand it better The following characteristics apply to the compensated bank
Im ^1  kvoltage  ˜V X  3 ˜V0 function
VB ˜ 1  k[3]:
AB   VC ˜ 1  k AC ` 0
assume 1
VOP operating
precise   k ACtakes
equation 
˜Vx afamiliar
3 ˜V0 simplified
VB ˜ 1 form 
the bank is perfectly balanced (kAB = 1, kAC = 1). If so, the
1  k AB k AB [1]: VC  ˜
1
I A V BANK ( A)

k AC AB  kZAC1unbalance
neutral
A  Z 2A
3
• The singleZ 1element ZA ˜ Z 2ZA2 A function does not indicate explicitly the
I DIF ( A) VBANK ( A) 1 A
effected phase. Z1 A ˜ ZIt2 A could, however, aid troubleshooting and
VOP VX  V0 (6) repairs by reporting Z ˜ Z the k-factors Z ˜ (pre-fault
Z and fault values).
VBANK ( A) I DIF ( A) ˜ 1 A 2 A I A ˜ 1 A 2 A
Z 1 A  ZZ12AA  Z 2 A Z1 A  Z 2 A
• I A TheV BANKfunction shall apply appropriate security measures for
Z1A ˜ Z 2 A
( A)
Equation (5) identifies the source of the inherent bank unbalance, sensitive but secure operation: appropriate restraint signal
and therefore allows for proper compensation. In addition, this key Z1 A ˜beZ 2 Aused with Z ˜ Z 2 A operating signal (5). Disabling the
I DIF ( could Z I A˜ Z 1 A the 0 ˜Z
equation allows analyzing the impact of imperfect compensation A)
VBANKrestraintZ1 AI  Zshould
2 A˜ 1 A be  ZI2 A˜ Zif1 Adesired
Z12AAallowed 2A so.
and/or errors of instrument transformers on sensitivity of Z1 A  Z 2 A Z1 A  Z 2 A
( A) DIF ( A ) A

protection as explained later in this paper. • SeveralZindependent pickup thresholds shall be provided for
1A  Z 2 A
A) Z1IAA˜ Z 2 Aand tripping.
I DIF ( alarming Z1 A0˜ Z 2 A
Equation (5) can be implemented using either derived neutral I DIF ( A) Z  ZI A2 A 0
Z1 A  Z12AA Z1 A  Z 2 A
component in the bus voltages (vectorial sum of the phase • The inherent bank unbalance constants (k-values) shall be
voltages calculated by the relay), or directly measured neutral Z1 A  Z 2 A X 1 A  X 2 A
voltage component (open-delta VT voltage). Slightly different kI A settable. Z1 A |Z 2 A
DIF ( A )Z I  A Z X 1A  0X
Z 1A  Z 2 A
1A 2A 2A
errors would occur in these two approaches. • Both 1 nauto-setting
VOP VTX
1  k AB and k ACself-tuning
Vx  3 ˜ Vapplications
 ˜voltage are possible as
0  VB ˜ 1  k AB   VC ˜ 1  k AC 
long as the
3 nZVT  Xk ˜ I Xneutral point is non-zero and is measured
When deriving the 3 • V0 internally the relay is presented with I OP ( A)Z1 1A I
near-nominal voltages under internal failures that require high
k
VOP A
with nadequate
DIF
VTX
1 X 1kAAB X 2kA AC  ˜ Vx  3 ˜ V0  VB ˜ 1  k ABto calculate
2(AA )
| 1AA
Z13A nZ 2kA automatically
accuracy.
A 2A
Provision could be made VC ˜ 1  k AC 
factors under manual supervision of the user,
protection sensitivity, typically has maximum accuracy of voltage VT

measurement under such conditions, and calculates the vectorial VI OP either

1 n locally
VTX
1  k˜orI˜ABremotely
nCTk AC  ˜ V(auto-setting),
n
 3 ˜ VT 0 V0 or B ˜ 1  k AB   VC ˜ 1  k AC 
 Vcontinuously in a
I OP Aslow
OP (( A
In adjusting
)) 3 I DIF
DIF (A )k k A loop ˜ IA x
(self-tuning). n
voltage sum with relatively high accuracy.
( A ) A A
1 n VT nCT DIF n VT
VOP VTX
1  k AB  k AC  ˜ Vx  3 ˜ VT 0 V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
When measuring the 3 • V0 directly the relay is presented with a Re ^1I k ABVT k AC  ˜VnCT
3 n
X  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  k AC ` 0
nVT
I ( A) I DIF  kA ˜ ˜ IA
very small signal under internal failures that require high protection kˆOP DIF ( A) ( A)
nCT DIF
Re ^1 IkA AB  k AC  ˜V X  3 ˜V0  VB ˜ 1 
k AB   VC ˜ 1  k AC ` (8a)
A
sensitivity. In order to keep high accuracy a high-sensitivity voltage
relay input shall be used. At the same time, this voltage could
Im ^1  k AB  k AC  ˜V X  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  k AC ` 00
I DIF ( A)
reach as high as system nominal voltage during external faults. kˆ
ImA ^1 I Ak AB  k AC  ˜V X  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  k AC ` 0
Therefore, the input range shall be high enough to measure this Z  Z2A
voltage correctly and balance it accurately against the VX signal.
I DIF ( A) VBANK ( A) 1 A
Z ˜Z (8b)
Z11AA  Z22AA
The VX voltage, in turn, is relatively small under internal failures that I V
Z1 A ˜ Z
DIF ( A ) BANK ( A )
The process ofZfinding the
2 A two unknown constants is based on
1A  Z 2 A
require high protection sensitivity. Therefore either the relay shall
be equipped with a high-sensitivity voltage input, or the VT ratio the
I following
V principle. When the bank is healthy, equation (5) is
Z 1 A ˜and
A BANK ( A )
perfectly balanced, Z 2 Atherefore it can be zeroed out. Writing the
is selected to create this signal and improve measuring accuracy Z1A  Z 2 A
of this signal, or both. In any case, the ratio must be selected such real
I A and imaginary
V BANK parts of the equation separately one obtains
˜ZZ 2unknowns.
( A)
two equations for Z 1 Atwo A˜ Z Z ˜Z
as the input voltage does not exceed the conversion range of a VBANK ( A) I DIF ( A) ˜ 1 A 2 A I A ˜ 1 A 2 A
given relay. Sometimes this requirement may be relaxed allowing Z1 A  Z 2 A Z1 A  Z 2 A
Z ˜Z Z ˜Z
VBANK ( A) I DIF ( A) ˜ 1 A 2 A I A ˜ 1 A 2 A
Z ˜ ZCapacitor Z1 A ZBanks
Z 2˜AZ Z1 A  Z 2 A
24 Fundamentals of Adaptive Protection of Large
I DIF ( A) 1 A 2 A  I A 1 A 2 A 0
Z1 A  Z 2 A Z1 A  Z 2 A
Z1 A ˜ Z 2 A Z ˜Z
I DIF ( A)  I A 1A 2 A 0
Z1 A ZZ 2 A Z Z1 A  Z 2 A
 Z ˜Z Z ˜Z
V
I ABANKV( ABANK I DIFZ( A1 )A ˜  Z1 A2 A 2 A I A ˜ 1 A 2 A
ZZ ˜ ZZ ZZ ˜ ZZ
)
VBANK ( A) I DIF (ZA1) A˜ ˜ Z11AA2 A 22AA I A ˜ 11AA 22AA
( A)

Z1 A  Z 2 A Z1 A  Z 2 A
Z1 A ˜ Z 2 A Z ˜ZZ1 A ˜ Z 2 A Z ˜ Z2 A
VI BANK  Icoefficient I Ato˜ 0the1 Adifferential
( A) ˜
The above is now solved for the two unknowns kAB and kAC while Dividing
DIF ( A ) both sides I˜DIF by the 1AA 2A
next current
( A )ZZ Z Z Z
Z ˜
 ZZ
treating the involved voltages as knowns (the k-values are treated I DIF ( A) 11AA 22AA ZI1AA 11ZAA2 A 22AA 0Z1 A  Z 2 A
gives:
as real numbers per equations (4)). The method works as long as Z1 A  Z 2 A Z1 A  Z 2 A
the Vx voltage is above the measuring error level. The procedure Z1 A  Z 2 A Z ˜ Z
does not call for the system to be unbalanced (V0 can be zero) as I DIF ( A)1Zn1IVTX AA ˜ Z 2 A (9e)
the unknowns (k) do not appear as multipliers for the V0 value in
I
VOP Z  1 
 
kZ I A k10A  ˜ V2 A  30˜ V  V ˜ 1  k   V ˜ 1  k 
I DIF ( A) 3Z1InAAVT Z1 A2 A 2 A Z10A  Z 2 A
DIF ( A ) AB AC x 0 B AB C AC
1 A 2 A
equations1(8).nVTX
VOP 1  k AB  k AC  ˜ Vx  3 ˜ V0  VB ˜ 1  k AB   VIntroducing C ˜ 1  k AC Z1 A  Z 2unbalance
the inherent A
compensating factor, k:
3 nVT 
4.3. Phase Current Balance (60P) Z Z  X A  X 2A
k A 1 nIVTX | 1 A Z 2 A Z 1
A ZZ22A1A  kZXAB11AAk0X AC22AA˜ Vx  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1 
1 A 2 A nVT 0
VI DIF ( AZ
11AA 
11kkABABk kACAC ˜V˜xVx 3˜3V˜0V0nVVT B V˜ B1˜1kABk AB  VC˜V1CkA˜ k1AC3k ACnVT 1 A | 2 A
1 nnVTX )
With
VVOP reference 
VTXto Figure 7, this function is based on the balance OP
n (10)
between
VOP 3phase
OP

to both 3
1 n
nnVTVT
grounded
currents
VTX
1  kofABthetwo  ˜ Vx banks,
k ACparallel
and ungrounded arrangements.
3 ˜ andVis0applicable
0
 VB ˜ 1  k AB   VZC1 A˜ 1 Zk2 AAC  X 1 A  X 2 A
Higher
VT

nVT nVT I OP^(A)Z1 A IDIFZ(2AA)  kXA 1˜AI A X 2 A

sensitivity can be achieved when using a window CT (compared Re
yields
k 1the following
k AB  k AC | ˜V X signal
operating 3 ˜V0 of the ˜ 1  kcurrent
VB phase AB   V C ˜ 1  k AC ` 0
balance
with the1 two individual CTs summated electrically).
11kAC k AC3˜V
nVTn0VT 0 With the two
VB1 C˜ 1˜kbeAB  ˜
1k AB C˜ 1 ˜1kA1ACk AC
nnVTX  
A
1 I I 2A 
k I
OP ^slightly
13 nk ABdifferent, kkABa˜Vcirculating
x V3Bn˜˜ 
VB 
AC `VC 0
protection:
xV 3 ˜ flows, 1V0kVAB ˜shall Z Z X X
VVOP AB X k AC ˜˜V 0  kV
VTX OP ( A ) DIF ( A ) A A
banks
Re 0current and V 1A 2A

compensated 3 nVTVTfor in order to increase sensitivity VTn

VT of the function. Im ^1  k AB  k AC  ˜V X nCT 3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  k AC ` 0
This protection element is founded on the following theory. I OP ( A) I DIF ( A)  k A ˜ ˜ I A
^      `  ˜ (11)
Re^^1  X 3 ˜ V0  0VB ˜ 1 B  k AB   ˜ 1C˜ 1 `
  ˜  ˜  ˜   ˜  I I k In n
Re
Im 1 1
 kkABABkkAC AC
k k ˜V
 ˜ VX 
V
 3
3
˜ V
V
 V
V
˜ 1
1
 k
k AB

V 
CV
V 1
k AC k
k I
AC
`
OP
0OP ( A )
0
0 I DIF ( A )
 k A
˜ A CT
CT DIF
˜ I
Both parallelABbanks AC work under identical voltage, and
AB therefore: ( A ) DIF ( A )
Z1 A 
A A
X 0 B C AC nCTZ DIF
2A
I DIF ( A) relations VBANK (apply
Im ^1  k AB  k AC  ˜VZX1 A3 ˜ZV20A VB ˜ 1  k AB   VC ˜ 1  k AC ` ˆ0 I DIF ( A) Identical to phases
Z1 A ˜ Z B and C.
A)

Im ^1( A) k ABVBANK
I DIF  k AC  ˜ V  3 ˜ V  V ˜ 1  k   V ˜ 
(9a)1  k k
I
ACThe` 0
A (operating I DIFsignal  k(11) ˜
nCT
2A
A implements ˜ I A proper compensation for
Z1 A ˜ Z 2 A
( A) X 0 B AB C OP A) ( A)
I I n
Z Z kˆA inherent unbalance
the
DIF A( A )
of the
Z 1 A toZ the
bank. The equation identifies that
CT DIF

