IEEE Transactions on Power Apparatus and Systems, Vol. PAS-97, No.

5, Sept/Oct 1978 1935

PRINCIPLES AND APPLICATIONS OF STATIC, THYRISTOR-CONTROLLED
SHUNT COMPENSATORS
L. Gyugyi R. A. Otto T. H. Putman
Westinghouse Electric Corporation
Pittsburgh, Pennsylvania
ABSTRACT system. Load compensation thus tends to reduce the undesirable
effects of a single load (or load group) on the ac system, without
In recent years, there has been a rapid increase in the number of attempting to change the external regulation of the terminal voltage.
thyristor-controlled shunt compensators used in industrial and utility (This is in contrast to transmission line voltage support, where the
systems for dynamic power factor correction and terminal voltage compensator is employed to reduce terminal voltage variation regard-
stabilization. These thyristor-controlled shunt compensators function less of its cause).
as variable reactances operated in both the inductive and capacitive In the following three sections, the requirements for load
domains. compensation are established and related to different load parameters.
This paper examines the theoretical foundations of thyristor-
controlled shunt compensation. Conditions for unbalanced load Fundamentals of Load Compensation
compensation and voltage stabilization are established with the use of
symmetrical components. Mathematical relationships applicable to Consider a general ungrounded three-phase load represented by
feedforward (computational) type controls are derived. Practical impedances Za, Zb, and Zc, connected to the terminals of an ac
approaches to both feedforward and feedback type controls are system with a set of symmetrical line to neutral voltages Va V, Vb = =

discussed. Vei2l/3 and Vc = Ve-j41r/3 as shown in Figure la. The delta equivalent
INTRODUCTION of the load represented, for convenience, by admittances Yab, Ybc,
and Yca is shown in Figure lb. Impedances Za, Zb, and Zc (or
Shunt compensators are used to control (or regulate) the admittances Yab, Ybc, Yca) are, in general, functions of time. In order
transmission voltage at a given terminal and to provide power factor to establish the basic compensation requirements, it will be assumed
correction. Two types of compensation problems are normally that the three load impedances are time-invariant and different from
encountered in practical applications. The first is load compensation each other, that is, they represent a general steady state unbalanced
where the requirements are usually to reduce or cancel the reactive load. This assumption does not exclude the eventual consideration of
power (VAr) demand of large, and fluctuating industrial loads, such as
time varying impedances. Indeed, the three steady state impedances
electric arc furnaces, rolling mills, etc., and to balance the real power may be viewed as representing the load at a particular time instant. A
drawn from the ac supply lines. These types of heavy industrial loads series of "steady state" impedances, representing the load at discrete
are normally concentrated in one plant and served from one network
time instants appropriately close to one another, can reconstruct a
terminal, and, therefore, can be handled best by a local compensator time varying load (having rates of change sufficiently below the ac
connected to the same terminal. The second type of compensation is supply frequency) just as a chain of still pictures can faithfully
related to the voltage support of transmission lines at a given terminal describe a moving object.
in the face of disturbances of both loads and generation. Here the
load is not localized; several load areas and generator units may be tied
by a transmission network and the objective is simply to regulate (or
control) the voltage at the compensated terminal.
Static shunt compensators are essentially controllable VAr
sources - or VAr generators - functionally similar to the traditional
rotating synchronous condenser. Several types of static compensators,
with different operating features, can be realized using various power
conversion concepts and thyristor circuits.6'9 In this paper only the
variable impedance types, that is those which essentially function as
Fig. 1- la) General three-phasewye connected load represented by
variable reactances (capacitive and inductive impedances), are con- impedances (b) Delta equivalent represented by admittances
sidered. Regardless of the type of shunt compensators employed the
basic compensation principles are the same, requiring the generation With the above assumption, the compensation problem becomes
of controllable VArs to supply the reactive and negative-sequence load that of finding a reactive'admittance network which when combined
currents (load compensation) or to induce an appropriate voltage with the load admittance will present a real and balanced load to the
across the ac system impedance to regulate the terminal voltage supply terminal.
(voltage support). Using the delta equivalent representation of Figure lb, the three
In this paper theoretical bases for both load compensation and load admittances can be compensated separately as if they were three
transmission line voltage support are investigated. The intention is the single-phase loads. Consider for example phase ab. If admittance Yab
derivation of analytical relationships that are applicable to the control is composed of a real component (conductance) and reactive
of practical shunt compensator schemes using thyristor circuitry. component (susceptance), that is
LOAD COMPENSATION Yab = Gab + j Bab (1)
the reactive part can, as a first step of the compensation, be cancelled
There are usually two main reasons for compensating large, by an appropriate compensating susceptance, -Bab, connected in
fluctuating loads: (1) the ac system is too weak to maintain the parallel with Yab as illustrated in Figure 2.
terminal voltage with acceptable variations and (2) it is not econom-
ical, or practical, to supply the reactive power demand from the ac a
Yab =
Gab+ i ab
E

vzYabbGab
b'

