Kasztenny Paper CapacitorBankProtection0

1
Capacitor Bank Unbalance Protection Calculations

and Sensitivity Analysis
Bogdan Kasztenny and Satish Samineni
Schweitzer Engineering Laboratories, Inc.
Abstract—In this paper, we introduce a method for performing Capacitor failures cause only slight changes in the bank
unbalance calculations for high-voltage capacitor banks. We voltages and currents. Therefore, these failures cannot be
consider all common bank configurations and fusing methods and detected based on the levels of voltages and currents but based
provide a direct equation for the operating signal of each of the
commonly used unbalance protection elements. This one-step on the unbalance in voltages and currents relative to a healthy
calculation method requires less data and is not only simpler but bank (hence the name unbalance protection). A distinct set of
also less prone to errors compared with multistep methods such as unbalance protection elements is available for each bank
the ones included in the IEEE Std C37.99 [1]. Our equations cover configuration.
both the fail-open and fail-short failure scenarios (fused, fuseless,
and temporarily repaired banks). The paper also derives
To set the unbalance protection elements, we must perform
equations for calculating the degree of internal overvoltage that a fault calculations for series failures inside the capacitor bank
failure puts on the healthy capacitor units in the bank. Next, we (capacitor units or elements failing open or short). Because
derive equations for the unbalance protection operating signals as capacitor bank equations are linear and there is no mutual
functions of the internal overvoltage. Our equations tie together coupling inside the bank, the underlying equations for the
the unbalance protection operating signals, the number of failed calculations are simple: the unit reactance ties the unit voltage
capacitor units, and the internal overvoltage caused by the failure.
Therefore, these equations provide a solid basis for setting the and current while Kirchhoff’s laws tie all voltages and currents
unbalance protection elements: we set the alarm thresholds to inside the bank. However, solving these underlying equations
detect a single (or partial) unit failure, and we set the trip by hand is tedious.
thresholds to trip when the internal overvoltage caused by the In general, we use short-circuit programs to set protection
failure threatens a cascading failure. The paper includes dozens of
elements, such as distance or overcurrent. However, the
equations because it covers a variety of bank configurations,
fusing methods, and unbalance protection elements. However, commonly used short-circuit programs do not include modules
when working on a particular bank configuration with a for unbalance calculations in capacitor banks. The IEEE Std
particular fusing method, the user is presented with just a handful C37.99 [1] advocates numerical multistep unbalance
of simple equations for direct one-step calculation of the calculations. Often, about a dozen calculation steps are required
unbalance protection operating signals. to obtain an unbalance protection element operating signal.
Some users develop their own short-circuit programs for
I. INTRODUCTION unbalance calculations in capacitor banks. Developing and
Shunt capacitor banks are assembled from capacitor units validating these specialized short-circuit programs is time
connected in parallel to form groups, groups connected in series consuming.
to form strings, and stings connected in parallel to form phases. This paper fills this void and provides equations for
In high-voltage applications, the phases are connected as unbalance calculations for common bank configurations, fusing
grounded or ungrounded single-wye, double-wye, or H-bridge methods, and unbalance protection elements. These equations
bank configurations. Capacitor units, in turn, are fabricated allow direct (one-step) calculation because they directly tie the
from capacitor elements encased together and connected in unbalance protection operating signals to the capacitor unit
parallel-series structures. Fuses may be applied to address arrangement parameters and the size (number of failed units),
failures of capacitor elements (internally fused banks) or units type (fail-open or fail-short), and location (above or below the
(externally fused banks). The method of fusing impacts how the bridge, left or right half of the bank, phase A, B, or C) of the
capacitor units are arranged in groups and strings. failure. Avoiding multistep calculations not only reduces time
Overall, capacitor banks are protected by a combination of and effort but also eliminates opportunities for errors.
fuses, which remove the failed unit or element, and protective The unbalance protection equations are remarkably simple.
relays, which alarm and trip the bank offline. From the We achieved this simplicity by working in per-unit values. It is
protective relay perspective, a capacitor failure and the apparent that an unbalance in capacitor bank voltages and
resulting fuse operation (if fuses are used) blend together into a currents is a result of a difference between the faulted and
single event to be detected (alarm for failures that can be healthy parts of the bank. As such, the per-unit voltage or
tolerated and trip for failures that may progress catastrophically current unbalance is independent of the absolute characteristics
because of the overvoltage condition that the failure puts on the of the faulted and healthy parts. We will show that the
remaining healthy capacitor units). unbalance in per unit is a fractional number: a ratio of two
2
integer numbers that depend on the number of failed capacitor Appendix B addresses the issues of multiple bank failures
units and the number of units, groups, and strings in the bank. and shows how to use our equations to leverage the principle of
We introduce the concept of an overvoltage factor. We superposition and perform unbalance calculations for multiple
define it as a ratio of the voltage elevated by the failure in the failures occurring sequentially in different parts of the bank.
most stressed part of the bank and the normal voltage in that Appendix C shows how to use our equations to perform
part. The overvoltage factor is a simple function of the bank unbalance calculations for capacitor element failures by
parameters and the size, type, and location of the failure. We treating them as partial capacitor unit failures.
use the overvoltage factor to better understand the impact of a
failure on the rest of the bank including the danger of breaching II. BANK CONFIGURATIONS AND UNBALANCE PROTECTION
the unit voltage rating and causing a cascading failure. More We consider the following common configurations of high-
importantly, we use the overvoltage factor to set unbalance voltage capacitor banks:
protection elements. We propose setting the alarm threshold to
• Grounded single-wye
detect a single unit failure (or even a fractional unit failure
because of the failure of some but not all capacitor elements • Ungrounded single-wye
inside the unit) and setting the trip threshold to trip before the • Grounded double-wye
internal overvoltage caused by the failure exceeds the unit
• Ungrounded double-wye
voltage rating and triggers a cascading failure.
• Grounded H-bridge
Unlike the numerical solutions (numbers in, numbers out),
our analytical equations directly tie the signals of interest to the • Ungrounded H-bridge
failure and bank parameters. As a result, these direct equations with the following fusing methods:
allow a multitude of applications and insights.
• Externally fused
The paper is organized as follows.
• Internally fused
Section II reviews the common high-voltage capacitor bank
• Fuseless
configurations and the applicable unbalance protection
elements. The section states the scaling and measuring polarity Advantages, drawbacks, and application considerations for
conventions for the unbalance protection elements. each of these bank configurations and fusing methods are out
of the scope of this paper.
Section III explains the capacitor unit arrangement that we
assumed when deriving our equations, explains the capacitor We consider the following unbalance protection elements:
failure scenarios, and introduces the per-unit system for the • Neutral overvoltage (59NT) for ungrounded banks
calculations. The capacitor unit arrangement assumed in this • Neutral overvoltage unbalance (59NU) for
paper covers most practical cases. The failures include the fail- ungrounded banks
open and fail-short scenarios. You can use these failure
scenarios to represent failures in the fused and fuseless banks • Voltage differential (87V) for grounded banks and
as well as temporary repairs in the bank (leaving the failed units ungrounded double banks
open or shorted until a proper repair can be performed). • Neutral overcurrent unbalance (60N) for grounded
Section IV uses an example to explain how we derived and and ungrounded double banks
validated the equations. • Phase overcurrent unbalance (60P) for grounded
Section V introduces the concept of the overvoltage factor and ungrounded double banks
and explains how to use it to set trip thresholds for the • Negative-sequence overcurrent (50Q/50QT) for
unbalance protection elements. grounded and ungrounded banks
Section VI provides several unbalance protection settings • Impedance (21C) for grounded banks
calculation examples to better explain and illustrate the new
Fig. 1 through Fig. 6 show the bank configurations and the
concepts.
applicable unbalance protection elements. The figures show the
Section VII gathers insights from the derived equations. It unbalance protection elements applicable to each bank
points to similarities between various unbalance protection configuration by denoting their ANSI device numbers (in
elements and compares their relative sensitivity. green) and the unbalance protection operating signal names and
Appendix A is a compilation of the derived equations. While measuring conventions (in blue). For simplicity, the figures do
the paper explains our methodology and teaches how to use the not show instrument transformers, unless required for clarity.
new information, Appendix A is a key output of our work. The The figures also show the failure location for which the derived
appendix is formatted for ease of use and reproduction in your unbalance equations directly apply. The term location refers to
project documentation. To appreciate the output of this paper, the faulted phase (A, B, or C), the faulted half of the double
consider looking at Appendix A before reading the remainder bank (left or right), and the fault position with respect to the tap
of the paper. or bridge (above or below).
3
Fig. 1. Grounded single-wye bank configuration and unbalance protection.
Fig. 2. Ungrounded single-wye bank configuration and unbalance

protection.
Fig. 4. Ungrounded double-wye bank configuration and unbalance
protection (a) and 60P protection and alternative connection of the 87V
protection (b).
Fig. 3. Grounded double-wye bank configuration and unbalance protection Fig. 5. Grounded H-bridge bank configuration and unbalance protection (a)
(a) and 60P protection and alternative connection of the 87V protection (b). and 60P and 87V protection (b).
4
where T is the per-unit tap position (the reactance between the

tap and the neutral point of the bank in per unit of the bank
reactance).
Using (3), the operating signal of the voltage differential
element is independent of the way of obtaining the reference
signal (T ⋅ VBUS). The ∆V87 signal obtained by measuring the
bus voltage (Fig. 1) and the V87 signal obtained by measuring
the voltage difference between taps in a double bank (Fig. 3b)
are equal:
V87 = ∆V87 (4)
As a result, we only need to provide one version of the 87V
unbalance equation, and the equation applies to both ways of
connecting the 87V element. If your 87V application is based
on a different tap matching convention than (3), then rescale the
87V operating signal obtained by using this paper.
Typically, the 87V tap position T is considerably less than
0.5 pu. For H-bridge banks (Fig. 5b), we use variable H for the
per-unit position of the bridge and assume that the 87V tap
position T is the same as the bridge position H. Typically, H is
about 0.5 pu.
We show the neutral (60N) and phase (60P) current
unbalance protection elements connected to low-ratio window
current transformers (CTs). These CTs measure the unbalance
Fig. 6. Ungrounded H-bridge bank configuration and unbalance protection
currents (difference between two currents) through magnetic
(a) and 60P protection (b). summation of the two fluxes. This measurement method is
To understand the scaling and polarity conventions for the considerably more accurate than the method of using two
protection input signals, consider the following points. differentially connected CTs to sum the secondary currents.
When using window CTs, the 60P operating signal balances
We define the operating signal for the 59NU element as
without errors during normal bank operation and external
follows:
faults, and the 60N operating signal balances without errors
V59NU = VN − 3V0 (1) during system ground faults.
where VN is the bank neutral-point voltage and 3V0 is the tripled The 50Q protection uses breaker CTs and measures the
zero-sequence bus voltage. negative-sequence current (3I2) at the bank terminals. For
double banks, the 50Q element measures the total bank current.
When only a small number of capacitor units fail, the bus
Of course, when set to detect capacitor unit failures, the 50Q
voltages remain balanced (3V0 = 0), and using (1) makes the
element uses time delay (50QT) for security during system
operating signals of the 59N/59NT and 59NU elements equal
faults. For simplicity and to avoid considerations regarding
during bank failures:
inverse-time or definite-time delay, we refer to this unbalance
V59NU = V59N = VN (2) protection element as 50Q in the context of its operating signal.
As a result, we only need to provide unbalance calculations We consider the 21C protection only for grounded banks. We
for the neutral-point voltage, and during capacitor unit failures, provide unbalance calculations (apparent reactance change) for
these calculations apply to both the 59NT and 59NU protection both the per-phase and per-string applications of the 21C
elements. Of course, the 59NT element uses time delay for protection. The latter approach offers significantly better
security during system faults, while the 59NU element can sensitivity but requires multiple CTs per phase (to reduce their
operate with little or no time delay. We refer to the operating voltage rating and cost, these CTs are installed in the bottom of
signal of the 59NT and 59NU elements that is common during the bank near the ground potential). For the grounded H-bridge
series faults as V59N. banks, we consider only the per-phase 21C element.
To simplify analysis and calculations for the two possible From the traditional protection perspective (alarm and trip
connections of the voltage differential protection element thresholds), phase angles of the unbalance protection operating
(compare Fig. 1 and Fig. 3b), we use the following 87V signals are irrelevant. In this paper, however, we pay attention
operating signal convention: to the instrument transformer polarity convention and the
unbalance protection operating signal phase angle. Treating the
∆V87 = VTAP − T ⋅ VBUS (3)
unbalance signals as phasors (considering both the magnitude
and angle) facilitates the application of the principle of
5
superposition in unbalance calculations for multiple bank power rating (kVAr). Often, the sum of R and P is also much
failures (Appendix B). greater than 1 to increase bank survivability during unit failures,
We use the following measurement and angle conventions: even though P or R can be as low as 1.
• Unbalance currents (I60N and I60P) are measured away The phase reactance (X) is the following function of the
from the failed half of the bank. capacitor unit reactance (XU) and the bank parameters S, P, and
R:
• Unbalance voltages are measured as a potential of the
failure area of the bank respective to ground (V59N), the S
X= ⋅X (5)
bus (∆V87 and V59NU), or the tap in the healthy half of P⋅R U
the double bank (V87). We assume a uniform bank unit arrangement as follows:
• Phase angles of all the unbalance protection operating • All phases (A, B, and C) are constructed the same.
signals are referenced to the faulted-phase voltage
• Both halves of a double bank are constructed the same.
angle.
• In H-bridge banks, the parts above and below the
• Apparent impedance is calculated as the phase-to-
bridge are constructed with the same capacitor units
ground capacitive reactance, i.e., a real positive value.
and the same P and R parameters, even though the
Phase angles of the unbalance protection operating signals number of groups in series may be different above and
are also useful when looking for the location of the failure below the bridge (the bridge position H may be
inside the bank: Which phase has a failure? Is the failure above different than 0.5 pu; S is the total number of groups in
or below the tap or bridge point? Is the failure in the left or right series).
half of the bank? See [2] for more information on fault locating
We recognize that when an 87V tap is created, the number,
in capacitor banks.
arrangement, and ratings of capacitor units are often different
in the parts above and below the 87V tap. Typically, the top
III. BANK UNIT ARRANGEMENT AND FAILURE SCENARIOS
(high voltage) part serves to provide reactive power and the
A. Capacitor Unit Arrangement bottom (low voltage) part acts only as a reference for the 87V
Fig. 7 shows a general arrangement of capacitor units in a protection element (Fig. 8). The reactive power of the low-
single phase (Φ) of a high-voltage capacitor bank. P capacitor voltage part is typically two orders of magnitude lower than the
units are connected in parallel to form a group. S groups are high-voltage part and can be neglected.
connected in series to form a string. R strings are connected in
parallel to form a phase. In fuseless banks, P = 1, and in fused
banks, R is typically 1 or 2.
Fig. 8. Nonuniform unit arrangement when creating an 87V tap.