I DIF ( A) VBANK ( A) Z1 A  Z2 A the

I error is
I
V BANK ( A)
Aproportional 2 A amount of the total phase current

I DIF ( A) VBANK ( AZ) 1ZA11AA˜ Z Z 2 A2 A (IAA) and the

I DIF (not
difference ˜ Z 2 A the impedances of the two banks
Z 1 Abetween
I A V BANK ( A) Z1 A ˜ Z 2 A 2A
(9b) (kˆA). When A ) compensated, the straight differential current
k A display a non-zero value “leaking” from the phase current.
Z1A ˜ Z 2 A would I
Z1A  Z 2 A SubtractingA the historical ˜ Z 2 Aof such leakage
Z1 Avalue Z1 A ˜ Zcurrent, often
( A ) ˜ sensitivity butI it A ˜is not a correct way of
I A Vthe
Utilizing BANKfact ( A ) the Z 1 A ˜Z Z 2 A
voltage is the same in expressions (9a) and VBANK ( Atoday, I DIF 2A
applied improves
Z Z1 A More  Z 2 Ain section 5
)
(9b)
I A oneVwrites: 1A 2A ˜ Z
Z Z ˜Z Z 2 A discussion Z1 Afollows
compensating this functions.
VBANK BANK I DIFZ(1AA) ˜ Z 21AA 2 A I A ˜ 1 A 2 A
( A)
of this paper.
Z Z Z Z
( A)
Z1 A 1˜AZ 2 A 2 A Z1 A ˜ 1ZA2 A 2 A
VBANK ( A) I DIF ( A) ˜ ˜
I A (9c) Note that Zequation 1 A ˜ Z 2 A(11) is a vectorial
Z1 A ˜ Z 2 Adifference between the two
Z1ZA1 A ˜ZZ Z1 AZ1 AZ˜2ZA 2 A I DIF ( A) However,
signals. as
theI Ak-factor 0
VBANK ( A) Z1IADIF ˜ Z( A2 )A ˜ Z1 A ˜ Z 2IAA ˜
2A 2A
Z1 A  Z 2 A Z1 A  Zis2 aA real number (very small
I DIF ( Athe
creating following balance
1IAA Zequation:
Z Z01 A  Z 2 A or zero imaginary part), the two currents are in phase and their
ZZ1 A1 A˜ Z2 Z 
) 2A
2 A Z Z ˜
1 AZ Z 2 A
magnitudes, not phasors, could be used as well.
I DIF ( A) A
 I A 1A 2 A 0 Z1 A to  Zmeasure
ZZ1 A ˜ Z Z1ZA  ˜ZZ2 A I DIF ( A) CTs
Typically I Aused 2A
0 the total phase current and the
1A Z 2 A
Z1 A  Z 2 A have drastically different ratios. The
2A
I DIF ( A) Z1 A  ZI A2 A 1A 2A
0 (9 d ) differential current would
I DIF ( A)Z1 AIA Z 2 A Z 0 Z 2 A differential CT might have much lower ratio in order to increase
Z1ZA 1A Z2ZA 2 A 1 A magnitude of the secondary current under internal bank failures
I DIF ( A)  I A 0  Z sensitivity X of X
Z1 A  Z 2 A requiring Z1 Ahigh protection. During external fault
Z1 A  Z 2 A kA
conditions,
2A
the differential| 1 Acurrent2 Aremains low further promoting
I DIF ( A) Z1IA A Z 2 A X 1 A 0 X 2 A the usage
Z1 Aof low-ratio
Z 2 A XCT.  Xthe2 Arelay side, a sensitive ground
1 A On
k A Z  ZZ1 A |XZ 2 A  X
1A  Z 2| 1A  X 2 A
Z
1A 2A 1XA 2A current input shall be used for better sensitivity and accuracy.
kA A
Z1 A  Z 2 A X 1 A  X 2 A When
I OP ( A)written )  k A ˜ I A terms, the key equation (11) when
I DIFin( Asecondary
Z1 A  Z 2 A X 1 A  X 2 A expressed in secondary amperes of the differential CT becomes:
| k ˜I
( A) X
k AI I DIF
OP ( AZ )  Z
I OP ( A) 1 AI DIF ( A2)A k A ˜ 1IAA
A AX
2A
n
I OP ( A) I DIF ( A)  k A ˜ CT ˜ I A (12)

I ( A()A k A ˜ I A CT ˜ I
I OP n n
I ( A) I DIF )  k A ˜nCT
CT DIF
I OPOP( A()A) I DIFDIF ( A)  k A ˜ nCT DIF ˜ IA A
The following characteristics apply to the phase current balance
nCT DIF
functionI DIF [3]:( A)
nCT kˆA
I OP ( A) I I DIF ( A)  k A ˜ ˜ IA • The element IA shall support individual per-phase settings.
ˆkˆ I DIFDIF ( A)
( A) nCT DIF
kAA
I AI A • The function indicates the effected phase, as well as reports
the change in the current division ratio, k (pre-fault and fault
I DIF ( A) values) to aid troubleshooting and repairs of the bank.
kˆA
IA • The element shall apply appropriate security measures for
Figure 7. sensitive but secure operation: appropriate restraint signal
Phase current balance application. could be provided to accompany the operating signal (11).
Disabling the restraint shall be possible if desired so.

Fundamentals of Adaptive Protection of Large Capacitor Banks 25

 Z1 A  Z 2 A Z1 A  Z 2 A I B1 VB  V X  ˜ YB1 ; I B 2 VB  VX  ˜ YB 2
Z1 A  Z 2 A
IICB11 VVCB VVXX ˜Y˜ YC1B;1 ; I CI2 B 2 VCVBVX V ˜XYC˜2YB 2
I DIF ( A)  I A 0 I C1 VC  V X  ˜ YC1 ; I C 2 VC  V X  ˜ YC 2
Z1 A  Z 2 A
IINC11 IV  V  ˜ Y ; I CV
A1C I B1 X I C1C1 V A  2 X ˜V YCA1 VVXB˜ YVCX2  ˜ YB1  VC  VX  ˜ YC1
• Several independent thresholds shall be provided per phase I
AtNthe I
1 same  I
A1 time  I
B1the total V
C1 currents A  V in 
X each  VBare
˜ YA1phase X  ˜ Yby
 Vdriven  VC  VX  ˜ YC1
B1 the
for Z1 A  Z 2and
alarming X1A  X 2 A
tripping.
IINN21 IIAA21  IIBB21 IICC21 V VAAVVX X ˜Y˜AY2A1 VBVB VXVX˜ YB˜2YB1VCVCVX V ˜XYC2˜ YC1
total admittance of the two banks in each phase:
kA A
|
1 A  Z dividers
TheZcurrent X(k)  Xindividually I N 2 I A 2  I B 2  I C 2 VA  VX  ˜ YA2  VB  VX  ˜ YB 2  VC  VX  ˜ YC 2
• 2A 1 Aare 2A set per phase.
I A I A1  I A2 V A  V X  ˜ YA1  YA 2  (15a)
• Both auto-setting and self-tuning applications of this method
IIDIF IIN 1 IIN 2 V V  ˜ 
I CA2 XVA AV1 X  ˜AY2A2  V
 Y  Y   V  V  ˜ 
B B  XV X  ˜ BY Y1B2 YB2V C 
VVC XV˜ XYC
I DIF I N 1  I N 2 VA  VX  ˜ YA1  YA 2   VB  VX (15b) ˜ YB1  YB 2   VC  V
N2 A2 B2
I OPare
( A ) possible. )  k A ˜ Icould
I DIF ( AProvision A be made to calculate factors k I B I B1  I B 2 VB  V X  ˜ YB1  YB 2 
automatically under manual supervision of the user, either IIA I A1I IA2 I V A VV XV ˜ YA1˜ YYA 2 Y   V  V  ˜ Y  Y   V
locally or remotely (auto-setting), or constantly in a slow
adjusting loop (self-setting).
II A IIA1 IIA2 VVA VVX  ˜X˜YY
DIF N1 N2 A
A1   YAY2  A2
A1

B X
(15c)
B1 B2 C

nCT IB IB1  IB2 VB  VX  ˜ YB1  YB2 

C C1 C2 C X C1 C2
I I k ˜ ˜I I A I A1  I A2 V A  V X  ˜ YA1  YA 2 
IB2 VB(15) X  ˜ equations
YB1  YB2 (14) allows eliminating the
The
OPprocess
( A) of
DIFfinding
( A) the
A balancing constants
A for each phase of
nCT DIF
protection is based on the following simple equation:
IB IB1 equations
Inserting  Vinto
I C I C1  I C 2 VC  V X  ˜ YC1 balance
voltages and derive the all-current  YC 2  equation for the two
IIB IBI1  IB2I VBV VX V˜ YB˜1Y YB2 Y 
banks:
I DIF ( A) C C1 C2 C X C1 C2
kˆA (under no-fault conditions) (13)
Y Y Y Y Y Y
IA IIDIF II A ˜ IA1 A2V IBV˜ B1˜ Y B 2YI C ˜ C1 C 2 (16)
C C1 YYAC 1A12YY 2 2C YY
X B1B1 B
YCY
12 C 2 YCY 1C1 CY
Y 2C 2
I DIF I A ˜ AA
 IB ˜ B2
 IC ˜
4.4 Neutral Current Balance (60N) YA1  YA2 YB1  YB 2 YC1  YC 2
Labeling:
YA1  YYA 2A1  YA 2 YB1  YYBB2 1  YB 2 YC1  YYCC21  YC 2
With reference to Figure 8, this function is based on the balance kI ADIF I A ˜ ; k B  I B ˜ ; k C  I C ˜
between interconnected neutral currents of two parallel banks, YYAA11YYAA2A21  YA 2 YYB1B1YYY 2B12  YB 2 YCY
BB 1C YYC2C1 2 YC 2
1 CY
and is applicable to both grounded and ungrounded installations. kA ; kB ; kC (17)
A window CT measuring the vectorial difference between the two YA1  YA 2 YB1  YB 2 YC1  YC 2
neutral currents allows for better accuracy/sensitivity. I OP YIADIF  k ˜ I A  kY
1  YA 2 A  YBk2C ˜ I C  YC1  YC 2
B B˜ 1I B
One ; koperating
k A gets the following equation ; kbalancing the protected
With the two banks possibly slightly different, a circulating zero- I OP YAI1DIF
bank: 
 YA 2k A ˜ I A  Y
B
k B 1˜ 
I BYB 2k C ˜ I C YC1  YC 2
C

sequence current may be present and shall be compensated for I OP I DIF  k 0 ˜ I 0  k1 ˜ I1  k 2 ˜ I 2 
in order to increase sensitivity of the function.
I OP
OP I DIF kk0A˜ ˜II0 Ak1k˜BI1˜ IBk2 ˜kIC2 ˜ I C 
DIF 
(18)
Proper inherent unbalance compensation is founded on the I OP I DIF  k1 ˜ I1
following theory.
DIF  1k 0 ˜1 I 0  k1 ˜ I 1  k 2 ˜ I 2 
Both parallel banks work under identical voltages, therefore their I OP
OP IIDIF k ˜I
ˆ
When
I
phase currents are driven by the individual admittances in each k1 the banks are identical, i.e. phases A are equal, phases B
DIF

phase of each bank: are equalI1 and phases C are equal, the operating equation (18)
IkˆOP I DIF
simplifies to astraight
I DIF k1 ˜ I1 overcurrent condition for the measured
1
neutral differential
I1 current.
III AA11 VVV AA  VVV XX  ˜˜˜ YYYAA11 ;;; III AA22 VVVAA  VVVXX  ˜˜˜ YYYAA22 (14a)
I OP I I DIF 
nCT
˜ k A ˜ I A  k B ˜ I B  kC ˜ I C 
A1 A X A1 A2 A X A2
kˆ1 DIF nCT DIF
II B1 V V  ˜ YBB11 ;; II BB 22 VVBB  VVXX  ˜˜ YYBB 22 I OP II1 DIF  CT ˜ k A ˜ I A  k B ˜ I B  kC ˜ I C 
n
VBBB  V XXX  ˜˜ Y
(14b)
I BB11 V V YB1 ; I B 2 VB  V X  ˜ YB 2 n DIF
VOP V1  k SET ˜CT V2
III C1 V V  V V ˜ Y ; ˜ YI C1 ;; V II C V  ˜YVCC  V XX  ˜˜ Y
A1
I CC11 V A
VCCC  V
X
V XXX  ˜˜ Y
A1
YCC11 ; I CC 222 V
A2 A
VC X
V VX ˜ YA2
YCCC 222 (14c)
I OP VI DIF  ˜ VCT ˜ k A ˜ I A  k B ˜ I B  kC ˜ I C 
n
V  k
VV VV ˜˜YY ;; II VV VV ˜˜YY nCT DIF
II OP 1 SET 2

Vcurrents X can  derived X  ˜˜ Y  V VXX  ˜˜ Y

A1 A X A1 A2 A X A2
II NN 11 sumII AAof11   IIneutral V ˜˜ Y
YAA11be VBB  1  VCC 
I N 1 V IAV1  IYB1; VAAAV V V
B1 B X B1 B2 B X B2
II BB11two V
The the
 ˜  II CCC111 VV V 
V ˜ X  ˜ YA1  VB  V X  ˜ YB1  VC  V X  ˜ YC1
X Vfrom
X Ythe
B1
B YCC11
Iabove
I
B1 Vequations:
B  V ˜ Y ; I
X B1 B2 V  V  ˜ Y B Y X B2
C1 C X C1 C2 C X C2
VOP V1  k SET ˜ V2
AV22   II BB 22; 
 II CC 22 V V A   X  ˜˜ Y A 22   B  X  ˜˜ Y 2   V  ˜ YCC 22
IIII N N 22 V I   VXV˜ YA˜ Y2 VVB VVX˜ Y ˜ YB 2  VCCC 
   VXXX  ˜˜ Y
V V V V
C1
N1 I N 2 I IAI2  I B 2 VI C2V ˜V
C
A1
I A X
B1
˜ Y C1
C1
I C2
A
V
Y A V
X
V
A C ˜VY
A1
X X Y AC2
B
V B
X
V X Y
B1
B2
B V
CV XYC 2 C1

V  V ˜Y ; I 
I II I I VV VV  ˜Y˜ Y  VV V
V  V ˜ Y
II NN 22 V X  ˜˜ Y VY˜ Y˜ Y  VV  V V ˜ Y(14d)
A 22  
 V VXX˜ Y ˜˜ Y YBB 22  
 V VXX  ˜˜ Y YCC 22 
I
N 11  A  A11  VBB  YBB11  VCC  YCC11 
IIA1 A X A1 A2 A X A2
II DIF
DIF I I   V  V Y V Y
II DIFV  VI N ˜1Y ; I NI 2 VVAV  ˜V Y X  ˜ YA1  YA A 2   VB  V X  ˜ YB1  YB 2   VC  V X  ˜ YC1  YC 2 
N1 A1 B1 C1 A X A1 B X B1 C X C1
N2 A2 I N
B2 C2 V A
A V X
X A2Y A Y B X B2 C X C2