F 78 096-0. A paper reccrmended and approved by

Fig. 2- Compensation of the reactive part, Bab, Of
the IEFEE Pcwer System Engineering Ccrnittee of the admittance Yab by susceptance Bab -

IEEE Poer Engineering Society for presentation at the

IEEE PES Winter Meting, New York, NY, January 29 - As a second step of the compensation, real admittance Gab has
February 3, 1978. Manuscript submitted Octcber 4, to be complemented with a reactive admittance network so as to
1977; made available for printing Noventber 4, 1977. obtain a resultant balanced load on the ac supply. This problem was
0018-9510/78/0900-1935$00.75 (i 1978 IEEE
1936
originally investigated by C. P. Steinmetz at the beginning of
century. It is easy to show (as is indeed well known) that the load
the ac supply becomes balanced (and remains real) if a capaci
susceptance SV
B(ab) = Gab! (2) c C =c ~ ~ ~ ~ ~~bCb SC

is connected between phases b and c, and an inductive susceptance P= 3V2 -

B(ab) = Gab/ (3) Fig. 5 - Compensation of the single- phase load, Y by susceptances - Bbc. B (bc ) and B(bC)
is connected between phases c and a as shown in Figure
(Superscript (ab) is used to indicate that the susceptances are nee
to compensate phase ab). Since the single-phase real power supp
by the ac source before compensation
pk=(/y V)2 Gab= 3V2 Gab (4)
must be equal to the three-phase real power demand after compensa-
tion, the compensated load may, in effect, be represented as three wye P = 3V2 Gca
connected resistors, each with a conductance of Gab as shown in
Figure 3. The mechanism of compensation is depicted in Figure 4 Fig. 6- Compensation of the single-phase load, Yca, by susceptances - Bab and BbCOl
where a phasor diagram illustrates how two (capacitive and inductive)
susceptances transform the single-phase real load into a balanced
three-Phase real load. Complete compensation of the three-phase load is readily
obtained by adding all compensating susceptances in each phase as
B(COab)
0 =_
Ga
- _6b
c
illustrated in Figure 7. Thus, the resultant compensating susceptances
J _~t
B(ab), B(bc), and B(c) are simply the sums of the susceptances used to
compensate the individual phases, namely,
S --
Gab (
B(c)ab ab + B(bc)
Bab ab + B(ca)
ab
la Va Gab V Gab
B(B
= +B(ca)+ (Bab) (9)
Ib VbGab
C V Gb exp(- i/3)
=I It Vc Gab = V Gab eap (-j 4 /3
P = 3 V2 GC
B(c)
ca =B ca
+ B(ab) + B(bc)
ca ca
Fig. 3- Compensation (balancing) of the single-phase real (resistive) load. Gab, bytwo (capacitive and where superscript (c) stands for "compensating."
inductive) susceptances. 8 lab) and Blabl

C
ca.
V
L,ca 'C. bc

.0 a
o~~~~~~~~~~~~~~~~~~~~~~

55.0~ =b~~~~~~~~
CCGbCB

C.b opestn
b. bcurens aCbc, IL caRadlv urns a a I. L
rtentoki sca o = -n
(bcC)c 0 b ca (c)Ga G_
b5C
ca__
b5c
BabCI
3
-- -
aC. (
A prcdr
oc) = -+ Gab
iiabt htotie abovanRbe folc edt
otherFig. 7twoApplicadion
- phases. With ot Eqs.
of the principles (2),ab(3)
single- phase and to an unbalanced
toad compensation ab) bo(),tehe
th ree- phase load

The first term in Eqs. (9) represents the reactive component of

the load in the phase considered, the second and third terms represent
B
Gbc/ -\,/c3ca
ca C.b the reactive requirements to balance the real part of the load in the
other two phases. With Eqs. (2), (3) and (5) to (8), the three
B ab GbcIV3 (6)~Rab susceptances, required to compensate a general unbalanced load, can
be expressed in terms BCCa)=of- BCa
the +real (BCaand reactive BCab parts of the load
admittance, that is,
B(C=- Ba + (Gc - b)V-

and
BICnn bc+(ab -Gca) /O(0)