The per-unit tap position T follows the voltage divider
principle. We calculate T as follows:
XBOTTOM
T= (6)
Fig. 7. General capacitor unit arrangement in the capacitor bank phase.
XTOP + XBOTTOM
Typically, S is much greater than 1 because the unit voltage where XTOP and XBOTTOM are the reactances of the top and
rating is a relatively small fraction of the system nominal bottom parts, respectively (use (5) to calculate these
voltage, and several units must be connected in series to match reactances).
the system voltage. The product of S, R, and P is much greater
than 1 to provide the desired rated power given the unit reactive
6
To remove this tap-related nonuniformity and use a uniform protect the capacitor unit from short circuits of the
unit arrangement in all our calculations, we introduce and use capacitor elements inside the case.
the equivalent bank parameters as follows: • Fuseless banks where no fuses are present and the
• The equivalent number of parallel units in a group, P, is capacitor failures are permanent short circuits.
the same as in the top (above the tap) part of the phase. In general, we need to consider two general categories of
• The equivalent number of parallel strings in a phase, R, failures: capacitors failing open (a short circuit blows the fuse
is the same as in the top (above the tap) part of the and the faulted capacitance becomes an open circuit) and
phase. capacitors failing short (no fuse is present and the faulted
• The equivalent number of groups in series in a string is: capacitance remains shorted out). We will use fail-open and
fail-short categories to represent temporary repairs in the bank.
STOP
SEQ = (7) In general, we can look at failures from the perspective of a
1−T
capacitor unit or a capacitor element. A failure of a single
Typically, (7) returns the equivalent value of S that is not an capacitor element, or even a few elements, does not necessarily
integer number. The noninteger value of S does not create any result in the loss of the entire capacitor unit. From this
problems in subsequent unbalance calculations but allows us to perspective, ability to perform unbalance calculations for a
neglect the differences in unit arrangement in the parts above partial unit failure is beneficial. It allows analyzing the
and below the tap. following cases:
Example 1 • A few capacitor elements fail short in a capacitor unit
A grounded single-wye capacitor bank is constructed with of an externally fused bank, but the current that the
one 580 kVAr, 17.5 kV capacitor unit per group, 8 groups in failed unit draws is below the fuse rated current, and
series in a string, and 6 strings in parallel per phase. The low- the unit does not fail open through the operation of the
voltage tap is created by using two 167 kVAr, 825 V capacitor external fuse.
units connected in parallel. We calculate the reactance values • A few capacitor elements fail open in an internally
of the top and bottom capacitor units from their rated power and fused capacitor unit (through the operation of internal
voltage and obtain 528.017 Ω and 4.0756 Ω, respectively. fuses), but the entire unit does not become an open
We use (5) and calculate the reactance values of the parts circuit yet.
above and below the tap: • A few capacitor elements fail short in a fuseless bank.
8 These failures could short the entire capacitor unit, less
XTOP = ⋅ 528.017 Ω = 704.023 Ω than one capacitor unit, or more than one capacitor
1⋅6
unit.
1
XBOTTOM = ⋅ 4.0756 Ω = 2.038 Ω Typically, capacitor element failures put the highest voltage
1⋅2
stress on the rest of the failed capacitor unit. As a result, the
We use (6) and calculate the per-unit tap position: failure progresses first inside the unit, resulting in a complete
2.038 Ω failure of the unit before the problem spreads to the rest of the
T= = 0.002886 pu bank.
704.023 Ω + 2.038 Ω
Ideally, to obtain better resolution of our calculations, we
We use (7) and calculate the equivalent number of series
should perform all derivations from the perspective of the
groups, S:
capacitor elements rather than the capacitor units. However,
8 such an approach would be complicated for at least the
SEQ = = 8.023
1 − 0.002886 following reasons: 1) for externally fused banks, we would see
Now we can neglect the nonuniformity of the capacitor unit combinations of the fail-short (element failures) and fail-open
arrangement and treat the bank as uniformly constructed with (blown unit fuse) scenarios, 2) in our calculations, we will need
the following equivalent parameters: P = 1, S = 8.023, R = 6, to consider fuses that protect various numbers of capacitors
and T = 0.002866 pu. such as external fuses protecting the entire capacitor unit and
internal fuses protecting one or a few capacitor elements, and
B. Failure Scenarios 3) the structure of the bank (how the capacitor units are
High-voltage shunt capacitor banks fall into the following arranged in the bank) and the structure of the unit (how the
three categories: capacitor elements are arranged in the capacitor unit) would
• Externally fused banks where fuses mounted outside create a very complicated overall structure of the bank to
each capacitor unit case protect the bank from short analyze.
circuits inside the units by disconnecting the shorted We solve this challenge as follows:
units. • We approach the calculations from the perspective of a
• Internally fused banks where multiple “simplified” capacitor unit and assume a failure (open or short) of
fuses are fabricated inside each capacitor case to the entire unit.
7
• We derive the unbalance equations by assuming an