II II  I V V V  ˜VY  ˜YY VV V V ˜ Y ˜ Y V  Y V  V
B1 B X B1 B2 B X B2
II
N2 A2 B2 C2 A X A2 B X ˜ Y  V  ˜ Y  Y 
B2 C X C2
II V I V  VVA  VX˜ Y  ˜˜ Y
˜ YI ; I
 VV Y YA 2 
DIF N1 N2 A X A1 A2 B X B1 B2 C X C1 C2
II AA II AA11  YA1 
I I A 2V V
I V X  ˜ Y A1  Y A 2  (14e)
C1 C X C1 C2 C X C2

1  V A 
A 2
˜ YVX Y  A1V  AV2  ˜ Y  Y   V  V  ˜ Y  Y 
A2 A
II A I I A
DIF
I
NA1 AAI  I I  I VV VV ˜˜ Y
11
N1
B1 A 2 C1
N2 A
AA XX Y  VY  V  ˜ Y  V  V  ˜ Y
X
A1A1
A1 A2
A2 X
B X B1 B2 C X C1 C2

VBBVVVV˜(14e)  ˜ YYYVBB11V(14d)

Y   V obtains:
B B1 C X C1
The differential current is a vectorial difference between the two
IIB IIB1  I IIIBBsubtracting
By
IIIBB I II IBB11I I
Icurrents.
B
A2
N
B1
A21
A BA
B2
22 22I V VV
B2V  V
C2
B VB˜ Y V˜XXXYYY ˜˜from
A
X
A X X
B1 YB1  YY ˜BBBY222 one
AA12
B2
A
B2 X  V ˜ Y
B2 C X C2

IIII I II IIII VIVVVV ˜V

II CC II CC11   V  ˜ Y V XY ˜˜ Y  Y 
˜YY  YY   V  V  ˜ Y  Y   V  V  ˜ Y  Y 
 
Figure 8.
 I C 2 VC  V X  ˜ YCC11  YCC 22 
DIF N1 N2 A X A1 A2 B X B1 B2 C X C1 C2
B B1 B2 B X B1 B2
C C1 C2 I C 2 C V CX V C1 Y C 1
C2 Y C 2 Neutral current balance application.
I C I  IC1 V C 2V  ˜ Y C Y  X (14f)
I CA IA1C1 AI2YC 2 AYVCX  V XA1Y ˜ YAC21 Y YC 2  YC1  YC 2
II DIFI II A ˜ VA1 YV A˜2Y Y IBY ˜  B1 B 2 Y
YXA1 B1 YAB22 1 IYB ˜2 Y
A1˜ YYAA12  Y A 2YB
 IC  ˜
1 YY YC1BB22Y YCC11 
Y Y YCC 22
C ˜˜ YC1  YC 2
B2 Y B 
B B1
II DIF I
I A ˜ YA1 A 2YB  I B ˜ B1 YY
Y B1
B  B2YC2 I C
C2 I
DIF
IIIDIFDIF I Y ˜ 
11V X ˜Y
 IY ˜   I ˜ 
YBBB111CYC
˜ Y  Y ˜ 
Y ˜ Y C YC 1  Y
Y
A B C 1
I C1 AI C 2 A VY
YYBBCAAA11222YBYY1CB22YB 2 Y
A 1 A 2 1 B 2
I I I
YA1 YYAA12 YYAA12  Y 1YCY YCC11  Y
A B
C C A
Y1YBBBC222Y2 C 2 YCCC 222
kA ; kB ; kC
YI A1˜ YA1YAY2 A2  I ˜ YYB1B1  YBY2 B 2 YC1  YYCC21  YC 2
A Y  Y  IC ˜  Y
A 2 YYBB1 1 1kY YYYBBCC2221 ;YkC 2 Y YCCCof111 
I DIF Y
A1YY AA111 2Y AY YBY2B 2Y Y Y
C1  Y Y
B

26 kk AA YA1YAY1 A2 YAB2 ;;; YkkkB1BB YB 2YB1 C YFundamentals

kkA YA ;2 kA2
A ;B1
B
B 2 ; Yk C Adaptive
YCCC 222 Protection of Large Capacitor Banks
   
Y C
C1 ; k
YA1  Y BC1  YC1  Y
I OPA Y IY YDIF  k Y ˜ I  k
2A  Y BB B YY ˜ I Y k ˜ I Y
 YCYB 2
C2 Y YC 2
YA1 ;k YAA 22 YBB11  YBB 22 YCC11  YCC 22
A AY C
A1 A2 B1 B2 C1 C2
kA ; kC
YA1  YAA12 Y Y Y Y
B

I OP I DIF  k A ˜ I AB1 kBB2 ˜ I B  k CC1˜ I C C2

I OP I DIF  k 0 ˜ I 0  k1 ˜ I1  k 2 ˜ I 2 
 I N1 I A1  I B1  I C1 VA  VX  ˜ YA1  VB  VX  ˜ YB1  VC  VX  ˜ YC1
IIBIDIF IB1  IINI1B˜2YAI1NV2BYAV2VXA ˜IYVB˜1XYB˜1YYAY1 B2 YA2  ˜ YVCB1  YVCX2 ˜ YB1  YI BN22  IVAC2  VI BX2 ˜ IYCC21  YVC A2  VX  ˜ YA2  VB  VX  ˜ YB 2  VC  VX  ˜
BV  VB2  ˜ Y  ICV
  XYC ˜2YB 2  VC  VX  ˜ YC 2
A I 
I NDIF2 I A 2 Y AB 1 2 Y AC
I 22 AYB1 X YB 2 A 2 YBC1 V
ICA ICA1  ICA2 VCA  V X  ˜ YCA1  YCA 2  I DIF I N 1  I N 2 VA  VX  ˜ YA1  YA 2   VB  VX  ˜ YB1  YB 2   V
ItIisDIF YIA1Nto
justified 1 YAI2N 2 the
assume Vbalancing
AY BV1 XY˜constants,
Y  YAk,2 are
B 2 A1 YC1real  V  ˜ YB1 5.
VBYnumbers.
C2 X B 2   VC  V X to
 YSensitivity ˜ YInternal
C1  YC 2  Bank Failures
k A this leaves the ;balance
Still, k B equation (18); with k C 3 unknowns. These
Iunknowns I  I   V  V  ˜  Y   Y   V A the X  ˜ Y A1  Y A 2 
Y
B1 A1 can YY
B2 AAbe  Y
12 calculated
Y Y Y  Y Y Y Y Y I A1  I A2defining  Voutlined
I B ˜based
B B A2 X B1B1 B1 B B 2on
2 B 2several Cmeasurements TheI Akey equations capacitor bank protection
IYA2  YV A conditions.
I AI1A˜unbalanced  V
X Y ˜ YA1 Y YA2 I C ˜ Y  Y
1 C1 C 2 C 2
IIDIF under
taken methods ((1), (5), (11) and (18)) allow not only proper compensation
A

    for the inherent bank unbalance, but also facilitate analysis of

A1 A2 B1 B2 C1 C2
 Iequation
I COP I CI1 DIF
Alternatively,  VkXBmay
˜ ICA(18)
C2 kA V ˜˜IYB Cbe
1  Y˜CI2C
k Cre-written from phase IB IB1 of protection. 
IB2 VB  VX ˜ YB1  YB2  
IB IB1  IBinto
coordinates, 2 V  
V
sequence
YA1  YA 2 B X ˜ Y B1 
Y
components:
YB1  YB 2
B2  YC1  YC 2
sensitivity

kA ; k ; k Each of the four methods as described in this paper is founded on

Y YA1k Y˜AI2 Yk ˜IYYB1 k Y˜BI2  Y Y C1Y YC 2(19)   
1  IC2 C  V X ˜ YC1  YC 2
B C
a balance
I C I Cequation thatVassumes:
IIIDIF I C1A˜ YI CA 22 0YVC0  VIBB11 ˜ ˜1YB 2 2 Y2 I C C˜1 C 2
OP YAII1 DIF
C X Y C 1 Y C2 Y Y
A1 A2 B1 B2 C1 C2 First, that the bank is intact in terms of experiencing a ground or
It is justified to assume that the positive-sequence current would
IIOP
leak I DIF 
OP into the

kk1 A˜ I˜1I A quantity
Yoperating
 k B ˜ I Bmore 
 k Cconsiderably
˜ IC compared
 YC 2
phase fault. Y  Y Y Y Y Y
1 Y B1  YBcomponents. I ˜ I ˜ I ˜
A1 A2 B1 B2 C1 C2
with
Ik DIF the ˜ Y
YAI1zero and  YI B1˜YY
negative-sequence  YC˜1YC1YTherefore I
YB1  Ybetween
 YA2 unbalance 1  YC 2
A
A 2
A2
B 2
2
I C2 Second,
DIF
thatA the
YA1 inherent
B C
the
YCcapacitor
only
A
;
A the positive-sequence
Y Y k B
leakage Y 
can ;
Y bek C
CeliminatedY toimprove phases does not change.
B2
 Y 2k ˜ I yields k1B1˜ I1aYslightly YC 2YofC 2the
YC1 form
B
IˆOP IYDIF
sensitivity.
k1
IA1DIF
A1

ThisAapproach
0
A 2
0
Y B 1
Bk
2
B

2 ˜ I simplified
2
2
C 1

A ground or phase fault violating the first assumption results

mathematically accurate equations (18) and (19): YA1  YA 2 in the Yoperating
B1  YB 2 equations,YC1  YC 2
YIA11  YA 2 Y Y Y Y in ksevere unbalance and leads to
I DIF  k;A k˜ IB A  kBB1 ˜ I BB2 k; Ck˜CI C  C1 C 2 (20)
; kB ; kC
1  YA 2 YB1  YThis  YC 2 is
I OP
k A
protection
YAoperation as expected.
I OP YI DIF Y  k1 ˜ I1 B 2 aspect YofC1operation
YB1  YB 2 YC1  YC 2
A
A1 A2 backed-up by overcurrent protection, and therefore is of secondary
importance.
˜ kbalancing ˜ IC 
nCT
I OP IWith  kunknown
DIF one 0 ˜ I 0  k1A˜ I 1 A k 2 B ˜ I˜2IB(k1)
˜ I  kfactor  the
kC auto-setting  k ˜ I  k ˜ I  k ˜ I 
 nkCTA ˜DIF 
OP
I DIF I I
 kbeB ˜ implemented
I B  k C ˜ I C simply
DIF
or self-tuning
Ikˆ OP I DIF procedures I A can as: A short
OP
or open
DIF
in a AsingleA
or several
B B
cans
C
violates
C
the second
1
I1 assumption, causes a minor unbalance in the operating equations,

V
I OP I DIF  k1 ˜ I1 (under no-fault conditions)
I OP VI1DIF k 0˜ V˜ I20  k1 ˜ I1  k 2 ˜ I 2 
 kSET
(21) I size of the
the OP
I DIFinternal
 k 0 failure.
˜ I 0  k1 ˜ I1  k 2 ˜ I 2 
and results in operation of protection set sensitive enough given
OP

I DIF  CT ˜ kthis A ˜ I A  k B ˜ I B  kC ˜ I C 
Unlike in previous methods,n compensating coefficient may be This latter way of responding to internal failures is critical for
I I DIF analysis
I OP ofI DIF  k1 ˜ I1sensitivity. For this purpose one could
protection
akˆOP
complex number.
I 1OP II DIF  kn1CT˜ IDIF 1 assume nominal system voltages and resulting currents, and use
Operating signal (18) or (19) implements proper compensation
1
the operating equations to determine the amount of the operating
for the inherent unbalance of the bank. Equation (20) is a good
VˆOP I V
practical 1  k SET ˜ V2
approximation.
ˆ inI DIF
signals
k1
response to any given unbalance in the bank.
DIF
k1
 CT ˜ k ˜ I A  kwhen C
n I1
I OP II(18)
Equation 1 DIFholds for primaryAcurrents, B ˜ IB  kC to
applied ˜ Isecondary 5.1 Sensitivity of the Voltage Differential
nCT following
amperes, it takes the DIF form: Function
˜ k A ˜ I A  k B ˜ I B  kC ˜ I C 
nCT on the application to grounded banks.
I DIFletus focus
ForI simplicity
V2 ˜ k A ˜ I A  k B ˜ I B  kC ˜ I C 
n
VI OP  kSET ˜ CT
VI1DIF (22) OP
Neglecting the phase nCTindex,
DIF
the operating signal in this method is
nCT DIF (equation (1a):

Typically the differential CT would be of lower ratio in order to VOP V1  k SET ˜ V2