(7) a
Bca + Gbc Gab)! V
It can be seen that if the real part of the load is balanced, that is,
bca)= -ca Gab = Gbc Gca, the compensation problem
= is reduced to that of
simply cancelling the reactive part of the load in each phase. In
general, the compensating susceptances given in Eqs. (10) transform
respectively. The compensation of admittances Ybc and Yca is any linear unbalanced load into a balanced real load that, from the
illustrated in Figures 5 and 6 respectively. standpoint of the ac source, can be represented as three wye
 1937
connected resistors, each with a conductance that is the sum of the I
conductances of the original delta load (G = Gab + Gbc + Gca). From I
this it follows that the power factor of the resultant load can be
changed from unity to any desired leading or lagging value by adding
the same capacitive or inductive susceptance to each compensating al = (la + alb +aIC) (14)
susceptance defined by Eqs. (10).
The fundamental considerations presented in this section lead to Ia2 = 13(3(Ia + aIb aI )
aI + ac)
the following conclusions:
where laO' Ial and Ia2 are the reference phasors of the zero, positive,
1. An arbitrary ungrounded unbalanced linear load can be trans- and negative-sequence current sets.
formed into a balanced three-phase load without changing the real Substitution of Eq. (13) into Eq. (14) gives the symmetrical
power exchange between load and source. components in terms of the load admittances:
2. The transformation requires appropriate modification of reactive
load power in each phase; this process is generally referred to as 'aO = 0
load compensation.
'al =(Yab+Ybc+Yca)V (15)
3. In general, load compensation involves two separable acts; one is
power factor correction and the other is load balancing. If both are
carried out, they determine together the total reactive power Ia2 =- (a2 Yab + Ybc + a Yca) V
needed for compensation. The objectives of load compensation can be stated in terms of
4. Any unbalanced linear steady state load can be compensated by the symmetrical line-current components as follows:
three appropriate reactive impedances (capacitive or inductive 1. Eliminate the negative-sequence components (balancing).
reactances) connected between the three phases. 2. Eliminate (reduce) the reactive part of the positive sequence
5. If the load changes in time, the compensation requires variable components (power-factor correction).
reactances (or other controllable VAr sources) that can be adjusted The results of the previous section indicate that these objectives
in sympathy with the load variation.
can be achieved by a delta connected compensating network formed
Compensation Requirements in Terms of the Symmetrical of reactive circuit elements (susceptances) as shown in Figure 8. The
Components of Load Current purpose of the present investigation is to find relationships between
the compensating susceptances and the symmetrical components of
In this section the compensation requirements of a general load current.
unbalanced load in terms of symmetrical components are investigated. The symmetrical components of current drawn by the delta-
This investigation complements the fundamental considerations in the connected compensating network can be expressed in terms of the
previous section and provides a different interpretation of the results compensating susceptances in forms similar to those given in Eqs.
obtained. (15):
Consider again the general unbalanced three-phase load fed from
a balanced three-phase (ungrounded) source as shown in Figure 1. In the 1(c)
a0 = 0
present analysis the delta equivalent of the load, defined by
admittances Yab, Ybc, and Yca is used again for convenience, as
shown in Figure 8. i(c)=
al i (B(c) bc + B(c))
ab + B(C) ca
V (16)

I(c)
a2
-j (a2 B(c) +
=
ab bc
+a
ca
B(c) B(c))
V
The compensation requirements (load balancing and power-
factor correction) can be formulated mathematically as follows:
Ia2 + a(c) 0 (17)
and
ImIal al - (18)
Compensator Load Inspection of Eqs. (15) and (16) indicates that both Ia2 and l(c)
a2
Fig. 8- Three-phase unbalanced load represented hy delta connected admittances and
the delta connected compensator generally have a real and imaginary part, while Im lal and I a)(c)
are, by
As stated, the source voltages are assumed balanced with a definition, imaginary. Therefore, the condition given in (17) repre-
positive phase rotation, that is sents two equations, namely

Va = V, Vb = aV, VcaV (1 1)
ReIa2+ReI(C)=0 (19)

where a is the customary complex operator + Im I(c) = 0

Im Ia2a2 (20)
a2
a = ei27/3 = -12+j (12) Substitution of I(c)
al ' Re I(c),
a2' and Im I(c)
a2 from Eq. (16) into (18),
The three currents drawn by the load are (19), and (20), respectively, and subsequent solution of the three
equations for the three compensating susceptances provides the
following results:
Ia =lYab (l-a2) -Yca (a1)t V -

Bab (3V) =-Im IaV

I
Ib =Ybc(a2_ a)-Yab(l -a2)}V (13) Im Ia2 + Re Ia2

Ic =Yca (a- 1)- Ybc (a2 _a)t V B(c)

bc(cV)
(
= - 2Im I a1 + Im
/3I Im I a2
2
(21)