integer number of units that failed open or short
(complete failure, not a partial failure).
• We devise a method to represent a partial capacitor unit
failure by using a fractional number of failed units.
This allows us to apply the unbalance equations
derived for capacitor unit failures to complex failure
scenarios with capacitor units and capacitor elements
failing in arbitrary combinations.
In this paper, we use two generic capacitor unit failure
categories: fail-open and fail-short.
The first category (fail-open) applies to both externally fused
banks, in which the fuse operation removes the entire unit, and
internally fused banks, when enough internal fuses operate to
remove the entire unit. A common assumption is that the fuse
operation is so fast that no unbalance protection element detects
the shorted state. Therefore, the unbalance protection is
Fig. 9. Fail-open unit failure model.
concerned with detecting the resulting fail-open state after the
fuse operation has removed the shorted unit.
The second category (fail-short) applies mainly to fuseless
banks in which a short circuit inside the unit prevails and the
unbalance protection is expected to detect the shorted unit(s).
This category also applies to externally fused banks if you want
to perform unbalance calculations at the capacitor element level
for scenarios where some elements are shorted but the unit
remains in service because the current of the failed unit is below
the rated current of the external fuse.
We also use the fail-open and fail-short failure categories to
represent banks that are left unrepaired or are temporarily
repaired after a failure. In this application, we use the unbalance
equations to calculate the inherent (standing) unbalance before
the next failure, as well as to calculate the unbalance protection
operating signals for a failure in an inherently unbalanced bank.
Addressing the inherent unbalance in capacitor bank protection
is out of the scope of this paper (see [3] for more details). Fig. 10. Fail-short unit failure model.
Fig. 9 shows our unit failure model for the fail-open scenario. In Section IV, we use variable F to represent the size of a
When the first unit in a group fails open, the other units in the failure that occurred in a single location, as shown in Fig. 9 and
same group are subjected to an overvoltage condition. As a Fig. 10. In Appendix B, we use the superposition principle to
result, it is most likely that the next unit failure will be in the perform unbalance calculations for failures at two or more
same group. Therefore, in the fail-open scenario, we assume different locations in the bank. Finally, in Appendix C, we use
that F units in the same group failed open (F < P). a fractional value of F to perform unbalance calculations for
Fig. 10 shows our unit failure model for the fail-short capacitor element failures.
scenario. When the first unit in a group fails short, the other
C. Per-Unit System
groups in the same string are subjected to an overvoltage
condition. As a result, it is most likely that the next failure will Because equations that tie the capacitor bank voltages and
be in a different group of the same string. Therefore, in the fail- currents are linear, we can select any unit convention for the
short scenario, we assume that F groups in the same string failed capacitor bank calculations. By working in per unit, we follow
short (F < S). a long tradition of short-circuit calculations in electric power
systems. We will see that performing calculations in per unit of
the bank nominal voltage and current yields simple results that
are applicable to banks of any voltage (VNOM) and reactive
power (QNOM) ratings.
8
We use the following base quantities. The base voltage is the current); write a set of high-level equations for a given bank
nominal phase-to-ground bank voltage: configuration, failure scenario, and unbalance protection
VNOM element; and solve these equations in the per-unit frame for the
VBASE = (8) unbalance protection element operating signal of interest.
√3
We will use the 59N element for an ungrounded single-wye
We selected the phase-to-ground voltage because the phase- bank as an example. In this case, the Kirchhoff’s current law is
to-ground voltage is applied to the phase reactance in the wye- the starting point (in an ungrounded bank, the phase currents
connected capacitor banks covered in this paper (we do not sum up to zero; see Fig. 2):
consider delta-connected banks).
VA − V59N VB − V59N VC − V59N
The base current is the nominal bank current: + + =0 (14)
−jXF −jX −jX
QNOM QNOM
IBASE = = (9) where XF and X are reactances of the faulted phase A, and the
√3 ⋅ VNOM 3 ⋅ VBASE
healthy phases B and C, respectively.
QNOM is the reactive power that the bank provides under A failure involving a few capacitor units does not change the
nominal voltage, not the sum of unit power ratings (kVAr) of bus voltages, and therefore, we can substitute:
all the capacitor units in the bank. In double-bank
VB = a2 ⋅ VA , VC = a ⋅ VA , a = 1∠120° (15)
configurations, the base current is the nominal current of the
entire bank, not half the bank. Inserting (15) into (14) and simplifying, we obtain:
Of course, the base reactance is: 1 2 1 a2 + a
V59N ⋅ � + � = VA ⋅ � + � (16)
VBASE XF X XF X
XBASE = (10)
IBASE
Of course, a2 + a = −1, and (16) becomes:
In this per-unit system, the healthy-phase reactance is 1 pu X + 2 ⋅ XF X − XF
and the faulted-phase reactance is slightly above or slightly V59N ⋅ � � = VA ⋅ � � (17)
X ⋅ XF X ⋅ XF
below 1 pu depending on if the units fail open or short.
We will see that the unbalance calculations performed in this We further rearrange (17) and obtain:
per-unit frame involve only the failure parameters (F, fail-open X − XF
or fail-short) and the bank parameters (P, S, R, and H or T). V59N = VA ⋅ � � (18)
X + 2 ⋅ XF
We perform unbalance calculations in the per-unit frame and
Switching to the per-unit frame, we write:
obtain the unbalance protection operating signals. When
converting the unbalance protection operating signals from per- X − XF
V59N(PU) ⋅ VBASE = VA(PU) ⋅ VBASE ⋅ � � (19)
unit values (PU) to secondary values (SEC), we apply the X + 2 ⋅ XF
following unit conversions.
where VA(PU) = 1∠0° (in all calculations, we assume the power
Voltage unbalance protection elements: system is at nominal conditions).
VBASE
VSEC = VPU ⋅ (11) Therefore, (19) becomes:
PTR
X − XF
Current unbalance protection elements: V59N(PU) = � � 1∠0° (20)
X + 2 ⋅ XF
IBASE
ISEC = IPU ⋅ (12) Because (20) contains only the ratio of reactances, these
CTR
reactances can be expressed in either ohms or per-unit values.
Reactance unbalance protection elements: We use Fig. 7 and calculate the healthy phase reactance, X. We
CTR use Fig. 9 (fail-open) and Fig. 10 (fail-short) to calculate the
XSEC = XPU ⋅ XBASE ⋅ (13) faulted-phase reactance, XF. In per unit, we obtain:
PTR
where PTR and CTR are, respectively, the ratios of the voltage
X(PU) = 1 (21)
and current transformers providing signals to the associated Fail-open:
protection element.
S ⋅ P − F ⋅ (S − 1)
XF(PU) = R ⋅ (22)
IV. CAPACITOR BANK UNBALANCE CALCULATIONS S ⋅ R ⋅ (P − F) + F ⋅ (R − 1)
In this section, we explain how we derived and validated the Fail-short:
unbalance calculation equations listed in Appendix A. S−F
XF(PU) = R ⋅ (23)
A. Principles S ⋅ R − F ⋅ (R − 1)
We start with first principles (Kirchhoff’s voltage and current
laws and the unit reactance that ties the unit voltage and
9
We insert (21) and (22) into (20) and obtain the V59N • We ran the symbolic math software to symbolically
operating signal for the fail-open scenario: solve the set of equations for a variable of interest.
F ⋅ 1∠180° • We formatted the output equations to publish them in
V59N(PU) = (24)
3 ⋅ S ⋅ P ⋅ R − F ⋅ (3 ⋅ S ⋅ R − 3 ⋅ R + 1) Appendix A in a consistent format.
We insert (21) and (23) into (20) and obtain the V59N C. Validation
operating signal for the fail-short scenario:
Our derivation process is based on the symbolic math
F ⋅ 1∠0° software and is therefore highly automated. Nonetheless, it is
V59N(PU) = (25)
3 ⋅ S ⋅ R − F ⋅ (3 ⋅ R − 1) still prone to human errors related to writing the basic
equations, feeding them into the software, reformatting the
For brevity, we omit the (PU) subscript from this point on
output equations, and typing the final equations into the text of
unless necessary for clarity. The phase angles in (24) and (25)
this paper.
are relative to the faulted-phase voltage angle. When a capacitor
unit fails open, the V59N signal is out of phase with the faulted- We have validated our equations as follows:
phase voltage. When a capacitor unit fails short, the V59N signal • We inspected each equation for expected results and
is in phase with the faulted-phase voltage. We will use this symmetry with other equations. For example, when
angle information when calculating the unbalance protection there is no failure (F = 0), all unbalance protection
operating signals for multiple failures in different parts of a operating signals shall be zero.
bank (Appendix B).
• For each bank configuration, we ran EMTP models for
Observe the following regarding the V59N signal: several banks (banks with different parameters) and
• When there is no failure (F = 0), the V59N operating compared the values from the numerical EMTP
signal is zero, as expected. solution to the values obtained by using equations in
Appendix A.
• The per-unit V59N signal is a simple function of the
bank parameters, S, P, and R, and does not depend on Using the above procedure, we verified all the equations as
the bank ratings (nominal voltage, nominal reactive entered in Appendix A of this paper.
power, and nominal frequency).
V. INTERNAL OVERVOLTAGE AND ITS APPLICATION IN
• When the units fail short, the V59N signal does not
SETTING THE UNBALANCE PROTECTION ELEMENTS
depend on the number of capacitor units in parallel (P).
A failure in a capacitor bank causes an internal overvoltage
B. Derivations inside the bank (see Fig. 9 and Fig. 10). This overvoltage may
In this paper, we consider six bank configurations and several cause more failures, which in turn creates even higher
unbalance protection elements for each configuration. We overvoltage, and eventually, leads to a cascading failure. We
consider the fail-open and fail-short failure scenarios for each propose using the overvoltage level to set trip thresholds of the
bank configuration and each protection element. As a result, we unbalance protection elements. The number of failed units is
derive several dozen equations to cover all the combinations. only a proxy of the internal overvoltage. The same number of
While the unbalance equations turned out to be simple (see failed units may stress different capacitor banks differently,
Appendix A), the derivation process to obtain these equations depending on the bank unit arrangement data and the capacitor
is tedious and therefore prone to errors. We have solved this unit voltage ratings. Tying the protection trip thresholds
challenge by using symbolic math software. Symbolic math directly to overvoltage rather than the number of failed units is
software does not solve equations for numerical values, but a simple and logical way to set trip thresholds.
instead, it manipulates equations symbolically to derive an A. Overvoltage Factor
equation for the sought variable. Symbolic math software does
We introduce an overvoltage factor, kOV, as the ratio of the
not output a numerical value for a numerical input. It outputs
present voltage across a capacitor unit and the voltage across
an equation based on the set of input equations. The software
the same unit when the nominal system voltage is applied to a
automates the derivation process and reduces the chance for
healthy bank. For example, when kOV = 1.15, the voltage across
human error.
the unit is 1.15 times higher than the normal voltage across the
We used the Symbolic Math Toolbox [4] in Matlab and same unit. When defining the overvoltage factor, we use the
followed this procedure: normal voltage across the unit rather than the unit voltage rating
• We wrote a set of fundamental equations for a given to keep the voltage rating out of the equations and avoid using
unbalance protection element, given bank two per-unit voltage bases.
configuration, and failure scenario. A capacitor unit can be safely operated when the sine wave
• To avoid human errors, we did not solve or simplify voltage magnitude across the unit is below 110 percent of the
these equations by hand (derivations (14) through (25) unit nameplate voltage rating and the voltage peak value is
are just for illustration). Instead, we fed the original below 120 percent [1]. Our unbalance calculations are
equations into the symbolic math software. concerned with bank failures rather than system harmonics and
voltage distortion. Therefore, the 110 percent limit applies.
10
The nameplate voltage rating must be higher than the 1/S kOV = 1.135. When three units fail open (F = 3), we obtain
fraction of the system nominal phase-to-ground voltage. For kOV = 1.217, which is above the permissible level of 1.192.
example, a capacitor unit may be rated at 105 percent of the 1/S We use (27) to calculate the number of units that, if failed
fraction of the bank nominal phase-to-ground voltage. If so, the open, would cause the maximum permissible overvoltage of
unit can operate with an overvoltage factor, kOV, of 1.192 and obtain F = 2.717 units (the bank can tolerate two
1.05 ⋅ 1.10 = 1.155 (the voltage across the unit can increase to failed units in the same group of 14, but not three). Ideally, the
115.5 percent of the normal value before the unit is in danger bank should be tripped when enough capacitor elements in the
of failing). third capacitor unit have failed and caused the maximum
We derive the overvoltage factor equations for all six bank permissible internal overvoltage.
configurations and the fail-open and fail-short scenarios and Remember that (26) and (27) are examples and only apply to
show how to use the overvoltage factor to set the trip thresholds the fail-open scenario in ungrounded single-wye banks.
of the unbalance protection elements. Appendix A lists the overvoltage factor equations for all bank
B. Calculating the Overvoltage Factor configurations and failure scenarios.
When F units fail open in a group (Fig. 9), the other units in C. Using Overvoltage to Set Unbalance Protection Elements
that group are exposed to overvoltage. When F groups fail short We combine the equation that ties the operating signal of an
in a string (Fig. 10), the other groups in the same string are unbalance protection element to the number of failed units:
exposed to overvoltage. We use the same approach as when
Operating Signal = f(F) (28)
deriving and validating equations for unbalance calculations
(see Subsections IV.B and IV.C) and obtain equations for the with the equation that ties the number of failed units to the
overvoltage factor for all capacitor bank configurations and the overvoltage factor:
two failure scenarios.
F = g(k OV ) (29)
For example, the overvoltage factor in an ungrounded single-
wye bank for the fail-open scenario is shown in (26). Also see and we obtain a direct relationship between the operating signal
Appendix A. and the overvoltage factor:
3⋅S⋅P⋅R Operating Signal = f�g(k OV )� (30)
k OV = (26)
3 ⋅ S ⋅ P ⋅ R − F ⋅ (3 ⋅ S ⋅ R − 3 ⋅ R + 1)
We use the same approach as when deriving and validating
Equation (26) tells us how much internal overvoltage occurs equations for unbalance calculations (see Subsections IV.B and
in an ungrounded single-wye bank when F units fail open. Of IV.C) and derive equations for the per-unit unbalance
course, when there is no failure (F = 0), the overvoltage factor protection operating signals as functions of the overvoltage
(26) equals 1, as expected. factor. For example, we obtain (31) for the 59N element
We solve (26) to see how many units would have to fail open operating signal for the ungrounded single-wye bank (see
to cause a particular overvoltage, and we obtain: Appendix A):
3⋅S⋅P⋅R k OV − 1 k OV − 1
F= ⋅ (27) V59N = (31)
3⋅S⋅R−3⋅R+1 k OV 3 ⋅ R ⋅ (S − 1) + 1
Equation (27) allows us to associate the voltage stress in the Equation (31) is very useful because it directly ties the per-
bank (kOV) with the failure size (F). Consider the following unit operating signal of the 59N unbalance protection element
example. with the internal overvoltage in the protected bank.
Example 2 Example 3
A 138 kV ungrounded single-wye capacitor bank has its Let us continue Example 2 and set the 59N element to trip
phase constructed with 14 externally fused capacitor units per when the internal overvoltage is at the maximum permissible
group, 4 groups in series in a string, and 1 string per phase level. We use 1.192 in (31) and obtain V59N = 0.0193 pu or
(S = 4, P = 14, and R = 1). The units are rated at 21.6 kV. 1.93 percent of the base voltage.
Under normal operating conditions, the voltage across the The VT ratio for the 59N element in this example is 332:1.
We use (11) and obtain the 59N trip threshold in secondary
units is 138 kV / (√3 ⋅ 4) = 19.92 kV. The unit rating is 21.6 kV
volts as follows:
or 108.4 percent of 19.92 kV. Therefore, the overvoltage factor
of safe operation is 1.1 ⋅ 1.084 = 1.192. The units can 138,000 V pri
V59N(SEC) = 0.0193 ⋅ = 4.63 V sec
continuously withstand a 19 percent overvoltage compared √3 ⋅ 332
with the healthy bank operated under nominal system voltage.
By using the threshold of 4.63 V sec, we ensure that the
Assume that a single unit fails open (F = 1). Using (26), we 59NT/59NU protection element operates when the failure
calculate the overvoltage factor and obtain kOV = 1.063. This causes the internal overvoltage to exceed 110 percent of the unit
value is less than 1.192 and the bank is not in danger of a voltage rating.
cascading failure. If two units fail open (F = 2), we obtain
11
Fig. 11 illustrates the relationships between the 59N A. Grounded Single-Wye Capacitor Bank
operating signal and the number of failed units, the overvoltage Table I shows the bank data. We used this bank in Example 1
factor and the number of failed units, and the combined and explained that this nonuniform bank (different unit
relationship between the 59N operating signal and the arrangement above and below the 87V tap) can be treated as a
overvoltage factor for Example 2 and Example 3. uniform bank with T = 0.002886 and the equivalent value of S,
S = 8.023.
TABLE I.
GROUNDED SINGLE-WYE CAPACITOR BANK DATA
Voltage (kV LL) 230
Bus Voltage PTR 2000:1
Bank Nominal Power (MVAr) 75.14
Breaker CTR 250:5
Units in a Group, P 1
Groups in a String, S 8
Strings in a Phase, R 6
Unit Power Rating, kVAr 580
Unit Voltage Rating, kV 17.5
Unit Type Fuseless
87V Tap See Example 1
87V Tap PTR 3.2:1
CTs in Each String No
The base units are as follows from (8), (9), and (10):
VBASE = 132.79 kV
IBASE = 188.62 A
ZBASE = 704.02 Ω
We use equations from Appendix A and apply the fail-short
Fig. 11. Direct relationship between the 59N operating signal and the
overvoltage factor (Example 2 and Example 3). scenario (fuseless bank) to calculate the unbalance protection
You may consider setting the 59N alarm threshold to alarm alarm and trip thresholds. We consider the per-phase 21C
when a single unit fails (F = 1), and you may consider setting protection element because no CTs are installed on the per-
the 59N trip threshold to trip when the internal overvoltage string basis. Fig. 12 and Fig. 13 plot the unbalance protection
approaches the highest permissible level (Fig. 11). A single unit element operating signals as functions of the number of failed
failure causes V59N = 0.00632 pu or 1.52 V sec (24). Therefore, units and as functions of the overvoltage factor, respectively.
you can set the 59N unbalance protection alarm threshold to
1.52 V sec and the trip threshold to 4.63 V sec.
Remember that (31) and Fig. 11 show an example and apply
only to the fail-open scenario in ungrounded single-wye banks.
Appendix A lists equations for all bank configurations,
unbalance protection elements, and failure scenarios.
VI. UNBALANCE PROTECTION SETTINGS CALCULATION

EXAMPLES
In this section, we illustrate the described concepts and show
their benefits by providing settings calculation examples for
several bank configurations and unbalance protection elements.
We set the unbalance protection elements to alarm when a
single unit fails and to trip when the internal overvoltage
exceeds 110 percent of the unit voltage rating. For simplicity
and uniformity, we do not apply setting margins or consider Fig. 12. Unbalance protection operating signals as functions of the number
alarming on partial unit failures. of failed units for the bank in Table I.
12
TABLE II.
UNGROUNDED SINGLE-WYE CAPACITOR BANK DATA
Voltage (kV LL) 138
Breaker CTR 100:5
Unit Type Externally fused
Neutral Voltage PTR 332:1
Note that the V59N and I2 operating signals are equal in per-
Fig. 13. Unbalance protection operating signals as functions of the
overvoltage factor for the bank in Table I.
unit values (see Appendix A). In other words, in per-unit
values, 3I2 and V59N have a 3:1 relationship. If we plotted I2
We set the unbalance protection functions to alarm when a
instead of 3I2, the two curves in Fig. 14 (and Fig. 15) would
single unit fails (F = 1) and to trip when an overvoltage due to
overlap.
unit failures reaches 110 percent of the unit rating of 17.5 kV.
The normal voltage across each unit is 230 kV / (√3 ⋅ 8.023) =
16.55 kV. Therefore, we calculate the trip thresholds by using
kOV = 1.1 ⋅ 17.5 kV / 16.55 kV = 1.163.
Alarm thresholds (F = 1; see Fig. 12):
3I2 = 0.023732 pu
∆X = −0.023181 pu
∆V87 = 6.8513 ⋅ 10-5 pu
Trip thresholds (kOV = 1.163; see Fig. 13):
3I2 = 0.027167 pu
∆X = −0.026448 pu
∆V87 = 7.8434 ⋅ 10-5 pu
For illustration, let us convert the per-unit trip thresholds to
secondary units by using (11), (12), and (13): Fig. 14. Unbalance protection operating signals as functions of the number
of failed units for the bank in Table II.
3I2 = 0.027167 pu ⋅ 188.62 A / 50 = 0.102 A sec
∆X = −0.026448 pu ⋅ 704.02 Ω ⋅ 50 / 2000 = –0.4655 Ω sec
∆V87 = 7.8434 ⋅ 10-5 pu ⋅ 132.79 kV / 3.2 = 3.26 V sec
The reactance change of −0.4655 Ω sec is a per-phase change
from the nominal value of 17.601 Ω sec. The 21C element can
use a blocking characteristic in the form of a circle centered at
−j17.601 Ω sec and having a blocking radius of 0.4655 Ω sec.
B. Ungrounded Single-Wye Capacitor Bank
Table II shows the bank data. We used this bank in
Examples 2 and 3. The base units are as follows from (8) and
(9):
VBASE = 79.674 kV
IBASE = 29.885 A
We use equations from Appendix A and apply the fail-open overvoltage factor for the bank in Table II.
scenario (fused bank) to calculate the unbalance protection
We set the unbalance protection functions to alarm when a
alarm and trip thresholds. Fig. 14 and Fig. 15 plot the unbalance
protection operating signals as functions of the number of failed single unit fails (F = 1) and to trip when an overvoltage due to
unit failures reaches 110 percent of the unit rating of 21.6 kV.
units and as functions of the overvoltage factor, respectively.
13
The normal voltage across each unit is 138 kV / (√3 ⋅ 4) = phase. Fig. 16 and Fig. 17 plot the unbalance protection
19.918 kV. Therefore, we calculate the trip thresholds by using operating signals as functions of the number of failed units and
kOV = 1.1 ⋅ 21.6 kV / 19.918 kV = 1.192. as functions of the overvoltage factor, respectively.
Alarm thresholds (F = 1; see Fig. 14):
3I2 = 0.018987 pu
V59N = 0.0063291 pu
Trip thresholds (kOV = 1.192; see Fig. 15):
3I2 = 0.0576 pu
V59N = 0.0192 pu
secondary units by using (11) and (12):
3I2 = 0.0576 pu ⋅ 29.885 A / 20 = 0.0861 A sec
V59N = 0.0192 pu ⋅ 79.674 kV / 332 = 4.607 V sec
C. Grounded Double-Wye Capacitor Bank
Table III shows the bank data.
TABLE III. Fig. 16. Unbalance protection operating signals as functions of the number
GROUNDED DOUBLE-WYE CAPACITOR BANK DATA of failed units for the bank in Table III.
Voltage (kV LL) 138