VOP Vthe
increase 1  k SETof ˜the
level V2secondary current for internal failures that
call for increased sensitivity of protection. During external faults,
the differential current will be increased but not dramatically. The actual voltage-dividing ratio during internal failures of the
bank is:
The following characteristics apply to the neutral current balance
function [3]: Z BUS TAP  ZTAPGND Z BUS TAP CTAPGND
k FAIL 1 1
ZTAPGND ZTAP GND C BUS TAP
• The single element function does not indicate explicitly the
effected phase. Z Z Z CTAPGND
k FAIL Z BUS TAP  Z TAPGND 1  Z BUS TAP 1  C(23a)
• The function shall apply appropriate security measures Vk2FAIL
The ˜
1 TAP
BUS
V1voltage
tap
ZTAPGND
during
TAP GND
the failure is:
1  ZBUS TAP
TAP GND 1  CTAP GND
BUS TAP

for sensitive but secure operation (provision for a restraint

Z
k FAIL TAPGND Z TAP GND C BUS TAP
signal). 1
V2 V1 ˜ 1 (23b)
• Several independent thresholds shall be provided that can be VV1 1˜ ˜k1FAILk
VOP
V 2  SET
freely used for alarming and tripping. k FAILk FAIL
and the operating signal becomes:
• The positive-sequence compensating factor k1 shall be a k SET
setting. V
VOP
OP V 1 ˜ 1 
k kk SET
VOP 1V1 ˜ 1SET
 ˜100
FAIL % (23c)
• Provision could be made to calculate the k-factor V1 k FAIL k FAIL
automatically under manual supervision of the user, either
locally or remotely (auto-setting), or continuously in a slow V k
adjusting loop (self-tuning). 'VOP
As aOP 'k FAIL
% 1
percentage of% the
SET 'full
˜100C% bus voltage '
BUS TAP %
the operating
CTAP GND %
signal is:
VVOP kkSET
1 1  FAIL ˜100%
V1 1 k FAIL
VOP 1  k AB  k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
'VOP %3 'k FAIL % 'C BUS TAP % 'CTAPGND %
'VOP % 'k FAIL % 'C BUS TAP % 'CTAPGND %
Fundamentals of Adaptive Protection of Large
1 Capacitor Banks 27
1 1  k  k  ˜ Vx
VOP 31 1  kABAB AC
k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
VOP
VOP 3 1  k AB  k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
3
 Z BUS TAP  ZTAPGND 'VV1 OP %V1 ˜' k % 'C BUS TAP % 'CTAPGND %
 k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
Z C
k FAIL 1  BUS TAP 1  TAPGND VV VOPOP 1 1 1kkkFAIL
SET
k ABSET
FAIL
ZkTAPGND ZTAP GND C BUS TAP 'VOP %31  'k FAIL
OP k FAIL˜100'%C BUS TAP % 'CTAPGND %
%
VOP V1 ˜ 1  SET V1 k
' 1 'kFAIL
 k AB% k'ACCBUS  3%˜V0'CVTAP B ˜ 1  k AB   VC ˜ 1  k AC 
k FAIL V
VOP 1 1 FAIL ˜VxTAP GND %
 
OP %
1 V 1  k  k ˜ V
3 '1kSET CBUS 1 %k AB   VC ˜ 1  k AC 
V2 V1 ˜ 31 k
'
VOPOP
OPOP %1
V k AB%˜100
AB
 k'%AC
AC ˜VxTAP
x
 %3 ˜V0'CVTAP B ˜ GND
k FAIL As both 1 the
3 FAIL are close to unity, the abovesimplifies to:
1 1 FAIL k AB  k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
V k-values
k
VOP k VOP 1
1  SET ˜100% (24d) V 3 1  k AB  k AC  ˜ Vx
V1 k FAIL VOPOP | 3 11Vx (26b)
VOP V1 ˜ 1 
k SET '
VV V
OP
OP OP %

'
11 k  kkAB
FAILAB %
 kk'ACC ˜ ˜VV x
AC BUS xTAP %
 3 ˜ V 0' C V B ˜ 1
TAP GND %
 k AB   VC ˜ 1  k AC 
33(3e) helps
VBx11   ˜ Vx
Equation 1 calculating the amount of the neutral point
k FAIL  k ACzero-sequence
AB   VC 1  k AC 
Equation (24d) yields a proportional relationship between kFAIL and V OP | V k system
'VOP % 'kvoltage:
the operating FAIL % a ' C BUS TAP
change by %
'C k-value,
1% in theTAP GND % yields an voltage.
V OP 3Assuming  k AB voltage nil, the equation
Vx be re-arranged
OP | Vx1 k ˜˜VVxx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
can 1 to calculate the value of Vx:
extra 1% of nominal in the operating signal. VVOPOP 11kkAB k kkAC
AB AB AC AC
VOP 1 k SET 3
What is more1 1interesting,
V the bus-tap and ˜100
AC  ˜ Vcapacitances
 khowever,
k ABtap-ground %
x  3 ˜ V0  V B ˜ the k AB   VinC ˜ 1 V
is the relation between changes
1  increase VOPk AC|VVBx1  k AB   VC 1  k AC  (27a)
inVOP1 3 FAIL k and
V x
aV 2
˜ V1 
| Vx1  k AB AB , k V   Va ˜ V
1  , k a  1 ‘ 120 0
 k AC AC˜ Vx
the operating voltage. Given equation (23a) one can write: VVOPB 1 B A 1 k
AB C k
C A AC
V
   VCk1AC k AC 
OP
 
x
V
3 1 1 k k
'VOP %1 ' kFAIL %  k'CBUS 'CTAPGND % (25) Vx B AB AB
V OP
3
1 k AB AC ˜ V TAP %
x
Assuming
V
V B a
V
2 a balanced
˜

V
˜
1 Aa

1
, 2
k
˜ V
k1C

AB

kaABk˜AC
V VA ,a ˜ a1  k1‘
bus voltage:
1  k   0
AC120
x | V
VVOP a 2Ax˜ VA , VC1  ak AB ˜ V a 1‘1200 (27b)
B AB C AC
The above signifies that a 1% change in either of the bus-to-tap A , k AC
 
xB

| 3Vx1 voltage.  k AC  ˜Vwould x  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  VBk AC a 2 ˜ VAa,2 ˜ V

or
1
tap-to-ground capacitances yield 1% of bus nominal in 1 k k
VVOP
the operating
k AB
Vx V A ˜ 2
One simplifies

further:
AB
1C k AB a ˜VA a, ˜ 1a  k1AC
AC
‘120  0

V 1 ak ˜ 11kVkAB 1 akkAC ˜AC13 k AC  0

OP

Depending on the serial/parallel arrangement of the cans, it will VVBxx aV21BA˜V˜ 1A ,2kAB V  k a ˜
C AB
V  , j a k1‘ AC  k AB 
120
VVOP VB 1change
1
take a certain
 amount
 k AB AB kV
of
C 1˜ V
 xk AC 
shorted/opened cans to cause a single V x
Vx V A ˜
2a 1 ˜ 
ABk
1
C

AB 1 k ACk
AB AB 
AC 
A
a k˜ 12
AC  k AC  (27c)
percentage in the capacitance and an equivalent increase
 
AC
3  k AB The V 1a 2 ˜ 1  k AB 
1 k  k
AB,aj ˜ACa1AC3kk1‘ AC  k0AB 
x
1 k k
VVVBxx aV21A˜V˜ 2A1,k AB
in the operating1voltage. k ACfinal assessment of sensitivity has to A
take into account the actual arrangement of the capacitor bank. VC kaAB AC˜ V AC 120
 1kare   kj AC k AC
2 3
 and using
A
V | V Observing 1 
the k-values k kACABreal numbers  ktoABunity
close
An
VB interesting
OP
a 2 ˜ VA , question
x
VC ais˜ Vthe optimum
a 1‘120 location
0 of the tap. VVAx
properties of12
AB
the a-operand 1  k AB yields kthe 23 following:
A,
11 a 2k˜ AB ACk aj˜ k31 kk AC  k AB 
AC
Regardless of the number of parallel cans, the longer the string,
the higher the impedance. If so a single can failure would cause a
VV
VVxxA VA ˜12 j
31 1kk1AB k ABj 2AC
       k AC
k 1
VAx 1  2 k AB 1kACkABAB j kAC2AC k AC  k AB 
3
V 1 k
a 2 ˜AB1change
V 1

k
a ˜AC1  (bus-tap 1 k  k
AC 
smaller percentage C in the overall 2 2 2 2
kimpedance/capacitance. V x
Vx B
 k AB portions
Vx VZ A ˜ 1  kZAB  k AC Z
For best sensitivity both the and
C GND
tap-ground) VA 1 3 2 1 k 3 2k (27d)
shall
k FAIL be kept as short
BUS TAP
1 GND
TAPas
kpossible
AB 1 k AC asBUSmeasured
TAP
1 in TAP
the number of V  j  1  k k  j k k
cans. In reality, ZTAPthe number of cans ZTAP isnot
GND a variable.
C BUS Within this Vx A
21 1 23 21
AB
233
AC k 1
   k 
GND
  
TAP
restriction,2 half of the total length is the smallest0possible length.  j  k  j k
232 AC2kk1AB
VB a ˜ VA , VC a ˜ VA , a 1‘120 1 k k j k
VVxAx 12 2 AB 23 2 12AC k
  k AC  k AB  VVxA 2the actual operating
kk  jequation
1 1 3 tap is optimum from the Because j k (5)2compensates for the
Therefore
V2 V1 ˜ 1the  exact k ABmiddle
 k ACposition
 j of the V 2 212 k  k
2 k  1k
V
point x of view
k FAIL2of sensitivity. Under the 2 mid-tap both the portions inherent
A bank unbalance, it is
AB further AC justified to assume the ratios
(bus-tap and Z
VAin˜BUS
a 2
˜ 1
tap-ground)
TAP 1 

 kZ 
are  a ˜ 
protected
GND
1  kwith
Z  the same sensitivity
 k AC1  AC BUS TAP 1  TAPGND other C of
V the
Vx ratio
A
1

impedances j
3 1

to
2 be kk a  perfect
j
3
unity
k (say
2  kk
k  1AB
), and treat the
VkVFAIL
measured
xA the number of cans.
AB
TAPk AB 2 as a variable 2 2(kAC correspondingly): 2
k Z 1  k  k AC ZTAP GND C BUS TAP V
VOP V1 ˜ 1  SETTAP GND AB A 2k 2k
Often, the tap kisFAIL installed below the mid-point in order to apply 1 3 1 3
lower voltage VTs. This creates a classical trade-off between j  k j k
optimum 1 1
performance and low cost of installation.
Vx 2 2 2 2 k 1 (27e)
VOP2 V11˜k SET k 3  k1   j 3 k  k 
1 3
V
V   j ˜ AB  k  j k V 2  k 2  k
k k  1Bank
1 100 % AC AC AB
Vx 2 2 of the 22 A
5.2V1x Sensitivity 2 Compensated
FAIL
2k FAIL
V 1kk AB  k AC 2k
Neutral
VAA Voltage2 Unbalance Function
'VOP % 'k FAIL % k SET 'C BUS TAP % 'CTAPGND % The above equations means that only 1/3rd of the percentage
The
VOPanalysisV1 ˜shall
1  start with the full operating equation (5): change in the ratio of impedances between any two phases will
k FAIL be seen as a percentage of nominal bus voltage:
VOP
1
11  k AB 3k AC 1˜Vx  3 ˜V03 VB ˜ 1  k AB   VC ˜ 1  k AC 
Vx 3 2  j 2  2 k  j 2 k k 1 'VOP %
1
'k%
1
'C% (28)
VOP k SET
1the following˜2100 3 3
1  k  k  ˜ Vk 2k
inVwhich 1 %
assumptions can be made:
VVA OP1 AB
FAIL AC x
- 3 1-k terms can be neglected for simplicity.
The For example it will take 3% in the drop of the phase A impedance,
to see 1% of'bus ˜ VBUS voltage as the Vx signal, and thus the
C% nominal
V'OPVOP
nVTX
- |The% 'k FAIL
Vxsystem % 'C BUS voltage
zero-sequence TAP % 'Cbe
can considered zero
TAP GND %
operating3 ˜ 3 ˜V the function.
signal of
(the system is practically always strong enough to maintain SEC ( MIN )

the balance at the bus despite few cans affected within the The operating signal has an arbitrary factor 1/3rd to comply
VB11itself).
 k AB   VC 1  k AC  with the common understanding of this method (equation (6)).
V bank
VxOP 11kk ABk k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜  1
Using 
0.01 ˜ 345kV
k AC microprocessor-based relay technology this scaling is not
3 to the following nVTX 1328
This leads AB AC
relationship: important 3as˜ any
3 ˜ 0scaling
.5V can be handled accurately. What is
important is the 1:3 ratio between the measured neutral point
VB ˜ VA , VC
a2 1 a ˜ VA , a 1‘1200 voltage and changes in the capacitor impedance.
VOP 1  k AB  k AC  ˜ Vx (26a) I OP I DIF  k SET ˜ I
3
a ˜ 1  k AB   a ˜ 1  k AC 
2