The symmetrical components in terms of the three line currents are

given by B(c)
ca (,./3V), N/-
- Im lal -71 Im'2-R
a
-R a2
1938
Equation (21) gives the compensating susceptances in terms of where R + jX is the (balanced) supply network impedance (refer to Fig-
the symmetrical (positive and negative-sequence) components of the ure 9). The conditions expressed for the negative-sequence currents in
load current and clearly shows the properties of the compensation Eqs. (19) and (20) are, of course, not changed. Thus, from Eqs. (19),
investigated. The three compensating susceptances are identically (20) and (25), with the use of Eqs. (16), the compensating
proportional to the imaginary part of the positive-sequence load susceptances can again be expressed in forms similar to those given in
current, each a different function of the negative-sequence load Eqs. (21). In fact, the results may be obtained directly from Eqs. (21)
current. The contributions from the positive-sequence, and negative- by simply replacing the term - Im Ial with R/X Re Ial - Im Ial This
sequence currents are independent. Therefore, the compensating means that the condition for zero in-phase voltage drop is satisfied by
network can be viewed as two independent deltas. One (balanced) providing leading (capacitive) current for the ac supply network in
delta provides the positive-sequence reactive load current, the other proportion to the real load current. In other words, the power factor
the negative-sequence load current, so that the ac source has to supply at the supply terminals is leading and equal to cos (tan-1 R/X).
only the positive-sequence real load current. The concept of load The phasor diagrams in Figure 10 illustrate the compensation in
compensation with a "positive-sequence delta" and a "negative- terms of the symmetrical components. At (a) the phasors of an
sequence delta"is illustrated in Figure 9. The sum of the susceptances is unbalanced set of three load currents are shown. At (b), the elimination
zero in the delta providing the negative-sequence compensating of the negative-sequence components, that is, balancing by a negative-
current. (This follows from Eqs. (16) which shows that the positive- sequence compensator, is illustrated. The compensation of the
sequence current is zero only if the sum of the susceptances is zero; it positive-sequence current for unity power factor is illustrated at (c).
can be observed that Eqs. (21) adheres to this condition). The sum of IC)2|= a2 12Ic2)
U,a Ic)2t=a 2 2a?
1 c)
the susceptances in the delta providing the positive-sequence compen- V =av it) v ~~~V =aVt~
sating current is proportional to the reactive power demand of the
load.
Real = lal + 'a2
VGa X R V a ,la
AC V V aGe, =la2=aO+
Ib a Ia2
source G-T'AX -.-b
R
v x Rea
- I-1 a lal
R _,a IC -a Ial + a21a2
Source