Bank Nominal Power (MVAr) 100
Breaker CTR 2000:5
Unit Type Externally fused
87V Tap One bottom group
87V Tap PTR 120:1
60N CTR 20:5
The base units are as follows from (8), (9), and (10): overvoltage factor for the bank in Table III.
VBASE = 79.674 kV Alarm thresholds (F = 1; see Fig. 16):
IBASE = 418.37 A 3I2 = I60N = 0.005882 pu
ZBASE = 190.44 Ω = 63.48 Ω sec ∆X = 0.0059172 pu
The bank has a uniform unit arrangement, and we calculate ∆V87 = 0.0019608 pu
the tap position as follows: Trip thresholds (kOV = 1.143; see Fig. 17):
T = 1/6 = 0.1667 pu 3I2 = I60N = 0.0143 pu
The units can safely withstand a voltage of: ∆X = 0.014507 pu
1.1 ⋅ 13.8 kV = 15.18 kV
∆V87 = 0.0047667 pu
Normally, the units operate under a voltage of:
79.674 kV / 6 = 13.279 kV secondary units by using (11), (12), and (13):
Therefore, the overvoltage factor when selecting the trip 3I2 = 0.0143 pu ⋅ 418.37 A / 400 = 0.0150 A sec
threshold is:
I60N = 0.0143 pu ⋅ 418.37 A / 4 = 1.50 A sec
kOV = 15.18 kV / 13.279 kV = 1.143
∆X = 0.014507 pu ⋅ 190.44 Ω ⋅ 400 / 1200 = 0.921 Ω sec
We use equations from Appendix A and apply the fail-open
scenario (fused bank) to calculate the unbalance protection ∆V87 = 0.0047667 pu ⋅ 79.674 kV / 120 = 3.1649 V sec
alarm and trip thresholds. We consider the per-phase 21C The reactance change of 0.921 Ω sec is a change from the
protection element because the bank has only one string per nominal value of 63.48 Ω sec.
14
D. Grounded H-Bridge Capacitor Bank

Table IV shows the bank data.
TABLE IV.
GROUNDED H-BRIDGE CAPACITOR BANK DATA
Voltage (kV LL) 345
Breaker CTR 1000:5
Unit Type Fuseless
Bridge Position (pu) 0.5 overvoltage factor for the bank in Table IV.
87V Tap PTR 1500:1 Alarm thresholds (F = 1; see Fig. 18):
60P CTR 5:5 3I2 = I60N = 0.012346 pu
60N CTR 5:5 I60P = 0.012346 pu (same as 3I2 because H = 0.5)
∆X = −0.012195 pu
The base units are as follows from (8), (9), and (10):
VBASE = 199.186 kV ∆V87 = 0.0030488 pu
IBASE = 219.042 A Trip thresholds (kOV = 1.210; see Fig. 19):
3I2 = I60N = 0.03 pu
ZBASE = 909.348 Ω = 60.623 Ω sec
I60P = 0.03 pu (same as 3I2 because H = 0.5)
The units can safely withstand a voltage of:
1.1 ⋅ 9.96 kV = 10.956 kV ∆X = −0.029126 pu
Normally, the units operate under a voltage of: ∆V87 = 0.0072816 pu

199.186 kV / 22 = 9.0539 kV For illustration, let us convert the per-unit trip thresholds to
secondary units by using (11), (12), and (13):
Therefore, the overvoltage factor when selecting the trip
threshold is: 3I2 = 0.03 pu ⋅ 219.042 A / 200 = 0.0329 A sec
kOV = 10.956 kV / 9.0539 kV = 1.210 I60N = 0.03 pu ⋅ 219.042 A / 1 = 6.571 A sec
We use equations from Appendix A and apply the fail-short I60P = 0.03 pu ⋅ 219.042 A / 1 = 6.571 A sec
scenario (fuseless bank) to calculate the unbalance protection ∆X = 0.029126 pu ⋅ 909.348 Ω ⋅ 200 / 3000 = 1.766 Ω sec
alarm and trip thresholds. Fig. 18 and Fig. 19 plot the unbalance ∆V87 = 0.0072816 pu ⋅ 199.186 kV / 1500 = 0.9669 V sec
protection operating signals as functions of the number of failed
units and as functions of the overvoltage factor, respectively. VII. INSIGHTS INTO THE UNBALANCE PROTECTION ELEMENTS
In this section, we provide insights into the unbalance
protection elements based on the equations in Appendix A.
A. Some Unbalance Protection Element Operating Signals
Are Identical or Proportional to One Another
For a series failure in one location, the per-unit operating
signals of some unbalance protection elements are identical or
proportional to one another. As a result, these protection
elements can be considered as redundant elements rather than
complementary elements that mutually cover their weak spots.
Consider the following three examples:
Any ungrounded bank (per-unit values):
1
V59N = −j ⋅ 3I2 (32)
3
Fig. 18. Unbalance protection operating signals as functions of the number
of failed units for the bank in Table IV.
15
Grounded double-wye bank: element or the 50QT element. Of course, using the per-string
I60P = I60N = 3I2 (33) rather than per-phase 21C protection element would
dramatically improve the 21C sensitivity and favor it over the
Ungrounded double-wye bank: 50QT element (see Subsection VII.G).
1 In general, the product of the P and R parameters is
I60P = 3I2 , I60N = ⋅ 3I2 (34)
2 significantly higher than 1 (for survivability and to obtain the
required power rating, we have multiple parallel units in a
We draw the following insights and observations regarding
group or parallel strings in a phase, or both). Therefore, all
the unbalance protection elements that have the same or
unbalance protection elements (for a given bank configuration
proportional per-unit operating signals.
and for a given failure scenario) have operating signals that are
Even though these elements have the same theoretical either identical, proportional to one another, or very similar. Of
sensitivity, their practical applications differ. For example, CT course, the elements’ settings in secondary units differ, but this
ratio and angle errors reduce the accuracy of measuring the 3I2 is because of the physical nature of the operating signals
operating signal. A capacitor bank protective relay may see a (voltage, current, and reactance), instrument transformer ratios,
standing 3I2 signal for a healthy bank because of small CT and tap-matching.
errors. By contrast, there will be no standing V59N, I60N, and I60P
signals (assuming that window CTs are used to measure the C. All Unbalance Equations Have a Common Format
unbalance currents). As a result, the 59NT/59NU, 60N, and 60P The unbalance equations display a significant similarity
protection elements can be set with more sensitivity than the across all bank configurations and protection elements.
50QT element. Consider the following examples:
These elements have the same theorical sensitivity, but only Ungrounded single-wye bank, fail-open:
for series failures in a single location. Their sensitivities differ F
for shunt failures, such as phase-to-ground and phase-to-phase |V59N | = (38)
3 ⋅ S ⋅ P ⋅ R − F ⋅ (3 ⋅ S ⋅ R − 3 ⋅ R + 1)
faults, and for multiple series failures at different locations (see
Appendix B). As a result, enabling unbalance protection Grounded double-wye bank, fail-short:
elements that have identical or proportional operating signals is
still justified. These elements are redundant for a single 1 F
|3I2 | = ⋅ (39)
capacitor unit failure but complementary for multiple capacitor 2 S⋅R−F⋅R
unit failures and for phase-to-ground and phase-to-phase faults.
Grounded H-bridge, fail-open (G = 1 – H for failures above
B. All Unbalance Protection Element Operating Signals Are the bridge, and G = H for failures below the bridge):
Similar H⋅G⋅F
As an example, consider the grounded single-wye bank |∆V87 | = (40)
4 ⋅ S ⋅ R ⋅ P ⋅ G − F ⋅ (4 ⋅ S ⋅ R ⋅ G − 4 ⋅ R + 2)
configuration under the fail-open scenario and examine the 3I2
operating signal and the per-phase ∆X operating signal: All these equations have the same general format:
F k⋅F
|3I2 | = (35) Y= (41)
S ⋅ R ⋅ P − F ⋅ (S ⋅ R − R) A−B⋅F
F where factors k, A, and B are positive numbers that are specific

|∆X| = (36) to each protection element and depend on the capacitor unit
(S
S ⋅ R ⋅ P − F ⋅ ⋅ R − R + 𝟏𝟏)
arrangement and bank configuration, and Y stands for a general
The two equations are almost identical and differ only by 1 unbalance protection element operating signal.
in the multiplier following the number of failed units, F, in the Equation (41) teaches us that when the failure size is small,
denominator (shown in red). The number of series groups in a the unbalance protection operating signals are near-linear
string, S, is much greater than 1. Therefore, the values of functions of the number of failed elements, F. When the failure
R ⋅ (S – 1) and R ⋅ (S – 1) + 1 are almost identical, and we size increases, the denominator in (41) decreases and the
conclude that in per unit: operating signal (41) becomes a steeper function of the failure
size (see Fig. 12, Fig. 14, Fig. 16, and Fig. 18).
|∆X| ≅ |3I2 | (37)
We can use (41) to evaluate sensitivity of the unbalance
In other words, the per-phase 21C element and the 50QT protection elements across bank configurations, especially for
element have almost identical theoretical sensitivities to small failures (alarm application). We want to know the
capacitor unit failures. The 21C element uses one voltage increase in the operating signal (∆Y) for an increase in the
transformer and one current transformer per phase and the failure size (∆F). We use the derivative of (41) and write:
50QT element uses three current transformers. Therefore, the
d k⋅F
operating signals of the 21C and 50QT elements may ∆Y ≅ � � ⋅ ∆F (42)
experience different measurement errors. Based on these errors, dF A − B ⋅ F
it may be possible to apply more sensitive settings to the 21C
16
The slope in (42) tells us how sensitive a given protection units before the element trips the bank offline or a cascading
element is. We calculate the slope from (41) and obtain: failure occurs.
d k⋅F k⋅A From Appendix A, we see that for all bank configurations,
� �= (43) factor C in (45) is inversely proportional to the product of S and
dF A − B ⋅ F (A − B ⋅ F)2
R in the fail-open scenario and inversely proportional to R in
When the bank is healthy, then F = 0. Therefore, when the the fail-short scenario. In other words, for a given bank
failure develops (F increases from 0 to 1), the slope is: configuration and fusing method, all applicable unbalance
protection elements have approximately the same theoretical
d k⋅F k
� � = (44) relationship to the level of internal overvoltage when
dF A − B ⋅ F F=0 A considered in per unit of the bank nominal values.
When more capacitor units fail, the numerator in (43) D. Approximation of the Unbalance
decreases and the slope increases (sensitivity increases). From the previous subsection, we know that the unbalance
Therefore, (44) is a conservative estimate of the slope and protection operating signals are proportional to the following
therefore sensitivity. The unbalance protection sensitivity is expressions:
proportional to k and inversely proportional to A.
Fail-open:
To illustrate this point, let us use the grounded H-bridge bank
with the fail-open scenario and consider the 50QT and 87V 1
∆Y ≈ ⋅ ∆F (46)
elements. From Appendix A, the 50QT element operating S⋅P⋅R
signal has an initial slope of G/(2SPRG) or 1/(2SPR). The 87V
Fail-short:
element operating signal has the initial slope of HG/(2SPRG)
or H/(2SPR). The two elements differ in slope only by the fixed 1
∆Y ≈ ⋅ ∆F (47)
multiplier of H. The VT ratio for the tap voltage effectively S⋅R
eliminates this fixed difference, and the two protection
elements have effectively identical sensitivities to capacitor The number of groups in series, S, and the capacitor unit rated
unit failures in a single location. That sensitivity is inversely voltage, VU, must be selected to satisfy the following condition
proportional to the product of the bank unit arrangement (for simplicity, we neglect the margin for sustained system
parameters S, R, and P. overvoltages):
From Appendix A, we see that for all bank configurations, VNOM
VU ≥ (48)
the k/A factor in (44) is inversely proportional to the product of √3 ⋅ S
S, P, and R in the fail-open scenario and inversely proportional The rated bank power is:
to the product of S and R in the fail-short scenario. In other
words, for a given bank configuration and fusing method, all VNOM 2
Q NOM = (49)
applicable unbalance protection elements have approximately X
the same theoretical sensitivity to the number of failed capacitor We use (5) for the phase reactance X and use (48) for the
units when considered in per unit of the bank nominal values. nominal voltage and rewrite (49) as follows:
The practical sensitivities differ because of different instrument
transformer errors and ratios. VU 2
Q NOM = 3 ⋅ S ⋅ P ⋅ R ⋅ (50)
When we consider the unbalance protection element XU
operating signals as functions of the internal overvoltage caused The squared unit voltage rating divided by the unit reactance
by the failure (trip application), we conclude that the common is the unit rated power (QU), therefore, we can write:
equation format is as follows:
Q NOM
Y ≅ C ⋅ (k OV − 1) (45) S⋅P⋅R= (51)
3 ⋅ QU
where the multiplier C is specific to each protection element, We insert (51) into (46) and (47) and obtain the unbalance
depends on the bank unit arrangement parameters, and depends protection sensitivity estimates by using the bank nominal
slightly on the value of the overvoltage factor, kOV (strictly voltage and power and the capacitor unit rated voltage and
speaking, C is not a constant). power:
Equation (45) teaches us that the relationship between the Fail-open:
unbalance protection operating signals and the level of internal
QU
overvoltage caused by the failure is near linear. For example, if ∆Y ≈ �3 ⋅ � ⋅ ∆F (52)
the overvoltage doubles, such as when kOV increases from 1.10 QNOM
to 1.20 (0.2 = 2 ⋅ 0.1), the operating signal doubles as well Fail-short:
(approximately). Adjusting an unbalance protection trip
threshold up or down proportionally increases or decreases the QU
∆Y ≈ �3 ⋅ ⋅ P� ⋅ ∆F (53)
additional voltage that the failure puts on the healthy capacitor QNOM
17
The product of QU and P is the rated power of a capacitor Example 5