Vx VA ˜
1  k AB  k AC 'I OP 1
VOP | Vx 'k%
I 100
1
1
k AB  k AC   j 3 k ACFundamentals
k 
28 Vx VB21  k AB   VC 12 k AC  AB of Adaptive Protection of Large Capacitor Banks
X1  X 2
Vx
V 1  k AB  k AC
A 1  k AB  k AC k
X1  X 2
 I OP 0.'01
nn'VTX
˜ VBUS
1C%˜ 345
1 ' k%1kV 1328
3 ˜
'VIOP % 3100 3 ˜
'k%SEC.5( MIN
V V'C) %
3˜'C3%˜ 0˜ V
VTX

nVTX 3BUS
3 ˜ 3 ˜VSEC ( MIN )
X 1I0.01 'X C˜%2345 kV
nkInVTX
OP DIF  BUS˜ I 1328
k˜ VSET
This reinforces using low-ratio VTs for measuring the neutral-point Again,Xthe
VTX
13
3˜˜above
X332˜V˜ 0SEC .5V
observation may be used to select the ratio of
voltage. the split-phase
0.01 ˜ 345
CT:
kV)
( MIN
the target accuracy allows calculating the
n'VTX I OP 3primary 1 3 ˜ 0operating 1328
Relation (28) can also be used to calculate the required ratio.
minimum
I OP I0DIF ˜
1.01˜ 345k'SET kkV. ˜ 1I
5 V signal; the minimum relay sensitivity
allows
' I
k%
nVTX determining
'X %
100
%the minimum
'1328
C% accurately measured secondary
For example, assuming target sensitivity for the function, one signal; ˜ 3 ˜dictates
the32ratio 0.5V2 the maximum CT ratio that can be applied
calculates the effective operating signal as percentage of the bus inI'OP this
I OP case: I DIF1  k SET ˜ I
voltage. Using relay accuracy claim, one determines the minimum X 1  X '2 k%
1 that is1required for the proper operation of the
secondary voltage
'VCombining 'the 'C%
kI OPI I DIF '
100C k˜ I ˜ I
relay. OP % k%two requirements allows calculating the ratio n'DIF X 1  X%2 SETNOM (31)
3 3 I OP 2 ˜1I
for the VT: X  1X 2 SEC ' (kMIN )
k'IIOP 1 100
%
1 1 'k% 1
'VOP % '' Ck%%˜ VBUS'C% 5.4 X 1 100
'kI%Sensitivity
1 X2
'XZ%  of ' theC%Neutral Current Balance
nVTX 3 3 (29) Z
A )1
3 ˜ 3 ˜VSEC ( MIN ) ItI DIF (X
is worth
2noticing
IXA 2 that 1 A 2 this
2 A
0
method is a derivative of the phase
k
current X 1 X
balance
Z 
1
approach
Z (60P), and as such it has identical
'k% X 1 'X %2
1 A 2 A
Forn example,'with C% ˜ V 'C%
1 2
k
the BUStarget sensitivity of 1% of impedance X 1 2'C
sensitivity. X 2% ˜ I NOM 2
change on30a˜.01 3˜˜V
345kV bus, kVand the minimum relay voltage of 0.5V
VTX
345 nDIF
nVTX the maximum
secondary,
SEC ( MIN )
1328
VT ratio is: The balance 2 ˜ Iequations
Z1(BMIN  )Zfor all three phases per the 60P protection
3 ˜ 3 ˜ 0.5V I DIF ( B ) '
principle
are:
1CI B SEC
˜ 1 2B 0
'
n'DIF k
k%%
1 ' I
'X % %1B 'C2%B %
% X Z NOM
1 Z ' C
0.01 ˜ 345kV 222˜ I 22
nVTX 1328 Z(1MIN A  Z2A
SEC )
(32a)
I OP I3DIF˜ 3 k˜ 0SET .5V˜ I I DIF ( A)  I A Z  Z 0
' 1C  Z 2 C
1 1 InDIF 'CCI%C%Z˜ ZI˜ NOM
1IANOM
Z 2 A2 A 0
'Vthis 'k% SLG fault 'C%on the bus, the secondary voltage n DIF ( C ) 1 A 
With OP %ratio, under I DIF ( A) 22˜I˜IAISEC ( MIN
DIF Z 1C ) Z 2 C 0
DIF1This
would I'OPI OPbe I150V. 3 k SETis˜ Iwell 3 within the range of modern relays. Z
SEC ( MIN  Z)
Assuming a relay' k% Z1 A1B  Z2 A2 B (32b)
conversion range of 260VRMS, the ratio can be I DIF ( B )  I BZ  Z 0
I
lowered to 1328*150/260
100
' ˜ = 766, yielding the operating signal of II DIF ( A)  I I A1Z1A I A22Z ,A I DIF I B1  I B 2 , I DIF ( C ) I C1  I C 2
'V 1 C V 1 A ZZ 1B ZZ 2 B 0 ( B )
0.87Vn' I OP
secondary 1 'at%k%1%BUS
k% change 'C%in the capacitor impedance. DIF((BA)) IIBA 1 A
DIF ( A )
Z 1B1 A Z 22AB2 A 0 0
˜ 3X 32'˜V
VTX OP % I DIF
I X 1 3 3 ZZ1B1 AZZ
I DIF I N 1 Z1CI N2 Z 2CI A1  I B1  I C1   I A 2  I B 2 (32c) IC 2  
100 SEC ( MIN ) 2 B2 A
5.3 k Sensitivity of the phase current balance
X1  X 2 II DIF ( C ) I I C 1B 2 B 0 0
Z  Z
function 'C˜ %345 ˜ VkV B ZZ 
nnVTX 1 01.01
X X  ZZZ Z2C2C
DIF ( B )
IIDIF ( C) III ZZ
I1AB21BZ2BI B2 1B 0I0B 2   I C1  I C 2 
2 1BUS 1328 1C1C
' k
OP % 31˜Xphase ' 1 'Cthe 
33˜'23k3%%˜V
V k
'VOP %X 1 1the
DIF ( B ) A1 CB Z
Neglecting ˜ 0SEC.index, 'C) % operating signal of this method is 1C  Z
VTX %
513V( MIN Observing thatZ in the 60P 2C
2 Bmethod:
'k% (11)):
(equation ' X
3 % 3 % ' C 1B
I DIF ( A) I AZ1 1C I AZ22,C I DIF ( B ) I B1  I B 2 , I DIF ( C ) I C1  I C 2
2 2 I DIF ( C )  I C 0
˜ IIDIFDIF( A ) I DIF Z A 2IZ,DIF
1)CI  ( B )( B) I DIF I B(1C) I B 2 , I DIF ( C ) I C1  I C 2
'
10 ' . 01
C ˜
345 V 1
% k BUS˜'I C%1328 kV I A1Z(A
1C  Z 2 C
I
nI OPk% I DIF ''CX%%˜ VSETBUS
2 C DIF
( C )  II C  I
N 1 Z N2 Z I A1  I B1  I C1   I A 2  IC 2  
nVTX I DIF 0
˜ I˜ 0NOM I B 2 (33a)
VTX
nVTX 3 ˜ 3%˜V 2' 3 C 3˜ SEC .5(2MIN
V) I DIF
'
DIF 3 ˜to 3
It isnjustified
I 2 ˜1 I ˜
assume
V SEC ( MINtotal
the ) capacitor current does not change I DIF
I DIF
While
( A) II A11  IImethod:
in theN60N
1C
AN22, I DIFI A( B1) IIBB11IICB12,IDIF
2C
) I BI2C1IICC22
I A(2C 
in response to
'C%˜'kI kNOMthe internal failure of limited size, therefore the
I A1  I A2 I  ,I BI1  I B 2  I I C1 I I C,2 I
OP SEC ( MIN )
% ˜I 
˜ 345 (AI) INI1IA1  I C1  I C 2
nI OP I 0Icurrent I A 2 I ADIF ( BB)1I CB11IIBA22  IDIF B2  IC 2  
operating as a percentage of the total capacitor current
A 2 N2 I B1 
.DIF
01100 kV I DIF
1 IB I
SET
˜ I˜SEC  I
DIF (C )
nVTX the02.percentage
equals 01 345 kV )change 1328 in the k-value: DIF
Z  Z 2 A1328
A1 2 C1 C2

I DIF ( A)33˜˜ I A33 ˜˜ 001..A55V

( MIN
nVTX 0 I DIF I I DIF  IIDIF( BI) IDIF
k
'I OPX 1  1X 2Z V 
Z'11AkA% Z 22AA
Z (30a) I I A1 I (IA(1AA)2) II NDIF
DIFDIF I DIF N 2B1( B )   (ICIB)C(1C1 
IBAI21DIF ) IIC 2   I
C1 A2  I B 2  I C 2  
I I
I OPDIF ( AI) DIFX   100
I A k2 SET ˜ I
X 0 (33b)
1 A ˜ I
1
I OP I DIF  kZZSET Z I DIF II DIF( AI)  IDIF
A 2  I B1  I B 2   I C 1  I C 2 
 Z22AB in the k-factor yields 1% of the full  ( B )  I DIF ( C )
ForI example, 1% of1B change
B )1
A1
11 XB 2Z1B by
DIF (X I 0 allows one to insert (33a) into (33b) and obtain:
current
'' kI OP as measured
'   1Zthe
'
split-phase CT.
'IIDIF k %X 
21to
X'2k%1B% 2 2 B % 0
Z
I BXunderstand
Z 2B C
Next IOPstep is
( B ) 1100 'Zk1B%  Z 2 Bthe impact of impedance/capacitance I DIF I DIF ( A)  I DIF ( B )  I DIF ( C ) (33c)
changes I on the
100 Z  Z changes in the k-factor. From equation (10):
I DIF ( C ) ' 1XC I C ˜ 1IC 1 2C 0
 ZZ'
Now inserting (32a-c) into (33c) yields:
k' k%X 1  X'2%XZZ%11CCNOM
nDIF X C
22 CC %
I
k DIFX(C1) XC2 SEC  2 I˜ I 2 0
Z  Z2 A Z  Z2B
1 2
Z Z
X 1  X 2 Z1C  Z 2C
( MIN )
I DIF I A 1 A  I B 1B  I C 1C 2C (34a)
ObservingI DIF ( A) thatI Athe  I A 2reactances , I DIF ( B ) areI Bvery 1 I B 2 , Ione I C1  I C 2 Z1 A  Z 2 A Z1B  Z 2 B Z1C  Z 2C
'C 1 ˜ Itwo  Z2A
similar, obtains:
DIF ( C )
nI DIF 1 I % Z 1A 1
NOM
'kDIF % ( A1 '˜XAIA1% (IMIN A 2 ,'C I DIF ( B )0 I B1  I B 2 , I DIF ( C ) I C1  I C 2
Z1 AI12')ZC2 A%%I  I  I   I  I (30b)
( A )) 2

B 2  I C 2  I  I
DIF
'k%I DIF2 'IXN%SEC Y1 A  Y2 A Y Y Y  Y2C
2 1
2N 2 A1 B1 C1 A2  I B 1B 2 B  I C 1C 0
I DIF I N 1  I N 2 I A1  I B1  I C1   I A 2  I B 2  I C 2   Y1 A  Y2 A Y1B  Y2 B Y1C  Y2C
DIF A

Equations'(30) ˜IZ Z
II DIF ( A' CI%mean
2 ofthe
Athat ZIfor   I Cthere 2  be increase in
A each % of change in the impedance/
ZI NOM
n I 1  I0 1  IC
11B 22 B
DIF ( B )) 
capacitance AC I%BA˜one
1 of IZANOM Bparallel 0B 2banks, will
I2A˜1 ISEC 11B

) by A I of the n ACTUAL nIDEAL ˜ b, b 1  0.005‘0.30
B 2  I C 1  I C 2 
˜ A
DIF Z
n
the DIF 2
differential I I AZ(2MIN
current IZB0.5%
221 B total bank current.
I DIF I DIF ( A)  I DIF ( B )  I DIF ( C )
SEC ( MIN )

IDIFII B(ZA1Z Z1BZZ2Z(A2B2BC)  I0DIF (C )

B ))  I A CZ1Z) AA1CIDIF
II DIF n
( A) 
I DIF DIF (( C ZZ 0 0 VOP ( A) V1 A  k A ˜ VT 2 ˜ V2 A
I DIF ( A)  I A Z1ZA11BC Z 22ZAA22BC 0
DIF
nVT 1
Z1 A  Z 2 A
1CI 
Fundamentals of Adaptive Protection of Large Capacitor Banks 29
Z11Z
DIF ( A )  II A
II DIF B  AZ
Z 2C
2 ,2 BI DIF ( B0) I B1  I B 2 , I DIF ( C ) I C1  I C 2 n
I DIF ( C) I CZ  0 VOP ( A) V1 A  k A ˜ b ˜ VT 2 ˜ V2 A
I DIF ( B )  I B Z11ZBB1C Z 22ZBB2C 0
(B) B Z nVT 1
Z1B  Z 2 B
I I I  I  I  I   I  I  I  
 YZ11AAYZ2 A2 A Y1ZB1 B Y2ZB 2 B Y1CZ1CY2CZ 2C
YY11AA Y1YB   Y Y
B IC IIC1CYZ1C1C 2CYZ2C2C0 0
YY2 A Y2YB
I Z 1 A  Z22AAI BI Z11BB  Z22B
DIF  IIAAA Y  Y
IIIDIF 
YY1n1AA I 
YY2 A2 Ab, bY1YZB1B1  Y.13C YY
22BB   Z1C1C02CYZ22CC2C 0
B Y Y
Y2Y0BZ2.005 YC1C0Y
 IA Z A  Z2˜2A A  I B 11BB  I‘
DIF B
nI DIF
ACTUAL 11IDEAL
A
B
Y1 A  Y2 A Y1B  Y2 B Y1C  Y2C
C

Z1 A  Z 2 A Z1B  Z 2 B Z1C  Z 2C ˜ b, b 1 0.005‘0.30

I DIF I A the relationbetween
Observing IB  IC nnACTUAL n
YnIDEAL ˜ b,nequation
 Ythe Y20B.perfectly
bY1B 1was 005 ‘ .
Y10Cbalanced
3 Y2C making 0
Z1 A  Z 2 A Z1B the
Z 2impedance
Z1C and
 Z 2admittance Now assume that
 I A V1 Asignal  VT  Ibut
 
ACTUAL IDEAL2A
I DIF  ˜ I B2a ˜perfect 0
one can re-write the above into: the operating A˜ b,
above bY1B 21AY20B.005 ‘
zero, C one of
B C VnOP ( A)
ACTUAL Y1n k
A  Y2 A n
1 AIDEAL V 30Y2the
Y10C . VTs, say
the tap VT (#2), works with
n VT2 1an error of b. If so, the operating signal
C