group. The product of the number of failed units (∆F) and the Let us use the fuseless, grounded single-wye bank from
unit power or group power is the power lost due to the failure. Subsection VI.A and recalculate the 3I2 operating signal (as an
Keeping these observations in mind, we can consolidate (52) example) for different combinations of P and R while keeping
and (53) into a general approximation of the unbalance the product of the two parameters at 6, as in the original bank
protection operating signal as follows: data. Table V shows the results assuming two failures: a failure
∆Q F Number of Lost Units of a single unit (alarm) and a larger failure that leads to the
∆Y ≈ 3 ⋅ ≈ (54) maximum permissible internal overvoltage (trip).
QNOM Number of Units per Phase
TABLE V.
where ∆QF is the reactive power lost because of the failure 3I2 VALUES FOR DIFFERENT P-R COMBINATIONS
(fused units became an open circuit and unfused units shorted 3I2 (pu)
the entire group). P R
F = 1 (alarm) kOV = 1.163 (trip)
You can use (54) to obtain a very quick estimate of the
unbalance protection operating signals (in per unit) for a small 1 6 0.0237 0.0272
number of failed units based on the simple count of units lost 2 3 0.0475 0.0543
relative to the total number of units in a phase. 3 2 0.0712 0.0815
The value of ∆Y in (54) refers to the 3I2 current. Calculate 6 1 0.1424 0.1630
other unbalance quantities by factoring in the proportion
between 3I2 and other signals (see Appendix A, Table V shows a clear increase in the operating signal when
Subsection VII.A and Subsection VII.B). the capacitor units are moved from strings to groups. This
tradeoff between P and R can be explained as follows. The
Example 4
parallel strings (larger R) obfuscate the unbalance caused by the
Let us consider the grounded double-wye bank from failure in the faulted string and make it more difficult to detect.
Subsection VI.C and use (54) to get the first approximation of At the same time, shorting a larger group of units (larger P)
the per-unit unbalance. The bank has 15 capacitor units in a increases the unbalance and makes it easier to detect.
group, 6 groups in a string, and 1 string per phase. The bank is
Of course, considerations other than unbalance protection
a double bank with two phases in parallel. The total number of
sensitivity apply when making a tradeoff between the number
units per phase is 15 ⋅ 6 ⋅ 1 ⋅ 2 = 180. The bank is externally
of units in a group and the number of strings in a phase [5].
fused, and therefore, a single unit failure removes one unit from
the bank (a shorted unit would remove the entire group, i.e., Let us now look at unbalance protection sensitivity as a
15 units). Using (54), we calculate the per-unit unbalance for function of internal overvoltage. Using (45) and obtaining
F = 1 as 1 / 180 = 0.00555 pu. factor C from Appendix A, we write the following
approximations:
For comparison, the exact values we calculated were: 3I2 =
I60N = 0.005882 pu and ∆X = 0.0059172 pu. The approximation Fail-open:
of (54) provided an accuracy with better than a 6 percent error. 1
∆Y ≈ ⋅ ∆k OV (55)
E. Optimizing Unit Arrangement for More Sensitive S⋅R
Unbalance Protection Fail-short:
Can we select the S, R, and P parameters to increase the
1
protection sensitivity? Equation (54) teaches us that the ∆Y ≈ ⋅ ∆k OV (56)
sensitivity to capacitor unit failures depends on the number of R
units lost due to the failure and the total number of units in the Equations (55) and (56) show that the unbalance protection
bank. When F units fail open in a fused bank, the number of lost sensitivity as a function of internal overvoltage does not depend
units is F. When F units fail short in a fuseless bank, the number on P and increases when R decreases. Therefore, for any given
of lost units is F ∙ P (F groups are lost). product of P and R required to obtain the target rated power
The above observation applied to the approximation of (54) given the value of S, it is advantageous to select a smaller R.
leads to the following conclusions: This means that it is better for protection sensitivity to have
• Sensitivity of unbalance protection for fused capacitor more units in a group than more strings in a phase (see
banks does not depend on how the capacitor units are Example 5). For example, if the preferred product of P and R is
divided between the groups and strings. 6, it is better to have 3 units in a group (P = 3) and 2 strings in
a phase (R = 2) than the other way around.
• Sensitivity of fuseless banks increases when more
capacitor units are placed in groups while the number F. Optimizing the Tap and Bridge Position for More
of strings is reduced. Sensitive Unbalance Protection
Another consideration related to optimizing protection
sensitivity is the placement of the 87V tap (T) or bridge (H). To
evaluate the impact of T and H on protection sensitivity, we can
18
use equations from Appendix A, consider the bank unit halves of the bank and reduces the sensitivity of the per-string
arrangement parameters and the failure size as constants, and reactance measurement. In ohms, the per-string reactance
treat T (or H) as a variable. We can then determine the value of change is the same for the single- and double-wye bank
T (or H) that maximizes the unbalance protection operating configurations. In the per-unit frame, these reactance changes
signal for failures above and below the tap or bridge point. differ by a factor of 2 because the base reactance is twice as low
Consider the grounded H-bridge bank configuration. The in the double-wye bank as in the single-wye bank.
unbalance protection operating signals are proportional to G. Consider the grounded single-wye bank and the fail-open
G = 1 – H for failures above the bridge and G = H for failures scenario. The sensitivities of the per-phase and per-string
below the bridge. We want to maximize the smaller of the two reactance changes are as follows:
operating signals (the signal for failures above and below the 1
bridge). Fig. 20 plots the (1 – H) expression (the multiplier for ∆XPHASE ≈ ⋅ ∆F (57)
S⋅P⋅R
failures above the bridge) and the H expression (the multiplier
for failures below the bridge). When H is small (below 0.5), the R
sensitivity for failures below the bridge is lower than for ∆XSTRING ≈ ⋅ ∆F (58)
S⋅P
failures above the bridge. This is because when located lower,
the bridge is better able to short the units located below the Comparing the two sensitivities (slopes in (57) and (58)), we
bridge and by doing so, obfuscates failures below the bridge. conclude that the per-string reactance change is R2-fold higher
When H is large (above 0.5), the sensitivity for failures above than the per-phase reactance change. The R2-fold difference is
the bridge is lower than for failures below the bridge. This is in per unit of the bank reactance or in secondary ohms. As a
because when located higher, the bridge is better able to short percentage of the string reactance, the per-string reactance
the units located above the bridge and by doing so, obfuscates change is R-fold higher than the percentage change of the phase
failures above the bridge. We obtain the best sensitivity when reactance.
the bridge is in the middle (H = 0.5), as is typically the case.
H. Accounting for Bus Voltage Fluctuations
For H = 0.5, the sensitivity is equal for failures above and below
the bridge. Because a capacitor bank is a linear circuit, the unbalance
protection operating signals are directly proportional to the
terminal (bus) voltage. Our equations provide the unbalance
protection operating signals in per unit. You can use a simple
multiplier to account for changes in the terminal voltage.
To illustrate this point, let us consider the alarm and trip
protection applications.
In the alarm application, the intent is to detect a single unit
failure (or a partial unit failure). The operating signal during the
failure decreases if the terminal voltage decreases. Therefore, it
is good practice to set the alarm threshold at 0.8 times the
Fig. 20. Relative sensitivity for failures above and below the bridge as a calculated value to account for a possible 20 percent reduction
function of bridge position for the H-bridge bank configuration. in the terminal voltage during stressed system conditions. This
We obtain a similar result for the position of the tap (T) and margin allows dependable pickup of the alarm function, and it
the 87V protection element. When T is small (tap located low), prevents deassertion of the alarm (if not latched) when the
the 87V element has a much higher sensitivity to failures below voltage decreases after the alarm is already set.
the tap than for failures above the tap. When T is large (tap In the trip application, the intent is to trip the capacitor bank
located high), the 87V element has a much higher sensitivity to before the internal overvoltage caused by the failure breaches
failures above the tap than for failures below the tap. The two the unit voltage rating. From this perspective, the system
sensitivities are equal, and therefore, the overall 87V sensitivity overvoltage (the terminal voltage is higher than nominal) and
is at a maximum when the tap is in the middle (T = 0.5). A the internal overvoltage (there is a failure in the bank that
higher T value increases the voltage level at the tap and requires distributes the voltage unequally among the healthy capacitor
using VTs of a higher voltage rating. Therefore, the value of T units) compound. However, because the unbalance protection
is often kept low (the tap is installed low, and the VT ratio is operating signals are proportional to the terminal voltage, there
selected low to boost the secondary voltage signal). However, is no need for additional margin to account for terminal voltage
the 87V sensitivity can be improved by locating the tap closer fluctuations. Any given failure (number of failed units, F)
to the midpoint. results in an internal overvoltage that is proportional to the
G. Per-String vs. Per-Phase Impedance Protection terminal voltage and therefore the system overvoltage. A failure
Appendix A lists the per-phase and per-string reactance at a time when the voltage is nominal may result in an internal
changes for the grounded single- and double-wye banks. We do overvoltage that is permissible and the unbalance protection
not derive the per-string reactance change for the grounded H- may not trip, but the same failure at a time when the system
bank because the bridge connects the strings in the left and right voltage is higher than nominal may result in an internal
19
overvoltage that is not permissible, and the unbalance 15 percent, you may set the trip threshold for the ∆X value at
protection may trip. 1/1.15 (or 0.87) times the threshold calculated with the
Example 6 Appendix A equations. This is because when the terminal
voltage is elevated, it takes a smaller impedance unbalance to
Let us continue the ungrounded single-wye bank example
cause the same internal overvoltage in the bank.
from Subsection VI.B. We consider a failure of two units
(F = 2) and inspect the 59NT/59NU element operating signal
VIII. CONCLUSIONS
during normal voltage and during a system overvoltage. The
element is set at 4.607 V sec with the intent to trip the bank In this paper, we derived equations for unbalance calculations
when the internal voltage reaches the maximum permissible for six common high-voltage capacitor bank configurations.
value of 110 percent of the unit voltage rating (kOV = 1.192). Expressed in per unit, the unbalance protection operating
Under nominal voltage, the failure of two units results in a signals are simple functions of the bank unit arrangement
V59N operating signal of: parameters: number of units in a group, number of groups in a
string, and number of strings in a phase. The per-unit unbalance
V59N = 0.013514 pu = 3.243 V sec protection operating signals are independent of many other
This operating signal is below the tripping threshold of factors, such as unit reactance and power rating, frequency, and
4.607 V sec and the 59NT/59NU element restrains. Under these bank nominal voltage and power.
conditions (two units failed under nominal system voltage), the For completeness, our equations cover both the fail-open and
overvoltage factor is 1.135, which is below the maximum fail-short failure scenarios for all bank configurations. Not all
permissible value of 1.192. fusing methods are applied to all bank configurations.
Assume first that after the failure, the capacitor bank is However, to calculate unbalance under pre-existing failures and
exposed to a 10 percent system overvoltage (terminal voltage is temporary bank repairs, we need both the fail-open and fail-
1.1 times nominal). Now, the 59N operating signal is short scenarios for all bank configurations regardless of the
proportionally higher because of the compounding of the failure fusing method.
and the higher terminal voltage: These equations allow analyzing bank failures as well as
V59N = 1.1 ⋅ 3.243 V sec = 3.567 V sec setting the unbalance protection elements to alarm and trip.
The 59N operating signal is still below the tripping threshold In the context of capacitor bank analysis, the equations allow
of 4.607 V sec and the 59NT/59NU element does not operate. unbalance calculations in a quick and convenient way. Our
Under these conditions (two units failed and the system voltage equations apply to a single failure. You can use the principle of
increased to 110 percent of nominal), the overvoltage factor is superposition to perform calculations for multiple failures
1 + 1.1 ⋅ (1.135 – 1) = 1.149. This value is below the maximum occurring sequentially in different parts of the bank
permissible value of 1.192, and therefore, the 59NT/59NU (Appendix B). Following this approach, you can also analyze
element restrains as desired. The system voltage multiplier (1.1) the impact of inherent bank unbalance due to past failures or
applies to the fractional value of the kOV factor because the temporary repairs.
unbalance protection operating signals are proportional to We introduced the concept of an overvoltage factor: the ratio
kOV – 1 not kOV. It would take a 42 percent overvoltage of the voltage across a capacitor unit elevated because of the
(4.607 / 3.243 = 1.42) to make the 59NT/59NU element failure and the normal voltage across that unit. The paper
operate. The smaller the failure size, the lower the sensitivity of derives overvoltage factor equations for the common bank
the unbalance protection elements to system overvoltage. configurations under both the fail-open and fail-short failure
Without a failure (or under equalizing failures), the scenarios. These equations allow evaluating the voltage stress
unbalance protection elements do not detect the system that a failure puts on the healthy capacitor units in the bank.
overvoltage at all. They detect system overvoltage only when In the context of setting the bank protection elements, we
the failure size causes an operating signal that is close to the trip propose to set the unbalance protection alarm thresholds to
threshold under nominal voltage. Therefore, unbalance detect a single (or partial) capacitor unit failure. We propose to
protection does not substitute for bank overvoltage protection, set the unbalance protection trip thresholds by using the
and you need to apply phase overvoltage elements (59P) to overvoltage factor. The trip is issued when the overvoltage
protect against system overvoltages without and with caused by the failure reaches the voltage rating of the units in
harmonics [1] [5]. At the same time, no additional margin is the bank. The provided equations allow calculating the trip
needed when setting the unbalance protection trip thresholds to thresholds directly from the highest permissible overvoltage
account for overvoltage because the increased terminal voltage factor given the bank unit arrangement parameters.
changes the internal overvoltage and the unbalance protection
We examined the derived equations and gathered the
operating signals proportionally.
following important conclusions:
The 21C element operates irrespective of the terminal
voltage, which is a slight disadvantage when considering • The unbalance equations are similar to one another in
voltage fluctuations. As a result, the 21C element shall be set terms of both the expression and the values they return
with an additional margin to account for system overvoltages. for typical bank unit arrangement data.
For example, if the maximum expected overvoltage is
20
• There is a near-linear relationship between the • Some but not all fuses have blown in internally fused
unbalance protection operating signals and the banks.
overvoltage level caused by the failure. • Some but not all capacitor elements have shorted in the
• There is a near-linear relationship between the fuseless banks and externally fused banks before the
unbalance protection operating signals and the failure external (unit) fuse blows.
size. Our method for calculating the unbalance for capacitor
• For small failures, the unbalance protection operating element failures is very simple because it decouples the bank
signals for all methods are almost identical or arrangement (units in groups, groups in strings, strings in
proportional to one another and depend on the amount phases, phase connections, and grounding) from the
of reactive power lost because of the failure relative to arrangement of capacitor elements inside capacitor units.
the nominal power in one phase. The paper uses many numerical examples to explain and
We also showed (Appendix C) how to perform unbalance illustrate the content. Appendix A is a compilation of all the
calculations for capacitor element failures (partial unit failures), unbalance equations that you can print and use when
including the following cases: performing protection calculations for high-voltage capacitor
banks.
21
IX. APPENDIX A. CAPACITOR BANK UNBALANCE where:

CALCULATION EQUATIONS XU
This appendix lists equations for unbalance calculations for 𝛼𝛼 = (A.3)
XUF
the six capacitor bank configurations in six separate tables.
Each table includes equations for the fail-open and fail-short XU is the reactance of a healthy capacitor unit
scenarios and shows the bank configuration, measurements, XUF is the reactance of a partially failed capacitor unit
and unbalance protection elements marked with the ANSI Refer to the body of the paper for more details about the
device numbers. The unbalance equations are in per unit. When following topics:
appropriate, the equations include phase angles that you can use • Tap-matching of the 59NU and 87V operating signals
to apply the principle of superposition when performing (Section II).
calculations for simultaneous failures at two or more locations.
• Equivalencing banks with nonuniform unit
The equations use the following symbols: arrangement above and below the 87V tap point
P Number of capacitor units in parallel in a group (Section III.A).
S Number of groups in series in a string • Base values for the per-unit voltage, current, and
R Number of strings in parallel in a phase reactance and conversion to secondary values
F Number of failed capacitor units (see Fig. 9 and (Section III.C).
Fig. 10 for the fail-open and fail-short scenarios,
Following are typical use cases for the material in this
respectively)
appendix:
T Per-unit tap for the 87V element (per-unit reactance
of the bottom part) • Assume the size and location of a unit failure (F) and
calculate the unbalance protection operating signals.
H Per-unit position of the bridge in an H-bridge bank
kOV Overvoltage factor (the ratio of the voltage across a • Assume the size and location of multiple unit failures
healthy capacitor unit during a failure and during and calculate the unbalance protection operating
nominal conditions, kOV = 1 when there is no failure) signals by using the principle of superposition.
Remember that the overvoltage factor, kOV, refers to the • Assume the size and location of a unit failure (F) and
normal voltage, and not the unit rating (VU). The unit rating is calculate the alarm threshold for the unbalance
typically higher than the normal voltage and the units are protection elements.
designed to withstand 110 percent of the rated value. Typically, • Assume the size and location of a unit failure (F) and
when calculating a trip threshold, you will use kOV of 1.1 times calculate the overvoltage factor for the healthy units
the unit voltage rating divided by the unit voltage during (kOV) because of the failure.
nominal system conditions: • Assume the maximum permissible overvoltage factor
VU (kOV) and calculate the number of failed units (F) that
k OV = 1.1 ⋅ would cause that level of overvoltage.
VNOM (A.1)
� �
√3 ⋅ S • Assume the maximum permissible overvoltage factor
(kOV) and calculate the trip threshold for the unbalance
where VNOM is the nominal (phase-to-phase) bank voltage.
protection elements.
Before using the equations, ensure that the capacitor unit
• Use the angle of the operating signals in the post-fault
arrangement in your capacitor bank conforms to the
analysis to identify the faulted phase and location
assumptions of this paper (see Fig. 7).
(above or below the tap/bridge, left or right half of the
When performing unbalance calculations for capacitor bank, and the faulted phase).
element failures (partial capacitor unit failures), calculate and
To keep the equations in a single table on one page, the
use the fractional failure size as follows (see Appendix C):
equations omit the multiplication sign (⋅). For example, because
Fail-Open Fail-Short there is no variable called SP, SP is a product of the two
𝛼𝛼 − 1 (A.2) variables (S ⋅ P).
FFRAC = 1 − 𝛼𝛼 FFRAC =
𝛼𝛼 − 1 + P
22
GROUNDED SINGLE-WYE CAPACITOR BANK

Fail-Open Fail-Short
SP S
k OV = k OV =
SP − F(S − 1) S−F
SP k OV − 1 k OV − 1
F= F=S
S − 1 k OV k OV
F F
3I2 = 1∠−90° 3I2 = 1∠90°
SRP − F(SR − R) SR − FR
k OV − 1 k OV − 1
3I2 = 1∠−90° 3I2 = 1∠90°
R(S − 1) R
F F
∆XPHASE = ∆XPHASE = −
SRP − F(SR − R + 1) SR − F(R − 1)
k OV − 1 k OV − 1
R(S − 1) + 1 − k OV R + k OV − 1
F R
∆XSTRING = R ∆XSTRING = − F
SP − FS S
k OV − 1 k OV − 1
∆XSTRING = R ∆XSTRING = −R
S − k OV k OV
TF TF
∆V87 = 1∠0° ∆V87 = 1∠180°
R−T R−T
SRP − F �SR − SR − F
1 − T� 1−T
T(1 − T)(k OV − 1) T(1 − T)(k OV − 1)
∆V87 = 1∠0° ∆V87 = 1∠180°
R(S − 1) − T(SR − (R − 1)k OV − 1) R − T�(R − 1)k OV + 1�
Notes:
All values are in per unit.
∆XPHASE and ∆XSTRING are both in per unit of the bank reactance.
Voltage and current phase angles are relative to the faulted-phase voltage.
∆87V differential signal uses bus voltage scaled down to the tap voltage (VTAP – T⋅VBUS).
XBOTTOM
T=
XTOP + XBOTTOM
For nonhomogeneous banks (different unit arrangement above and below the tap), P and R
are parameters of the top part and S is an equivalent value, as follows:
STOP
S=
1−T
23
UNGROUNDED SINGLE-WYE CAPACITOR BANK

3SPR 3SR
k OV = k OV =
3SPR − F(3SR − 3R + 1) 3SR − F(3R − 1)
3SPR k OV − 1 3SR k OV − 1
F= F=
3SR − 3R + 1 k OV 3R − 1 k OV
3F 3F
3I2 = 1∠−90° 3I2 = 1∠90°
3SPR − F(3SR − 3R + 1) 3SR − F(3R − 1)
k OV − 1 k OV − 1
3I2 = 3 1∠−90° 3I2 = 3 1∠90°
3R(S − 1) + 1 3R − 1
F F
V59N = 1∠180° V59N = 1∠0°
3SPR − F(3SR − 3R + 1) 3SR − F(3R − 1)
k OV − 1 k OV − 1
V59N = 1∠180° V59N = 1∠0°
3R(S − 1) + 1 3R − 1
1
V59N = −j 3I2
3
Notes:
Voltage and current phase angles are relative to the faulted-phase voltage.
For internal faults, the 59NU operating signal is the same as the 59N operating signal.
24
GROUNDED DOUBLE-WYE CAPACITOR BANK

SP S
k OV = k OV =
SP − F(S − 1) S−F
SP k OV − 1 k OV − 1
F= F=S
S − 1 k OV k OV
1 F 1 F
3I2 = 1∠−90° 3I2 = 1∠90°
2 SPR − F(SR − R) 2 SR − FR
k OV − 1 k OV − 1
3I2 = 1∠−90° 3I2 = 1∠90°
2R(S − 1) 2R
F F
2SRP − F(2SR − 2R + 1) 2SR − F(2R − 1)
k OV − 1 k OV − 1
2R(S − 1) + 1 − k OV 2R + k OV − 1
F 2R
∆XSTRING = 2R ∆XSTRING = − F
SP − FS S
k OV − 1 k OV − 1
∆XSTRING = 2R ∆XSTRING = −2R
S − k OV k OV
TF TF
∆V87 = V87 = 1∠0° ∆V87 = V87 = 1∠180°
R−T R−T
SRP − F �SR − SR − F
1 − T� 1−T
T(1 − T)(k OV − 1) T(1 − T)(k OV − 1)
∆V87 = V87 = 1∠0° ∆V87 = V87 = 1∠180°
R(S − 1) − T(SR − (R − 1)k OV − 1) R − T�(R − 1)k OV + 1�
I60P = I60N = 3I2
Notes:
∆XPHASE and ∆XSTRING are both in per unit of the
bank reactance.
Voltage and current phase angles are relative to the
faulted-phase voltage.
∆87V differential signal uses bus voltage scaled
down to the tap voltage (VTAP – T⋅VBUS).
XBOTTOM
T=
XTOP + XBOTTOM
For nonhomogeneous banks (different unit
arrangement above and below the tap), P and R are
parameters of the top part and S is an equivalent
value, as follows:
STOP
S=
1−T
25
UNGROUNDED DOUBLE-WYE CAPACITOR BANK

6SPR 6SR
k OV = k OV =
6SPR − F(6SR − 6R + 1) 6SR − F(6R − 1)
6SPR k OV − 1 6SR k OV − 1
F= F=
6SR − 6R + 1 k OV 6R − 1 k OV
3F 3F
3I2 = 1∠−90° 3I2 = 1∠90°
6SPR − F(6SR − 6R + 1) 6SR − F(6R − 1)
k OV − 1 k OV − 1
3I2 = 3 1∠−90° 3I2 = 3 1∠90°
6R(S − 1) + 1 6R − 1
F F
V59N = 1∠180° V59N = 1∠0°
6SPR − F(6SR − 6R + 1) 6SR − F(6R − 1)
k OV − 1 k OV − 1
V59N = 1∠180° V59N = 1∠0°
6R(S − 1) + 1 6R − 1
6TF 6TF
V87 = 1∠0° V87 = 1∠180°
6R − 5T − 1 6R − 5T − 1
6SRP − F �6SR −
1−T � 6SR − F
1−T
6T(1 − T)(k OV − 1) 6T(1 − T)(k OV − 1)
V87 = 1∠0° V87 = 1∠180°
6R(S − 1) − T(6SR − 6(R − 1)k OV − 5) + 1 6R − T(6(R − 1)k OV + 5) − 1
I60P = 2I60N = 3I2

1
I60N = 3I2
2
1
V59N = −j 3I2
3
Notes:
XBOTTOM
T=
XTOP + XBOTTOM
For nonhomogeneous banks (different unit arrangement
above and below the tap), P and R are parameters of the
top part and S is an equivalent value, as follows:
STOP
S=
1−T
26
GROUNDED H-BRIDGE CAPACITOR BANK