Y1 A  Y2 A Y Y Y  Y2C becomes
V V  k ˜ VT
non-zero: nVT 2˜ V
I DIF  I A  I B 1B 2 B  I C 1C 0 (34b) V OP ( A )
V1n1AA  kA˜Ab˜,nnVT 1b ˜ V 12A 0.005‘0.30
2A
Y1 A  Y2 A Y1B  Y2 B Y1C  Y2C n OP ( A)
VOP ( A) V1 A  k A ˜ VTn1VT˜ V
ACTUAL IDEAL n VT 2
VOP ( A) V1 A  k A ˜ bnVT ˜ 1 2 2˜AV2 A (35a)
˜ b, b 1  0.005‘0.3
Which is precisely the 60N balance equation as
0 derived in section n
n n n˜ n2VT 2 ˜ V
VT 1
4.4ACTUAL
(equationIDEAL (18)). VVOP V11AA  kkAA ˜˜ b VT
V n˜VTV2 A 2 A
1 A  k Abalance, ˜nbVT˜n1VT ˜ V2 A (1c) can be solved for the
OP (( A A))
VOP ( A) aVperfect
Assuming nnVT1 equation
The above proves, that neglecting CT and relay accuracy the 60P tap
V voltage: V  nk ˜ b ˜ VT 1 ˜ V
2
n
n
0OP (V 1A  k A ˜ ˜ V2nAVTo k2AA˜ VT 2 ˜ V2 A V1 A
A) 1A A2
and
VOP (60N functions
V1 A  k A have ˜ VT 2identical
˜ V2 A sensitivity. Specifically, per each
VT
A) 1
percent of change innVT the1 impedance/capacitance of one of the nnVTVT2 1 nVT 2 nVTn2VT 1
banks, the differential CT would see an increase of 0.5% of the V 0OP (V 1 A  V k A ˜
 knA ˜ b˜ V˜ 2 A o ˜ k
V A ˜ n ˜2V2 A V1 A (35b)
0 V1 A  k A ˜nnVTVT1 2 ˜ Vn2VTA 1 o k A n˜VTnVT ˜ V2 A V1 A
A) 1A 2A
total bank current. 1

A )1 A Vk 1 AA ˜ nbVT ˜V 1 1˜AV Vo

2 A 1A k A ˜
˜ 1  nbVTVT 1 ˜ V
nVT 2 V0OP (V 2 2
V1 A
1 A  k Aof ˜ bthe
˜ method ˜ V2 A(60P) is easier to compensate for VT
VOP (phase
The A) Vvariant Substituting (35b) n into (35a) yields: n
2A

the inherent bank unbalance. n VT 1 The neutral variant of the method VOP ( A) V1 A nb ˜ V1 A V1 A ˜ 1  nb VT 1 VT 1

0V V1 A Vk A ˜ VT ˜ V o k ˜ VTb2 ˜ V2 A V1 A
1 A nb ˜ V1 A2 A V1 A A˜ 1 
2
(60N) requires 1 CT and relay input, compared with 3 sets for the V (35c)
phase version (60P). If applied concurrently on one relay, the two
OP
OP ((AA))
 ˜ nVT 1
n 2 nVT 2 redundant using different V
VOP V  b ˜ V1 A V1 A ˜ 1  b
1 b VT 100
1 %
functions
0 V1 A may k A ˜ beVTtreated˜ V2 A o as k partially
A˜ ˜ V2 A V1 A V1((AAA)) 1 1 Athe
OP
Or expressing b ˜ 100error% as a proportion of the bus voltage:
CTs and relay inputs. nVT 1 nVT 1
VVOPVOP1 A( A) V  b ˜ V
( A) 11 A b ˜ 100 1 A % V1 A ˜ 1  b
VV
VOP1 A( A) 1  b ˜ 100%
OP ( A )
(35d)
6.
VOPSensitivity V1 A  b ˜ V1 to V1 A ˜ 1  b
Instrumentation Errors V V 1  1  0.005‘0.030 ˜ 100% 0.72%
V1(AA) 1  1  0.005‘0.3 ˜ 100% 0.72%
( A) A
VOP
OP 1 (AA )

This section analyses impact of finite accuracy of Instrument V 1  b ˜ 100%

V
V
Transformers
OP ( A )
1  (ITs)
b ˜ 100and the relay on the four protection methods. For VOP11example,
VV
A( A )
A 1 with 1 negative 0.005‘ 0.5%0.30magnitude
˜ 100% error 0.72and % 0.3 deg
% angleOP ( A error,
)
11  1  0.005‘0.3 ˜ 100% 0.72%
the spurious operating voltage would read:
nVTX ˜ 1  k AB  k AC  ˜ Vx  nVT ˜ VA  k AB ˜ nVT ˜ VB  k AC ˜
0
It V
1A
is1 Aimportant to notice that errors of instrument transformers VOP
OP1(AA ) 3 ˜ nVT nVTX ˜ 1  k AB  0k AC  ˜ Vx  nVT ˜ V A  k AB ˜ nVT ˜ VB  k AC ˜ nVT ˜
V 1
VOP
3 ˜ n1VT 1  0.005‘0.3 ˜ 100% 0.72%
and the relay can be accounted for when tuning the coefficients. V
IfVthe tuning coefficients (k) are implemented as real numbers, the
nVTX ˜ 1  k AB  k AC  ˜ Vx  nVT ˜ VA  k AB ˜ nVT ˜ VB  k AC ˜
V1 A 1
OP ( A )
magnitude1errors  1  0can .005 be‘ 0.3 ˜ 100and
eliminated,
0
% the 0.72 impact
% of angular VOP
V1 A could be reduced. If the coefficients are implemented as
errors V 3is˜ 11natVTthenlevel˜that
VTX ˜ 3 1˜Vencroaches
k AB n k˜AC
x 
 ˜ Vthenx VTtargeted
nVT ˜Vsensitivity
A  k AB ˜ nVT ˜ VB  k AC ˜
VAVA nVTwould ˜ V˜BVaccommodate
B nVTn˜VTVC˜ VC
The
VOP error 1 on
nnVTX
˜1nnVT however,
OP
VTX˜ 3 ˜ Vthat
x  this
nVTVT˜method
complex numbers, both magnitude and angle errors can be V 3Note,
˜
˜ nVTVT nVTX ˜ 1  k AB  k AC  ˜ Vx  nVT ˜ VA  k AB ˜ nVT ˜ VB  k AC ˜ nVT
settings.
OP 3
accounted for. VOP of3the
some 3 ˜ n1VTerror in the matching factor k, leaving only a small
n ˜ 1  k AB  k AC  ˜ Vx  nVT ˜ VA  k AB ˜ nVT ˜ VB  variable
1
VOP kVAC ˜ n fraction ˜ V nofVTXthis ˜ 3error
˜ Vx unaccounted
nVT ˜ VA  nfor. VT ˜ VB  nVT ˜ VC
Assuming 0.15%
However, 3 ˜ nthe ITVTX
and relay errors will slightly change with the OP VT 1C
˜ ˜ ˜  and˜the
AV VTn ˜ V˜BV n
 ˜V
magnitude
V 3 1n
error nfor both 3 the
V ITs n V relay,
n and 0.2deg angle
VTX ˜ b ˜ 3 x˜ V VTn ˜ VTVTn˜VT
VC˜CVC
VT
magnitude of the signal and /or other factors such as residual flux 1 VT
V VOP
OP
error
OP gives 3 ˜1 n0.38% n nVTX of ˜ bus
b ˜ 3 ˜
voltage
V  n read˜ V as  an ˜
spuriousV B Bn operating
˜
VTX x VT A VT
or temperature. Even if tuned at one particular operating point, VOP 3 ˜ nVTVT nVTX ˜ 3 ˜ Vx  nVT ˜ VA  nVT ˜ VB  nVT ˜ VC
3 n VT x VT A VT
signal.
the method
VOP
1 will show some errors at different operating point due
nVTX ˜ 3 ˜ Vx  nVT ˜ VA  nVT ˜ VB  nVT ˜ VC 3 ˜ n1VT
to the IT3and ˜ nVTrelay inaccuracies. It is important to realize, though, ItVOPis important nVTX
to ˜ b ˜ 3 ˜ Vxthat
understand  nVT the˜ Vmethod
A  nVTcompares˜ VB  nVTtwo ˜VC
˜˜ ˜x1nx VTerrors
3˜˜VBoth ˜ ˜ ˜  ˜ ˜ ˜ ˜
   ˜ ˜ ˜ ˜  VT ˜ VC
that these errors occur regardless of the protection principle. By nVTX
voltages.
V VTX 3 3 V n n
n
VT VT V V
will
A bA 3n
playn
VTVVT a V V
role.
B n
B n n
They
VT V
VT V may
C
V n
C cancelV mutually,
n
3 ˜1nVT n ˜ b ˜ 3 ˜ V  n ˜ V  n ˜ V  n ˜V
OP VTX x VT A VT B
compensating for bank inherent unbalance, and partially for IT and or add up.
relay errors, 1 the methods
V
 Z˜ b2˜A3 ˜presented Zn1VTB ˜in
VAZthis
2 Bnpaper VBare
Z1Cnalready
 ˜ZVC2Cless OP
31˜˜VnVT n ˜ V  n ˜ V  n ˜ V
VTX x VT A VT B VT C
VIx Berrors.  ˜Ianalysis
VOP Z nAVTX n ˜ 3
susceptible to 1instrumentation Detailed 1
bbx 1of 1n˜VT˜Instrumentation
VVA AAVV VCVCB b VT
I DIF 3I˜ nAVT VTfollows.
6.2 VTXImpact
bErrors
1˜1VC˜0 V0on the
VT
B B 
VT
Z1 A  Z 2 A Z1B  Z 2 B
C
Z1C  Z 2C VOP OP ˜ 3 ˜ V
nVTX 3 VT ˜ V A  nVT ˜ VB  nVT ˜ VC
Magnitude and angle errors of ITs and the relay can be modeled as Compensated
nVTX ˜ 3 ˜ Vx nVT ˜Bank
3 x
VA  nNeutral Voltage Unbalance
ancomplex multiplier applied for the analysis purposes to the ideal VT ˜ VB  nVT ˜ VC
˜ ˜ ˜  ˜  ˜
VOP 1 b  1 ˜ VA  VB  VC  b  1 ˜ V0
Function
3 V n V n V n V 1
 Y2 Aof a given  Y2 B For example, Y1C  Ya2Cnegative
VTX x VT A VT B VT C
transformation Y1 A ratio Y1Bsignal.
   1 b illustrated 1 ˜ VA inVthe B  VC  subsection
b  1 ˜ V0applies to this
I
0.5% I
DIF magnitude error combinedI I
with a 0.3 Cdeg angle error can 0 VOPapproach
The
3 previous
A
YZ  Y Z B
Y Z  
Y Z Y Z  Y Z 3 method
b  1 ˜ as  VC  the
 VBExamining b key 1 ˜operating
beI
VOP modeled 1 IAb  1 ˜ VA  VB BVC  b  1 ˜ V0 C
as:
11A A 2 2
A A
 I 1 B 1 B 2 B 2 B
 I 1C 1C 2 C 2 C V
protection
OP
3
VAwell. V0 equation
3 Z1 A  Z 2 A Z1B  Z 2 B Z1C  Z 2C
DIF
for secondary voltages (7) leads to a conclusion that during
n ACTUAL nIDEAL ˜ b, b 1  0.005‘0.30
normal system conditions four voltage components, each of a
very small or zero magnitude, are added as vectors: neutral point
Y1 A  Y2 A Y1B  Y2 B Y1C  Y2C bank voltage, system neutral voltage and two phase voltages
I DIF  I A  IB  IC 0 – the latter two with very small multipliers.
6.1 Impact Y1 Aof  YInstrumentation
2 A nVT 2 Y1B  Y2 B Errors Y1C  onY2the
VOP ( A) V1 A  k A ˜ ˜ V2 A
C
Voltage Differential These four voltages are delivered by four VTs: (A,B,C,X) in case of
nVT 1Function implementation (7a) with internally derived system zero-sequence
n ACTUAL
For simplicitynIDEAL consider b 1  0.005
˜ b, applications ‘0.30 banks. The voltage; and (0,X,B,C) in case of implementation (7b) with externally
on grounded
operating signal in secondary volts is (equation (1c)): supplied system zero-sequence voltage. For the purpose of error
n analysis, each of the VTs shall be represented with its own ratio,
VOP ( A) V1 A  k A ˜ bn˜ VT 2 ˜ V2 A potentially slightly different than the nominal value.
VOP ( A) V1 A  k A ˜ VTn2VT˜ V
1 2A
nVT 1
n n
0 V1 A  k A ˜ VT 2 ˜ V2nA o k A ˜ VT 2 ˜ V2 A V1 A
VOP ( A) V1 A nkVTA 1˜ b ˜ VT 2 ˜ V2 A nVT 1
30 nVT 1 Fundamentals of Adaptive Protection of Large Capacitor Banks
VOP ( A) V1 A  b ˜ V1 A V1 A ˜ 1  b
nVT 2 nVT 2
VOP ( A) Vn1VTA1  k A ˜ b ˜ nVTVT1
1
˜ V2 A
nVT 1
VOP ( A) V1 A  b ˜ V1 A V1 A ˜ 1 nb
V0OP ( A)V V1 Akb ˜ V˜1 AnVT V
2 1A ˜ 1  b
˜V o k A˜ n ˜ V2 A V1 A
VT 2
n
0 V1 A  k A ˜nVTVT1 2 ˜ V2 A o k A ˜VT 1VT 2 ˜ V2 A V1 A
1A A 2A
n
VVOPOP A ) 1 
nVT nVT
Ab˜%˜ 100%
VT 12 ˜ V
2A o kA ˜
0 ( A)( V VT 12 ˜ V
1 A b1˜k