2SPRG 2SRG
k OV = k OV =
2SPRG − F(2SRG − 2R + 1 − G) 2SRG − F(2R − 1 + G)
2SPRG k OV − 1 2SRG k OV − 1
F= F=
2SRG − 2R + 1 − G k OV 2R − 1 + G k OV
GF GF
3I2 = 1∠−90° 3I2 = 1∠90°
2SPRG − F(2SRG − 2R + 1 − G) 2SRG − F(2R − 1 + G)
G(k OV − 1) G (k OV − 1)
3I2 = 1∠−90° 3I2 = 1∠90°
2SRG − 2R + 1 − G 2R − 1 + G
GF GF
2SRPG − F(2SRG − 2R + 1) 2SRG − F(2R − 1)
G(k OV − 1) G (k OV − 1)
2SRG − 2R + 1 − Gk OV 2R + Gk OV − 1
HGF HGF
∆V87 = 1∠180° ∆V87 = 1∠0°
4SRPG − F(4SRG − 4R + 2) 4SRG − F(4R − 2)
HG(k OV − 1) HG (k OV − 1)
∆V87 = 1∠180° ∆V87 = 1∠0°
2(2SRG − 2R + 1 − Gk OV ) 2(2R + Gk OV − 1)
1 F 1 F
I60P = 1∠−90° I60P = 1∠90°
2 2SPRG − F(2SRG − 2R + 1 − G) 2 2SRG − F(2R − 1 + G)
1 k OV − 1 1 k OV − 1
I60P = 1∠−90° I60P = 1∠90°
2 2SRG − 2R + 1 − G 2 2R − 1 + G
1
I60P = 3I
2G 2
0 for failures above the bridge
I60N = �
2I60P for failures below the bridge
Notes:
H is the per-unit position of the bridge relative to
the neutral point.
For failures above the bridge, use G = 1 – H. For
failures below the bridge, use G = H and add 180°
to the phase angle. The 180° angle shift does not
apply to 3I2 and ∆X.
27
UNGROUNDED H-BRIDGE CAPACITOR BANK

6SPRG 6SRG
k OV = k OV =
6SPRG − F(6SRG − 6R + 3 − 2G) 6SRG − F(6R − 3 + 2G)
6SPRG k OV − 1 6SRG k OV − 1
F= F=
6SRG − 6R + 3 − 2G k OV 6R − 3 + 2G k OV
3GF 3GF
3I2 = 1∠−90° 3I2 = 1∠90°
3G(k OV − 1) 3G(k OV − 1)
3I2 = 1∠−90° 3I2 = 1∠90°
6SRG − 6R + 3 − 2G 6R − 3 + 2G
GF GF
V59N = 1∠180° V59N = 1∠0°
G(k OV − 1) G(k OV − 1)
V59N = 1∠180° V59N = 1∠0°
6SRG − 6R + 3 − 2G 6R − 3 + 2G
3F 3F
I60P = 1∠−90° I60P = 1∠90°
3(k OV − 1) 3(k OV − 1)
I60P = 1∠−90° I60P = 1∠90°
6SRG − 6R + 3 − 2G 6R − 3 + 2G
1
I60P = 3I
G 2
I60N = I60P
1
V59N = −j 3I2
3
Notes:
H is the per-unit position of the bridge relative to the
neutral point.
For failures above the bridge, use G = 1 – H. For
failures below the bridge, use G = H and add 180° to the
phase angle. The 180° angle shift does not apply to 3I2
and ∆X.
28
X. APPENDIX B. UNBALANCE CALCULATIONS FOR MULTIPLE The V59N signal for the second failure (B-phase, F = 1) is:
FAILURES 1 ⋅ 1∠(180° − 120°)
In Section VI, we explained how to perform unbalance V59N = =⋯
3 ⋅ 4 ⋅ 14 ⋅ 1 − 1 ⋅ (3 ⋅ 4 ⋅ 1 − 3 ⋅ 1 + 1)
calculations for a failure that may involve many capacitor units
but that occurred at a single location. In this appendix, we … = 0.006329 pu ∠60°
explain how to apply the superposition principle to perform The V59N phase angle is relative to the faulted-phase voltage.
calculations for failures at multiple locations. Using this Therefore, we subtracted 120° when calculating the V59N
approach, you can perform calculations for evolving and operating signal for the failure in the B-phase.
equalizing failures and for failures under a pre-existing
The V59N operating signal with both failures present is:
unbalance.
V59N = 0.01351 pu ∠180° + 0.006329pu ∠60° = ⋯
Capacitor unit failures create only small changes in the bank
voltages and currents. Equations that tie the capacitor bank … = 0.01171 pu ∠152°
voltages and currents are linear. Therefore, we can use the The V59N operating signal decreased from 0.01351 pu after
superposition principle to solve the bank equations for multiple the first failure to 0.01171 pu after the second failure. This is
failures. expected because the second failure equalized the bank to some
We can think of a derivation of an unbalance equation as an extent.
application of the Thevenin principle. An unbalance protection Assume next that the second failure involves another unit
operating signal is a superimposed quantity in the Thevenin (F = 2 in the B-phase). We can recalculate the B-phase failure
method. The system sources are removed (shorted) as per the by using F = 2, or we can consider the second unit failure in the
Thevenin principle and the change in voltage at the failure B-phase as the third failure in the bank. If we follow the latter
location drives the unbalance signal. approach, we can calculate the V59N operating signal as follows:
Because a unit failure does not greatly affect the voltages V59N = 0.01171 pu ∠152° + 0.006329 pu ∠60° = ⋯
inside the capacitor bank, we can assume nominal conditions
and neglect the previous failure(s) when calculating the … = 0.01311 pu ∠123°
unbalance protection signals for the subsequent failure. Example B.2
Therefore, simultaneous failures at multiple locations can be Consider the grounded double-wye externally fused
treated as a superposition of a set of single failures, where each capacitor bank in Subsection VI.C (P = 15, S = 6, and R = 1),
single failure occurs separately and under nominal conditions. and calculate the negative-sequence (3I2) and the neutral
Use the following procedure to perform unbalance unbalance (I60N) currents for the following evolving failure: 1)
calculations for multiple failures: one unit fails in the A-phase of the left half of the bank, 2)
1. Calculate the unbalance protection signal of interest followed by another unit failure in the A-phase of the right half
separately for each failure, each time assuming a of the bank, 3) followed by another unit failure in the B-phase
healthy bank prior to the failure (use the equations in of the left half of the bank (Fig. B.1).
Appendix A).
2. Observing the measuring polarity convention, failure
location (phase, above or below the bridge, etc.), and
the unbalance signal phase angle, add (vectorially, as
phasors) the signals obtained separately for each failure
and obtain the unbalance signal for multiple failures.
The above approach is not perfectly accurate but is accurate
enough for practical engineering calculations, especially when
considering failures that do not result in very large voltage or
current changes in the bank.
Example B.1
Let us continue Example 2 from Subsection V.B and
calculate the V59N operating signal for the following failure
scenario: 1) two units fail open in the A-phase, 2) the protection
system issues an alarm but does not trip, and 3) a single unit
Fig. B.1. Progressing failure in the grounded double-wye bank in Example B.2.
fails in the B-phase before the bank can be repaired.
We use equations from Appendix A for the fail-open scenario
The V59N signal for the first failure (A-phase, F = 2) is:
(fused bank) and separately calculate the operating signal
2 ⋅ 1∠180°
V59N = =⋯ components resulting from each failure.
3 ⋅ 4 ⋅ 14 ⋅ 1 − 2 ⋅ (3 ⋅ 4 ⋅ 1 − 3 ⋅ 1 + 1)
… = 0.01351 pu ∠180°
29
First unit failure in the A-phase of the left half of the bank:
1 1
3I2(1) = ⋅ 1∠−90° = ⋯
2 6 ⋅ 15 ⋅ 1 − 1 ⋅ (6 ⋅ 1 − 1)
… = 0.00588241 pu ∠−90°
I60N(1) = 3I2(1) = 0.00588241 pu ∠−90°
Second unit failure in the A-phase of the right half of the bank
(we invert the I60N polarity to account for the I60N measuring
convention from the left to the right half of the bank):
3I2(2) = 3I2(1) = 0.00588241 pu ∠−90°
I60N(2) = −I60N(1) = 0.00588241 pu ∠90°
Third unit failure in the B-phase of the left half of the bank
Fig. B.2. Progressing failure in the grounded H-bridge bank in Example B.3.
(we subtract 120° to account for the B-phase being faulted):
We use equations from Appendix A for the fail-short scenario
3I2(3) = 3I2(1) ⋅ 1∠−120° = 0.00588241 pu ∠150° (fuseless bank) and separately calculate the operating signal
I60N(3) = I60N(1) ⋅ 1∠−120° = 0.00588241 pu ∠150° components resulting from each failure.
First unit failure in the A-phase above the bridge in the left
We apply the principle of superposition and sum the above half of the bank (G = 1 – H = 0.5):
operating signal components (vectorially, as phasors) and
obtain the following magnitudes of the two operating signals 0.5 ⋅ 0.5 ⋅ 1
∆V87(1) = 1∠0° = ⋯
for the three stages of the failure: 4 ⋅ 22 ⋅ 2 ⋅ 0.5 − 1 ⋅ (4 ⋅ 2 − 2)
Per-Unit Unbalance Protection Signal After Failures … = 0.00304878 pu ∠0°
First Failure Second Failure Third Failure 1 1
I60P(1) = ⋅ 1∠90°
3I2 0.00588241 0.01176482 0.01556339 2 2 ⋅ 22 ⋅ 2 ⋅ 0.5 − 1 ⋅ (2 ⋅ 2 − 1 + 0.5)
=⋯
I60N 0.00588241 0.00000000 0.00588241
… = 0.01234567 pu ∠90°
The 3I2 operating signal increases after each failure because,
from the point of view of the bank terminals, the bank becomes Second unit failure in the A-phase below the bridge in the left
more unbalanced after each failure. The I60N operating signal half of the bank (we invert the polarity to account for the failure
goes back to zero after the second failure. This is because at that below the bridge, G = H = 0.5):
time, the left and right halves of the bank are symmetrical – they
0.5 ⋅ 0.5 ⋅ 1
both have a single unit failure in the A-phase. When, ∆V87(2) = 1∠180° = ⋯
subsequently, the left half of the bank suffers another failure in 4 ⋅ 22 ⋅ 2 ⋅ 0.5 − 1 ⋅ (4 ⋅ 2 − 2)
the B-phase, the I60N signal increases again to a value as for a … = 0.00304878 pu ∠180°
single unit failure. This is because at that time, the difference
between the failed units in the left and right halves of the bank 1 1
I60P(2) = ⋅ 1∠−90°
is 2 – 1 = 1. 2 2 ⋅ 22 ⋅ 2 ⋅ 0.5 − 1 ⋅ (2 ⋅ 2 − 1 + 0.5)
=⋯
This example shows that even though the 3I2 and I60N signals
are identical for a single unit failure, when multiple failures … = 0.01234567 pu ∠−90°
occur at different locations in the bank, the 3I2 and I60N signals
Third unit failure in the A-phase above the bridge in the right
may differ depending on how the failures equalize inside the
half of the bank (we invert the I60P polarity to account for the
bank.
I60P measuring convention from the left to the right half of the
Example B.3 bank, G = 1 – H = 0.5):
Consider the grounded H-bridge fuseless capacitor bank in
0.5 ⋅ 0.5 ⋅ 1
Subsection VI.D (P = 1, S = 22, R = 2, and H = 0.5 pu), and ∆V87(3) = 1∠0° = ⋯
4 ⋅ 22 ⋅ 2 ⋅ 0.5 − 1 ⋅ (4 ⋅ 2 − 2)
calculate the phase differential voltage (∆V87) and the phase
unbalance current (I60P) for the following evolving failure: 1) … = 0.00304878 pu ∠0°
one unit fails in the A-phase above the bridge in the left half of
1 1
the bank, 2) followed by another unit failure in the A-phase I60P(3) = ⋅ 1∠−90°
below the bridge in the left half of the bank, 3) followed by 2 2 ⋅ 22 ⋅ 2 ⋅ 0.5 − 1 ⋅ (2 ⋅ 2 − 1 + 0.5)
another unit failure in the A-phase above the bridge in the right =⋯
half of the bank (Fig. B.2). … = 0.01234567 pu ∠−90°
30
We apply the principle of superposition and sum the above A. Partial Fail-Open Scenario (Internally Fused Banks)
operating signal components (vectorially, as phasors) and In the unit fail-open scenario (Fig. 9), the group reactance
obtain the following magnitudes of the two operating signals changes as follows:
for the three stages of the failure:
XU XU
XGROUP : → (C.1)
Per-Unit Unbalance Protection Signal After Failures P P−F
First Failure Second Failure Third Failure Assume a partial unit failure (some but not all internal fuses
∆V87 0.00304878 0.000000 0.00304878 in the unit are blown). This partial failure increases the unit
reactance from the nominal value of XU to some other value,
I60P 0.01234567 0.000000 0.01234567
XUF. The affected group now comprises (P – 1) healthy units
The second failure (one unit failed above and one unit failed and one unit that has a different reactance. The equivalent
below the bridge in the A-phase of the left bank) makes the reactance of the group is:
parts above and below the bridge of the A-phase of the bank XU ⋅ XUF
symmetrical. Therefore, the operating signals of both the 87V (C.2)
XU + (P − 1) ⋅ XUF
and 60P elements drop to zero. The third failure makes the A-
phase unbalanced (two units failed above the bridge and one We equate the group reactance from our failure model (C.1)
unit failed below the bridge), and the two operating signals and the actual group reactance (C.2):
increase again. XU XU ⋅ XUF
The 3I2 operating signal increases after each failure because, = (C.3)
P − FFRAC XU + (P − 1) ⋅ XUF
from the point of view of the bank terminals, the bank becomes
more unbalanced after each failure. The I60N operating signal We solve (C.3) to obtain the equivalent partial failure size,
also increases after each failure because, from the point of view FFRAC, as follows:
of the neutral current, the A-phase of the bank becomes more XU XU
unbalanced after each failure compared with the B- and C- FFRAC = 1 − = 1 − 𝛼𝛼, 𝛼𝛼 = (C.4)
XUF XUF
phases.
You can perform unbalance calculations for fail-open
Example B.2 and Example B.3 show how multiple protection
capacitor element failures by using the fractional failure size,
elements can become blind to equalizing bank failures. It is
FFRAC, obtained by using (C.4).
beneficial to use several unbalance protection elements to
mutually cover their weak spots during equalizing failures. If the capacitor unit fails open completely, then XUF = ∞ and
FFRAC = 1, as expected. If there is no failure, then XUF = XU and
XI. APPENDIX C. UNBALANCE CALCULATIONS FOR PARTIAL FFRAC = 0, as expected.
CAPACITOR UNIT FAILURES B. Partial Fail-Short Scenario (Fuseless and Externally
In the main body of the paper, we assumed that a failure Fused Banks)
affects the entire capacitor unit or multiple units. In this In the unit fail-short scenario (Fig. 10), the string reactance
appendix, we introduce a simple method for unbalance changes as follows:
calculations at the capacitor element level (partial unit failure).
S S−F
We represent a partially failed unit by using a fractional value XSTRING : ⋅ XU → ⋅ XU (C.5)
of F. This approach allows us to use the unbalance equations P P
derived for failed units to perform unbalance calculations for Assume a partial unit failure (a short-circuit of some but not
failures of capacitor elements. all capacitor element groups inside the capacitor unit). This
When an element fails short in a fuseless or externally fused partial failure decreases the unit reactance from the nominal
bank, the unit reactance reduces slightly. Similarly, when a fuse value of XU to some other value, XUF. The affected string now
blows in an internally fused bank, the unit reactance increases comprises (S – 1) healthy groups and one group that has a
slightly. We calculate an equivalent fractional failure size, different reactance. That group has P – 1 healthy capacitor units
FFRAC, that represents the change in the unit reactance because and one partially failed unit. Therefore, the equivalent reactance
of the element failure(s) and use this fractional failure size, of the string is:
FFRAC, in the unbalance calculations. S−1 XU ⋅ XUF
⋅ XU + (C.6)
This method is very convenient because it decouples the P XU + (P − 1) ⋅ XUF
arrangement of the bank (units, groups, strings, phases, and
We equate the string reactance from our failure model (C.5)
phase connections) and the arrangement of the capacitor unit
and the actual string reactance (C.6):
(capacitor elements and fuses). Moreover, our method allows
unbalance calculations for banks that include capacitor units of S − FFRAC S−1 XU ∙ XUF
∙ XU = ∙ XU + (C.7)
different types. P P XU + (P − 1) ∙ XUF
31
We solve (C.7) to obtain the equivalent partial failure size, The healthy capacitor unit reactance is 8/5 of the capacitor
FFRAC, as follows: element reactance. When two capacitor element groups in a
XU − XUF 𝛼𝛼 − 1 capacitor unit are shorted, the capacitor unit reactance becomes
FFRAC = = (C.8) 6/5 of the capacitor element reactance. Therefore, α = (8/5) /
XU + (P − 1) ⋅ XUF 𝛼𝛼 − 1 + P
(6/5) = 8/6. We apply (C.8) and calculate the fractional failure
You can perform unbalance calculations for fail-short size, FFRAC:
element failures by using the fractional failure size, FFRAC, 8/6 − 1
obtained by using (C.8). FFRAC = = 0.0233
8/6 − 1 + 14
If the capacitor unit fails short completely, then XUF = 0 and
FFRAC = 1, as expected. If there is no failure, then XUF = XU and Note that the shorting of two capacitor element groups
FFRAC = 0, as expected. removes 10 capacitor elements out of the total of 50. The
calculated fractional failure size of 0.0233 is close to the simple
C. Unbalance Calculations for Partial Unit Failures proportion of 10/50 = 0.025. This observation leads to an
Apply the following procedure to perform unbalance opportunity of using the ratio of removed capacitor elements
calculations for capacitor element failures (partial capacitor and the total number of elements as an approximation of the
unit failures): fractional unit failure.
1. Calculate the nominal (XU) capacitor unit reactance and Next, we use the equation for the 59N operating signal in an
the reactance of the partially failed unit (XUF). Because ungrounded single-wye bank for the fail-short scenario and
only the ratio (α) of the two reactances matters, these calculate:
calculations are very simple and can be done by hand
0.0233
by just inspecting the internal connection diagram of V59N = =⋯
the capacitor unit and counting the capacitor elements 3 ⋅ 4 ⋅ 1 − 0.0233 ⋅ (3 ⋅ 14 − 1)
(see Example C.1 and Example C.2). … = 0.0021 pu = 0.504 V sec
2. Calculate the fractional failure size, FFRAC, by using
For comparison, when we assumed that an entire unit failed
(C.4) for internally fused banks and (C.8) for externally
open (F = 1), we obtained V59N = 0.0063291 pu. When the
fused and fuseless banks.
entire unit fails open, the V59N operating signal is out of phase
3. Use the unbalance equations in Appendix A to with the faulted-phase voltage; when there is a partial fail-short
calculate the unbalance signals of interest based on the failure in the unit, the V59N operating signal is in phase with the
fractional failure size, FFRAC. faulted-phase voltage (see Appendix A). When the capacitor
4. Use the superposition principle described in elements inside the capacitor unit fail short in a cascading
Appendix B to perform calculations for partial and fashion, there is a phase inversion of the unbalance protection
complete unit failures, including cases where the two operating signals at a time when the progressing fail-short
failures occur at different locations in the bank. capacitor element failure becomes a permanent fail-open
Example C.1 capacitor unit failure because of the operation of the external
fuse.
Consider the ungrounded single-wye capacitor bank in the
setting calculation example in Subsection VI.B. The bank is Assume you want to calculate the 59N operating signal for
externally fused. Assume that each capacitor unit has 5 the case when one unit failed open (F = 1) and some other unit
capacitor elements in parallel in a group and 8 groups connected failed short partially (FFRAC). You can use the principle of
in series (Fig. C.1). Calculate the 59N element operating signal superposition and calculate the V59N signal separately for the
for the case where 2 groups of capacitor elements fail short two failures (F = 1, fail open, and F = FFRAC, fail short). To
inside an A-phase capacitor unit. obtain the final V59N value, you must add (vectorially, as
phasors) the two V59N signal components.
Example C.2
Consider the grounded double-wye capacitor bank in the
setting calculation example in Subsection VI.C, but assume the
bank is internally fused (unlike in Subsection VI.C). Assume
each capacitor unit has 15 capacitor elements in parallel in a
group and 5 groups connected in series (Fig. C.2). Calculate the
60N element operating signal for the case where 4 capacitor
elements fail open in a single group inside a capacitor unit in
the A-phase of the left half of the bank.
Fig. C.1. Externally fused capacitor unit in Example C.1.