100 2A V1 A
V1 A n n
When deriving the system zero-sequence voltage internally the In other words, 1/3rd of the bus voltage “leaks” as a spurious
V
VOP1(AA) V1 A  b ˜ V1 A
VT 1 V1 A ˜ 1  b
VT 1
three
VOPOP( A)( Aphase Vvoltages
 b ˜ Vare added
V1 A as˜ 1vectors
 b – small errors could operating signal due to errors in the measurement. For example,
V
yield 1 1 A0.005‘significant
a) 1 relatively 0.30 ˜ 100% spurious
1A
0.72% system zero-sequence assume 0.3% magnitude error and 0.2 deg angle error. These
VVVOP 1 A  b ˜ Vderivative
Vfollowing V1 Aof0˜ 1equation
 b (7a) is useful:
11 b1˜100% ‘0.3 ˜ 100%
voltage.
OP
1A ( A A ) The
( A ))
1A errors in the A-phase voltage with all the other measurements
0.005 0.72% intact, i.e. with errors not adding and not canceling, would yield
VVVOP1 A( A)1
VOP 1 A n1VTX˜b1 ˜ 100
k AB % according to equation (37c) 0.18% of bus voltage as an error in the
k AC  ˜ Vx  nVT ˜ VA  k AB ˜ nVT ˜ VB  k AC ˜ nVT ˜ VC
VV OP (3A˜)nVT operating signal of this protection method.
1A 1  b ˜ 100%
VOPV1(AA) 1 1 n ˜ 1  k 0 k  ˜ V  n ˜ V  k (36a)
 1˜ 3ofVTX .005˜V‘A AB AB ˜ nVT ˜ VB  k AC ˜ nVT ˜ VC
V When using externally derived system zero-sequence voltage
VOPOP
For the ˜ 1nnVTX
purpose
3 ˜ V0x  nVTanalysis,
error  ˜ 100
˜V
0.n3VT the AC n%˜xV 0.72
B k-factors
VT % A
can be assumed to (equation (7b)), requirements for the bank and system neutral
˜
VT C
VVunity, 3 n
 
VT
1 A( A ) VT
beOP and 1 therefore:
 1  0 . 005 ‘ 0 . 3 0
˜ 100 % 0 . 72 % voltage measurements are relaxed, and the accuracy of
VV
n1VTX˜b1˜ 3˜ V0x .005
nVT ˜V‘
measurement of the two phase voltages becomes secondary
VOP
OP1 ( A )1
A
1 0n.VT3 ˜V˜B100
A 
0
 nVT% ˜VC 0.72%
VOP
V V 3 ˜ nVT 1
VTX ˜˜3
nnVTX 1 ˜Vkx AB nVTk ˜ VA˜ V nVTn˜ VB˜ V nVTk˜ VC˜ n ˜ V because of the low value of multipliers applied to the B and C
 k AC ˜ nVT ˜ VC
OP1 A 3 ˜ n B voltages.
3 ˜ nVT
AC x VT A AB VT
VT

VOP nVTX ˜ 1  k AB  k AC  ˜ Vx  nVT ˜ VA  k(36b)

nVTX ˜ 3 ˜Vx n1VT ˜ VA  nVT ˜ VB  nVT ˜VC
AB ˜ nVT ˜ VB  k AC ˜ nVT ˜ VC
Generally speaking the method is most impacted by the accuracy
3 ˜ 1n nVTXis˜ b˜ perfectly
13˜ Vk AB nbalanced An˜ Vand ˜ nVT of
˜ Vthe measurement of the system neutral voltage. This quantity is
VOP 1 the
Assume  11 ˜VT
above VB ˜3V˜C˜V 
k AC
˜ xn  n ˜an ˜VnA in˜kVAB
VT error the B k AC ˜ nVT ˜of VCthe method applied (internally, externally to
V
V
V b3˜1n
OP 3 3 ˜ n VT VTX
OP
OP VAnnVTX  bxn1 ˜ V˜VT
0V 
V ˜ V
VT 
V B n ˜ V
VT C derived regardless
measurement 3 ˜ nVT VT of the bank neutral voltage is added, represented
x VT A VT B VT C
the relay) out of three vectors each having significant magnitude
by the complex 1 number b: compared with the target sensitivity. Small magnitude and angle
VOP nVTX ˜ 3 ˜ Vx  nVT ˜ VA  nVT ˜ VB  nVT ˜ VC errors in sensing any of the three vectors would become significant
nVTX ˜ 33˜ V˜1xnVT nnVT ˜ V˜A3˜ V
1 nVT ˜ VnB ˜nVVT  ˜ VC ˜ V  n ˜ V
V
V n VTX˜ b ˜ 3 ˜ Vx   n VT ˜ V A  nnVT˜ V B n VT ˜ V for this sensitive protection function.
˜
OP C
OP 3
3 ˜ nVT n VT
VTX x VT A VT B VT C

6.3 Impact of Instrumentation Errors of the

VOP
1 1
b  1n˜VTX VA˜b V˜ 3B ˜VVx CnVT b˜VA1˜ Vn0VT ˜V B  nVT(36b) ˜VC Phase Current Balance Function
˜ 33˜3V˜x1nVT (36b): ˜ VA˜ b˜n3VT˜ V˜ V B  ˜VC n ˜ V  n ˜ V
OP
From
nVTX equation nVT nVT˜ V 1
V nVTX nVT
OP
3 ˜ nVT
x A VT B VT C
When nVTX ˜ 3 ˜V
VOP using a window-type  measure
CTx to nVT ˜ b ˜ Vthe
A  nVT ˜ VB current,
differential nVT ˜ VC
this method 3 ˜ nisVTquite immune to instrumentation errors. From
VTX ˜ 3 ˜ 
1 Vx nVT ˜ VA  nVT ˜ VB  nVT ˜ VC (36c)
VnOP b  1 ˜ VA  VB  VC  b  1 ˜ V0 equation (12) the method balances the differential current with
nVTX ˜ 33˜ Vx nVT ˜ VA  nVT ˜ VB  nVT ˜ VC a small fraction of the total bank current. Both signals are low:
˜3 ˜ V
Substituting (36c) into (36b) gives:
nformer
the VTX x  nVT of
because
˜ Vthe
B near-zero
nVT ˜ VC circulating
nVT ˜ VA current; the latter
VOP
1
b  1 ˜ VA  VB  VC  b  1 ˜ V0 because of the multiplier. As a result the errors are decimated
VOP
13
b  1 ˜ VA  VB  VC  b  1 ˜ V0 (36d)
when they1“leak” into the operating signal.
VOP b  1 ˜ VA
3 Analysis of3equation (12) yields the following expression the error
In other words, the error in the operating signal is proportional to analysis:
the system unbalance, with a small multiplier. As a result, errors
in the measurement of the bank neutral voltage are of secondary I OP b  1 ˜ I DIF b 1 ˜ k ˜ I (38)
importance. For example, assume a system unbalance (V0) of 3%
of bus nominal voltage, and a 5% magnitude and 1 deg angle For example, assume 2% of full bank current circulating in the
error for the neutral point transformer. Using equation (36d) one ' Vx CTV0(k =!0.02),
window P and 5% magnitude and 3 deg angle error in
concludes that this error introduces about 0.16% of bus nominal the phase CT. According to equation (38) the spurious operating
voltage as a spurious operating signal. signal will reach 0.14% of the total bank current.
' I DIF ! P
Bus VTs must be much more accurate to facilitate sensitive 6.4 Impact of Instrumentation Errors of the
protection. Assume, a phase A VT is now exposed to measurement
errors: Neutral1Current Balance Function
VOP
As explained
1  k AB  k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
1 3 in the previous section, the neutral and phase current
VOP nVTX ˜ 3 ˜Vx  nVT ˜ b ˜ VA  nVT ˜ VB  nVT ˜ VC balance methods are equivalent. The differential neutral current
3 ˜ nVT is compensated for inherent unbalance by all three currents (per
(37a) equation Vx but
VOP (18)), V0similarly to the phase current balance method
1 the multipliers for the phase currents are small. Therefore, equation
V OP ˜3
nVTX
From equation ˜ V ˜3 n˜ Vx ˜V nVTn˜VTb ˜˜V
˜ Vx  n(36b):
nVTX VAA  nVT ˜ VB  nVT ˜ VC (38) applies to this method, and yields the same results as to the
3 ˜ nVT VT B VT C
impact Vx  V0 errors.
VRESTof measurement
nVTX ˜3 ˜ Vx  nVT ˜ VB  nVT ˜ VC nVT ˜VA (37b)
1 Overall the relative insensitivity of the current balance methods to
nOP
V b 1 ˜ V
VTX ˜3 ˜ Vx  nVT ˜AVB  nVT ˜ VC nVT ˜ VA
3 (37b) into (37a) gives: instrumentation
0.2 pu‘00 errors0can .17be
puunderstood
‘50 by realizing only small
Substituting portions of the phase currents are used for compensation, while
the differential currents – if measured via window CTs – are not
I OP
V OP
b1 b1˜ 1I DIF
˜ VA b  1 ˜ k ˜ I (37c) exposed
V to0any significant
.2 pu ‘00  0errors.
.17 pu‘50 0.034 pu
3 OP

' Vx  V0 ! P
I OP b  1 ˜ I DIF b 1 ˜ k ˜ I VREST 0.2 pu‘00  0.17 pu‘50 0.37 pu
' I DIF ! P
' Vx  V0 ! P V0 0.05 pu‘00 , Vx 0.05 pu‘0 0  0.02 pu‘1800 0.07 pu‘00 .

V'OPI
1
1  k AB  k AC  ˜Vx  3Fundamentals
˜V0  VB ˜ 1 ofkAdaptive
AB   VC ˜ 1  k ACV
Protection of Large Capacitor 0Banks 31
DIF 3! P OP 0.07 pu‘0  0.05 pu‘0 0 0.02 pu

VOP 1Vx  V0 0 0
 VOP nVTX ˜ 3 ˜Vx  nVT ˜ b ˜ VA  nVT ˜ VB  nVT ˜ VC 3 ˜ nVT VT
3 ˜ nVT ' I DIF ! P n ˜3 ˜ V  n ˜ V  n ˜ V n ˜ V
VTX x VT B VT C VT A
nnVVTX
VTX ˜3˜3˜ V1 ˜x VbxnVT n 1˜ VVT˜B V˜ V
nVT
B ˜ VnC VT n˜VT VC˜ VA nVT ˜ VA
OP A
nVTX ˜3 ˜ Vx  nVT ˜ VB  nVT ˜ VC nVT ˜ VA 13 VOP 1 b  1 ˜ VA
VOP 1 1  k AB 3 k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
VOP 13b  1 ˜ VA
7. Comparison with Traditional Methods V VIn other 3b b111 ˜1the
words, In˜DIF A ˜ 3 ˜bV 1
Voperating ˜ k ˜is˜Iba ˜vectorial
signal difference of two
x  V and ˜˜ V  ˜ VC
I
 nnwith n n
OP
1 ˜31˜˜VIcope b ˜1berrors
˜˜kV˜AAI VT ˜ VC
b  1 ˜ VA V OP
voltages. 3 In order n to bbetter nVT avoid VBB penalizing
nVT
33 ˜˜ nnVT OP VTX
OP I VTX VT
VOP OP xDIF VT VT
3 either a given function is desensitized to account V
Traditionally, OP b V
Isensitivity
OP 1x˜an IVTDIF 0 b  1 ˜ krestraining
Voptimized ˜I signal can be created as
for inherent bank unbalances and instrumentation errors. Or, a follows:
I'OPVx bV0 1!˜'PIVDIF x  V0 ! P
b 1 ˜ k ˜ I
Vx ˜3 V0˜˜ V !x P VT ˜˜ V B  n ˜˜ V ˜ VA
historical value of the non-zero operating quantity is subtracted n n n
I
(D-changes)
OP b  1 ˜ I
before b  1 ˜ k ˜ I
comparing with a pickup threshold (P) resulting
DIF
' n
VVTX VTX ˜3 V V 
x x  VT n V0 B nVTV  VT VC
C nVT VT ˜ VA (40b)
REST
in the rate-of-change mode of operation: ' !
'IVI xDIF!1VP!0 P! P DIF
I P
'Note DIFthat1the above signal is not a classical restraint in the form of
' Vx  V0 ! P (neutral unbalance) (39a) 2 puor‘ 1 0b0n1 ˜ ˜V3A˜0V.17
VT ˜‘
Va0
VOP.OP
sum b  VTX 1 ˜ VofA1the x magnitudes.
npu b ˜5 V0A  nThis VT ˜ V B  nVTaffect˜VC sensitivity
V ˜3naverage would
OP
of
33 VT VOP  1  k  k  ˜ V  3 ˜ V  V B ˜ 1of AB two
 kthe  VC ˜ 1  k AC 
'OPthe
V I DIF1function.
Vvoltages.
1!1 Pk Instead
1  k  k 3  ˜Vtherestraint
 k 
AB
3˜ ˜
V V
AC

 V 3
is a vectorial sum
˜˜V
x
1   k V 
0
˜ 
V 1 
˜ 1k k   VC ˜ 1  k AC 
' I DIF ! P OP 3 AB AB 0AC x
AC x0 B AB C AC
(phase or neutral current unbalance) (39b) 0 0 B AB
VInOP ˜3 ˜ V
b 30 

. 2 n
1 pu˜ ˜I ‘
V 0 n  ˜0
bV . 
17 1 n ˜
pu k˜‘V ˜ 5I 0 .034 pu
b  1 V˜ better Vx bVthis  ˜ k ˜ I works, consider external
VTX
OP x VT B VT C VT A
I OP
To understand I DIF
OP
DIF how 0 1 approach

1 Vx1Vk0 AB  k AC  ˜ Vx  3 ˜ V0  VB ˜ 1  k AB   VC ˜ 1  k AC 
The rate-of-change approach improves sensitivity to some extent V fault and
OP V 1 x V
internal0 bank failure.
V b  1 ˜VVA
but has 1
1  k AB  k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1 VVAssume
limitations.