32
[3] B. Kasztenny, J. Schaefer, and E. Clark, “Fundamentals of Adaptive

Protection of Large Capacitor Banks - Accurate Methods for Canceling
Inherent Bank Unbalances,” proceedings of the 60th Annual
Conference for Protective Relay Engineers, College Station, TX,
March 2007.
[4] MATLAB Symbolic Math Toolbox, MathWorks. Available:
https://www.mathworks.com/products/symbolic.html.
[5] R. Natarajan, Power System Capacitors (Power Engineering), CRC
Press, 2005. ISBN-10: 1-57444-710-6.
XIII. BIOGRAPHIES
Bogdan Kasztenny has over 30 years of experience in power system protection
Fig. C.2. Internally fused capacitor unit in Example C.2. and control. In his decade-long academic career (1989–1999), Dr. Kasztenny
taught power system and digital signal processing courses at several
The healthy capacitor unit reactance is 5/15 of the capacitor universities and conducted applied research for several relay manufacturers. In
element reactance. When 4 capacitor elements fail open in a 1999, Bogdan left academia for relay manufacturers where he has since
designed, applied, and supported protection, control, and fault-locating
group inside the capacitor unit, the capacitor unit reactance products with their global installations numbering in the thousands. Bogdan is
becomes 1/(15 – 4) + 4/15 = 59/165 of the element reactance. an IEEE Fellow, a Senior Fulbright Fellow, a Distinguished CIGRE Member,
Therefore, α = (5/15) / (59/165) = 0.9322. We apply (C.4) and and a registered professional engineer in the province of Ontario. Bogdan has
served as a Canadian representative of the CIGRE Study Committee B5 (2013–
calculate the fractional failure size, FFRAC: 2020) and on the Western Protective Relay Conference Program Committee
FFRAC = 1 − 0.9322 = 0.0678 (2011–2020). In 2019, Bogdan received the IEEE Canada P. D. Ziogas Electric
Power Award. Bogdan earned both the Ph.D. (1992) and D.Sc. (Dr. habil.,
Note that the failure removes 4 capacitor elements out of the 2019) degrees, has authored over 220 technical papers, holds over 55 U.S.
patents, and is an associate editor of the IEEE Transactions on Power Delivery.
total of 75. The calculated fractional failure size of 0.0678 is
close to the simple proportion of 4/75 = 0.0533. The two values Satish Samineni received his bachelor of engineering degree in electrical and
are not identical because the capacitor element failures electronics engineering from Andhra University, Visakhapatnam, India, in
redistribute the voltage inside the unit and affect the unit 2000. He received his master’s degree in electrical engineering in 2003 and a
Ph.D. in 2021 from the University of Idaho. Since 2003, he has been with
apparent reactance. Schweitzer Engineering Laboratories, Inc. in Pullman, Washington, where he
Next, we use the equation for the 60N operating signal in a is a principal research engineer in the research and development division. He
has authored or coauthored several technical papers and holds multiple U.S.
grounded double-wye bank for the fail-open scenario and patents. His research interests include power system protection, power system
calculate: modeling, power electronics and drives, synchrophasor-based control
applications, and power system stability. He is a registered professional
1 0.0678 engineer in the state of Washington and a senior member of IEEE.
I60N = ⋅ =⋯
2 6 ⋅ 15 ⋅ 1 − 0.0678 ⋅ (6 ⋅ 1 − 1)
… = 0.000378 pu
For comparison, when we assumed that an entire unit failed
open (F = 1), we obtained I60N = 0.005882 pu.
Let us go back to (54) and approximate the I60N signal based
on the number of capacitor elements lost because of the failure.
The bank has 15 ⋅ 5 = 75 capacitor elements in a capacitor unit
and 15 ⋅ 6 ⋅ 1 ⋅ 2 = 180 units in a phase, for a total of 75 ⋅ 180 =
13,500 capacitor elements in a phase. When 4 capacitor
elements fail open (are removed), the per-unit unbalance can be
approximated as 4/13,500 = 0.0002962 pu (a 20 percent error
compared with the accurate value of 0.000378 pu). When the
entire capacitor unit fails open, the per-unit unbalance can be
approximated as 1/180 = 0.005555 pu (a 5 percent error
compared with the accurate value of 0.005882 pu).
XII. REFERENCES
[1] IEEE Std C37.99 IEEE Guide for the Protection of Shunt Capacitor
Banks, 2012.
[2] J. Schaefer, S. Samineni, C. Labuschagne, S. Chase, and D. J. Hawaz,
“Minimizing Capacitor Bank Outage Time Through Fault Location,”
proceedings of the 67th Annual Conference for Protective Relay © 2022 by Schweitzer Engineering Laboratories, Inc.
Engineers, College Station, TX, March 2014. All rights reserved.
20220920 • TP7076-01

Kasztenny Paper CapacitorBankProtection0

Uploaded by

Copyright:

Available Formats

Kasztenny Paper CapacitorBankProtection0

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kasztenny Paper CapacitorBankProtection0

Uploaded by

Copyright:

Available Formats

1

Capacitor Bank Unbalance Protection Calculations

Fig. 1. Grounded single-wye bank configuration and unbalance protection.

Fig. 2. Ungrounded single-wye bank configuration and unbalance

where T is the per-unit tap position (the reactance between the

Fig. 8. Nonuniform unit arrangement when creating an 87V tap.

• We derive the unbalance equations by assuming an

VI. UNBALANCE PROTECTION SETTINGS CALCULATION

Voltage (kV LL) 138

D. Grounded H-Bridge Capacitor Bank

Normally, the units operate under a voltage of: ∆V87 = 0.0072816 pu

F where factors k, A, and B are positive numbers that are specific

The product of QU and P is the rated power of a capacitor Example 5

IX. APPENDIX A. CAPACITOR BANK UNBALANCE where:

GROUNDED SINGLE-WYE CAPACITOR BANK

UNGROUNDED SINGLE-WYE CAPACITOR BANK

GROUNDED DOUBLE-WYE CAPACITOR BANK

I60P = I60N = 3I2

UNGROUNDED DOUBLE-WYE CAPACITOR BANK

I60P = 2I60N = 3I2

GROUNDED H-BRIDGE CAPACITOR BANK

UNGROUNDED H-BRIDGE CAPACITOR BANK

Fig. C.1. Externally fused capacitor unit in Example C.1.

[3] B. Kasztenny, J. Schaefer, and E. Clark, “Fundamentals of Adaptive

You might also like