OP
VOP
'
OP
 .! REST‘fault 0V.017 pu‘50
0V0 xproducing
VOP ' k V
VACx  33 V
Van00external
!2P Ppu 20% of system 0.37zero-sequence
pu
First, it is3an approximation. As derived in section 4, the “leaking”
REST
Vx0  V0
voltage.
RESTx Assume further, the bank neutral point voltage is
values are proportional to present values of some other signals Vmeasured
I OPREST b  V01xV as˜xI DIF
0V.20Vpu 00b‘010 while˜ k ˜ I0the pu‘50 zero-sequence voltage
.17system
' ! ‘ ‘0accuracy  0.02ofpuinstrument ‘1800 0.07 pu‘00 .
OP 0
related
VOP Vx  V0 to the bank (example: differential current in the phase Vis I ‘
measured
0 . 05 P pu as 0 , V ‘ 0
due
0 . 05 to pu finite
balance method proportional to the total bank current). When
0
' . 0 I DIF ! P
2 pu 0
transformers and the relay, transients,
DIF
0 . 17 pu x 5
etc. If so, the function even
'perfectly
 V‘ 0 !
0P VOP 0.2 pu‘00  00.17 pu‘50 0.034 pu
0‘5
the currents do not change, the delta method works satisfactory. V2x pu
Vif0
VOP . 0
V
REST 0.2 pux‘0 0 0.17 
compensated 0V 0 . 17pu‘5
for
pu the bank inherent unbalance would
0.034 pu
But when the Vx currents
V0 change, such as during close-in external V see an operating 110.07 pusignal
 ‘00k of:0.05 pu‘0 0
 50BB ˜˜ 1 AB   VC ˜ 1  k AC 
VREST V  ˜  ˜˜V 0.02 pu
I DIF !3P1 VREST pu‘ 0  10. AB   VC ˜ 1  k AC 
faults, subtracting an old value will not compensate correctly. V'OP
OP 1 k
k AB  0k.2AC ˜ V
V 0xx 03
0
3.17 V pu ‘V V 37kkpu
OP AB AC 0
Time delay or other inhibit method may be needed to ride through ‘0‘0 0.. 5‘
0 0
V0 VREST 2 pu‘
.OP 300.02.20pupu 0
0 17
17 0pu .pu
17 ‘‘ 0
5pu 0 05.37 pu0.034 pu
such conditions.
0.2 pu‘0 0
0.17 pu‘5 0
VREST 1 0.07 pu0‘ .050pu
0
‘00.05 0
, Vxpu0‘ 00pu‘000.12 the0.pu pu‘180 0
0.07 pu‘00 .
Second, the nrate-of-change
1 approach will not provide for a V
V
V VIfOPused to
OP
V 1
Vpuxset
x 
trip 
‘
k
V V
0Vabove
AB0 0 k AC  ˜ V x  3 ˜without
instantaneously
0 V 0 VB a
.05
˜ 
0.05 pu‘0  00.020 pu‘180 0.07 pu‘0 .
0 1 
restraint
k AB  
02
0VC ˜function
1  k AC will
0
0OP 0to .05 0, Vx 0 this
VTX ˜ 3 ˜ Vx  nVT ˜ b ˜ V A  nVT ˜ VB  nVT ˜ VC have 3 be
V
sustained
OP
3 ˜ nVToperating0 signal. When 0the delta-t window slides V OPREST 0.0 2.pu 2 pu ‘‘ 0 00 0level. .017 .17 pupu ‘‘ 5 5 0.0034 .37pu pu
VOP into
entirely 0.2the ‘0 the
pufault,  0operating
.17 pu‘signal 5 will0.034
reset.pu
This creates a
VOP 0.07 pu‘00  0.05 pu‘0 0 0.02 pu
Calculate the V0proposed
V VOP 0V.07 x  V x ‘ 0V00 0.05restraining pu‘0 0 0.02 pu
signal:
x  V0 0 0
V
problem when time-delayed operation is assumed.
nVTX ˜3 ˜ Vx  nVT ˜ VB  nVT ˜ VC nVT ˜ VA
VOPREST
REST Vpu
1 V 0.050.pu 2Vpu ‘0‘0, 0V.07 x 0pu . 0‘.05
17 0pu0 pu‘
‘ 0.5
0 0
050pu‘ 000.037
.02pu 0pu.12‘ pu180
0
0.07 pu‘00 .
Methods for0.inherent‘0bank 0compensation
‘5 Vpresented puinB section 4 0
Vx  V0 0
˜of3the
REST REST
nVT ˜VC V0VREST
0 0
VREST
identify
2 pu nVTX ˜V  npu
.x17 VT ˜ b ˜and  VT ˜ V
0.n37
are 
REST
‘ 0.07 0 pu‘0  0.05 pu‘0 0 0.12 pu
‘
0
OP
1the 3 ˜ true cause
n unbalance, A as such accurate
Note
.
0.2 pu2 pu
that ‘the 0 0
0 applied0definition 0 . 17 pu
.17 pu‘of5the 5 0
restraint practically doubles
under
V b  1VT
system ˜balanced
VA conditions, minor unbalances, and major
OP
3
system events such0 as close-in faults. Their
the
V0 . 2 two
pu 0‘ . involved
05
0 00
. 07pu ‘
pu 0 0‘
signals.
. 0
17 0
,
0
puV 
‘ Assuming
50 0
. 05
0 . 05pu pu‘a0
0
‘0 0 
slope is00used
.02.02pu for ‘1800 it 0.07 pu‘00 .
putripping,
0 operating signals 0are OP
‘0 , Vx delayed 0.05 pu‘0 and pu‘180 ‘00.034/0.37
0 0 x
V0 0.05 pu 0.02 will putake . pu‘0=00 9.2% of slope to restrain the operation.
sustainable
n ˜3 ˜ V allowing  n ˜ Vtime  n ˜ V alarming n ˜V tripping with no 0.07V  0.17 pu‘500 0
I OPVTX b  1x˜ I DIFVT b B 1 ˜ kVT˜ I
restrictions. C VT A VOP 0 0..2 2 pu‘
‘0  0bank 0 0  0.17 0pu‘5
‘5 under 0..034
034 pu pu
Consider
VOP
OP 0.2an
puinternal .17 pufailure 0.0345%
pu of system unbalance
V
(system 0.007 .07
REST zero-sequence ‘‘
pupu 0000
0.005 pu
.05
voltage). ‘‘
pu 0 000further,
Assume 0.002 pupu
.12
the bank failure
pu‘0  0is.05 ‘0 condition
0 0 OP
Major
VOP system0.07 unbalance an pu
important 0.02 pu to consider.
pu0 ‘  pu‘
' Vxexample, changes the neutral point
0 voltage by 2%
0 of bus voltage at the
VOP  V0 !bPassume
1 V 0.0
For 1 ˜ VA a close in ground fault elevating both the 02 ..pu
2
2‘pu ‘0 0
0..17
 pu ‘5 pu ‘0.5
537 pu 00..37
37 pu
0 0 0 0
VREST
V REST 00.17 17 pu
angle
REST of 180 deg (worst case):
system zero-sequence voltage and the bank neutral point voltage.
The
3
compensated neutral
0 unbalance method
VREST 0.07 pu‘0 0  0.05 pu‘00 0.12 pu
' REST
V I !P 0 . 07 pu ‘ 0  0.05 pu‘00 is0based
.12 puon equation
‘00 ,0V00x , V0.05 pu‘0 0pu
 0.02 0pu‘180 0 pu‘000 . .07 pu‘0 00 .
05 pu‘ 05 pu‘‘00 0  pu‘
(5): DIF
V0 0.05 pupu
V 0.07180
I OP b  1 ˜ I DIF b 1 ˜ k ˜ I V0 00..05
0 ‘0 , Vx 00..05 x  00..02
02 pu ‘1800 0 0.07 pu‘0 .
The operating signal is:
VOP
1
1  k AB  k AC  ˜Vx  3 ˜V0  VB ˜ 1  k AB   VC ˜ 1  k AC  VOP 0.07 pu‘0  0.05 pu‘0 0
0
0.02 pu
pu‘
‘000  pu‘
0 0
3 V
VOP 00..07
07 pu  00..05
05 pu‘00 0 00..02
02 pu
pu
' Vx the
During
! P ground fault event, V
V0outlined and V assume significant
OP
The
VRESTrestraining ‘0 is:0.05 pu‘0
0.07 pusignal 0 0
0.12 pu
x will
x 0
V
values
OP V
and V0 balance perfectly as long as the relay uses proper
pu‘
‘00  pu‘
0 0
settings for the inherent bank unbalance compensation (k-values) V 00..07
07 pu  00..05
05 pu
0
‘000 0
0..12
12 pu
' I DIF
and the V!P
instrumentation errors are low enough compared with
VREST
REST pu
V REST x  V0
the applied setting. The other two voltage components are of Assume a 10% slope setting is applied. The ratio between the
secondary importance as they use small multipliers. operate and restraining signals is 0.02/0.12 = 17% allowing for
0.2 pu‘01  0.17 pu‘5    ˜ 
0 0
VOP
Simplifying 1  k AB  k AC ˜Vx  3 ˜V0  VB ˜ 1  k AB  VC 1
sensitive
k AC operation given the slope of 10%.
function:
3 one can write the following balance equation for this
Change in the voltage at 180 degrees is the worst case. Under the
VOP 0.2 pu‘0  0.17 pu‘5
0 0
0.034 pu
best case scenario one obtains 0.08pu of restraint, or 0.02/0.08 =
VOP Vx  V0 (40a) 25% of the operate-to-restraint ratio.
VREST 0.2 pu‘00  0.17 pu‘50 0.37 pu Careful application of restraint allows further improvement of
VREST Vx  V0 security while maintaining good sensitivity of the capacitor bank
V0 0.05 pu‘00 , Vx 0.05 pu‘0 0  0.02 pu‘1800 0.07 pu‘00 . protection functions.

0.2 pu0‘
VOP
00 0.17 pu‘50
.07 pu‘00  0.05 pu‘0 0 0.02 pu

VVREST
OP 00..07
2 pu‘000 0.005.17
pu‘
0
pu0‘
pu‘ 0 5
0.12 0
0
pu.034 pu

32 VREST 0.2 pu‘00  0.17 puFundamentals

‘50 0.37ofpuAdaptive Protection of Large Capacitor Banks

V 0.05 pu‘00 , V 0.05 pu‘0 0  0.02 pu‘1800 0.07 pu‘00 .

 8. Summary
This paper derives correct balance equations for short circuit
protection of shunt capacitor banks taking into account inherent
unbalances in the protected bank. Four methods are derived:
voltage differential, compensated neutral voltage unbalance, phase
current balance, and neutral current balance.
The paper analyses sensitivity of the developed methods and derives
practical equations for the amount of the operating signals given
the size of the bank failure. Also, impact of instrumentation errors
(instrument transformers and relays) is analyzed quantitatively
allowing one to optimize the secondary system design, and select
settings based on data.

As can be seen from key equations (1), (5), (11), and (18) the proper
way of balancing the bank (or banks) involves instantaneous
9. References
values of currents or voltages. Subtracting the residual unbalance [1] IEEE Std. C37.99-2000: “Guide for the Protection of Shunt
as a time-delayed signal (a historical, or a constant value), and Capacitor Banks”, June 2000.
responding to the delta changes does not constitute a proper,
sensitive and secure operating equation for protective relaying [2] Kasztenny B., Brunello G., Wester C.: “Capacitor Bank
purposes. Fundamentals and Protection”, Proceedings of the 56th Annual
Conference for Protective Relay Engineers, College Station, TX,
The methods presented in this paper compensate for both bank
April 8-11, 2003.
and system unbalances. Therefore they are insensitive to major
system events such as close-in faults. Presently used relaying [3] Capacitor Bank Protection and Control Relay, Instruction
techniques might misoperate on such system conditions, as they Manual, General Electric Publication, 2006.
typically disregard system unbalances and compensate for the
bank unbalance assuming no, or minor system unbalances.

The exact balance equations developed in this paper open a chance

to perform manual, or automated adjusting of the operating logic
in order to accommodate the inherent unbalance of the bank either
due to un-repaired failures, temperature or seasonal changes, or
changes due to removing, shorting, or repairing the cans. This can
be done as auto-setting, i.e. one time adjustment after the repair
and under user supervision, or as self-tuning, i.e. a continuous
tracing of the slightly changing capacitor reactances in order
to maintain optimum sensitivity to internal failures, and security
during system unbalances.

The voltage differential, phase and current balance methods are

subject to self-tuning under any conditions; the neutral voltage
unbalance is subject to self-tuning as long as the neutral point
voltage is above the measuring error level. When applied in the
self-tuning mode the methods continuously compensate for
temperature and seasonal changes, in a slow loop of modifying
their balancing coefficients based on actual values. Note that the
majority of the balancing coefficients developed in this paper are
ratios of impedances. As such they are already greatly insensitive
to temperature and seasonal changes.

If implemented in the self-tuning mode a given method shall still

monitor the total drift in the operating signal even if very slow,
and alarm if the amount of the drift signifies a danger of possible
future failure, or a series of minor failures that went undetected or
unattended to.

The involved balancing factors although in theory are complex

numbers, could be very well represented by real numbers (uneven
loss tangents of the capacitors in the bank, and errors of instrument
transformers cause small imaginary parts of the matching factors).
With the matching factors being real numbers, inherent unbalance
of a capacitor bank can be easily zeroed out in the protection
equations using only 1, 2 or a maximum of 3 coefficients. These
coefficients can be tuned by measurements, and simple engineering
calculations.

0925-v4
Fundamentals of Adaptive Protection of Large Capacitor Banks 33

View publication stats