ImNdaCe ~ X 'C _, CE
a o c
4,
Fig. 10- Phasor diagrams illustrating load current compensation In terms of the symmetrical components.
a a
ta) A set of three unbalanced load currents I4*1a and Ib and IC C
with the a system voltages Va,
a'bVh.
and Vc hi Balancing of the ioad currents by the negative- sequence compensating currents I a2c Ib2c
b b Load aI 5t2. and C) = a I a2 (c) Elimanation of the reactive compionents from the balanced load currents
Positive- Sequence Negative- Sequence
Compensator Cornpensator by the positive- sequence compensating currentsIal Ibl= aI 2 1c1 and Icl aIal a
B(cl B(c) B(c)
P ab P, bc P, N
Bc)IC)8() 06(c) =o
NB ab + N bc+ N, ca
Compensation Requirements in Terms
Pig. 9- Representation of the compensator by positive- sequeece and negative- sequence admittance deltas of Load Current and Power Quantities
The susceptances of the "positive-sequence delta" and "negative-
sequence delta" can be expressed in terms of the real and imaginary In the previous two sections, the compensating susceptances are
parts of the load admittances from Eqs. (21) with the use of Eqs. (15), expressed in terms of the load admittances or the symmetrical
namely components of load currents. While both types of expressions can
provide bases for practical control approaches, equations relating the
(c) (c) (c) 1 compensating susceptances to directly measurable load currents and
BP,ab BP,bc = Bp,ca = - ab + Bbc + Bca) (22) powers are often preferred in actual compensating systems. Several
different types of expressions can be derived (with more or less labor)
B(c) 2 1 1 from either of the two basic sets of relationships given in Eqs. (10)
N,ab 3 Bab + 3 Bbc + Bca + 1 (Gca Gbc) and (21). (As already indicated, these equations themselves can be
converted into each other). Four sets of equations which appear to be
particularly attractive for practical implementation are now presented.
Nbc 3Bab1 + Bbc
c) j 2 +- Bca1 +71 (Gb Gca) (23)
In these sets of equations the compensating susceptances (or corre-
sponding currents and reactive powers) are given in terms of in-phase
N,ca 3 Bab bbc ca + bc a) and quadrature components of load currents (measured with respect
to appropriate line-to-line or line-to-neutral voltages) or corresponding
where subscripts P and N stand for positive-sequence and negative- real and reactive power quantities. The first two sets satisfy the
sequence, respectively. It can be observed that the sum of Eqs. (22) conditions for zero in-phase voltage drop [Eq. (25)] and zero
and (23) results in the same total compensating susceptances given in negative-sequence supply current [Eqs. (19) and (20)]
Eqs. (10) derived by a different approach. It is pointed out here that,
for economic reasons, the compensating network is normally not
K Re Vb*c) I-K
b-K2 R(Va
broken into two deltas. Indeed, there is no apparent reason to do so
since one compensating delta with controllable susceptances can
Bab(c) 1kVXr3/jVI ReoV/ I a
provide both the positive and negative-sequence currents required for
the compensation.
The fact that compensation of the positive-sequence and B(CC)
( XV) = K1 Re
c
V Ic K2Re (b))b (26a)
negative-sequence load currents can be carried out independently (by
a single controllable compensating delta) makes control of the reactive
power provided for the ac supply quite simple. This may be required
in practice for power factor adjustment and compensation of voltage
drop across the supply (ac network) impedance.
B(ca)(/V) = - K1Re\Vt Ia- K2Re V\vjIc
Frequently, the compensation requirement is not simply cancel-
lation of the reactive component of the positive-sequence load where K1 = 1/ -/3 - R/3X, K2 = 1/V + R/3X, R is the real part and
current, as stipulated in Eq. (18), but, rather, the reduction of the X is the imaginary part of the supply impedance, and Vat, VbC*, VcA
in-phase voltage drop across the balanced supply impedance to zero. are conjugates of the line-to-line voltages, that is, Vat = (Va - Vb3
The condition for this can be written simply as Vb* = (Vb - Vc)* and Vca = (Vc - Va)*. Expressions Va*/Nj3V,
Re [(I + I(c)) (R +jX)l = 0 (24) Vbc/NF V, and Vc*al/v3V represent unit phasors. Thus, the real part of
each phasor product in Eqs. (26a) gives a component of the relevant
or load current that is in-phase with the line-to-line voltage in question.
For example, Re (Va*b Ia)/V3`V specifies the component of load
R
Re Ial Im Ial I(a)= 0
current 'a that is in phase with line-to-line voltage Vab.
 1939

Equations (26a) can be written in a different form, which implies dva/dt > dvb/dt > 0 dvc/dt >0
a way to measure the load current components sought, by converting The expressions ia(t) va = ,lib(t) -b=0 , ic(t) v =Q
the phasor quantities into appropriate time functions. Thus
mean to take the value of each load current at the instant of time
vca > 0

a(b B= -K1I ib (t) when the corresponding voltage is zero, and its derivative is positive,
vbc

/V)
dvbc/dt o -K
K2 ia(t) dvca/dt =
0
that is, at those zero crossings at which the voltage changes from
negative to positive. (At negative-going zero crossings the signs of Eqs.
t)vca (28b) should be changed). Of course, these values represent the
BgC)B(c
> 0
vab >O
(,/3V)
(V/V) i K

ib(t) (2b
= - K1 ic(t) dv
ca/dt K2 amplitudes of reactive load current components.
The following two similar equations express the compensating
susceptances in terms of the reactive phase powers of the load.
B(ca(c (.3V) =-K1 ia (t) |dvab/dt
> >0
0 K2 ic(t) dvbc/dt =
vab vbc
/V iat

0
B(c) (3V2) = - Im
(Va 1a) - Im V 1b) + Im (Vc* Ic (29a)
Alternatively
The expression, for example, ib(t) dvbc /dt 0 simply means to take
ab( ( 2) T f [vbc(t) ia(t) + vca (t) ib(t) - vab(t) ic(t)] dt
the value of current ib(t) at the instant of time when the derivative of (T) (29b)
vbc(t) is zero and vbc(t) is positive, or, in practical terms, to sample
the current ib(t) at the positive peak of voltage vbc(t). Evidently, the
quantity so obtained is the amplitude of the component of current
ib(t) that is in phase with voltage vbc(t). (Note that the sampling can TRANSMISSION LINE VOLTAGE SUPPORT
also be carried out at the negative peak of the voltage if the signs in
equations (26b) are changed from negative to positive.) In the previous sections the compensating susceptances are
Inspection of Eqs. (26a) indicates that, by multiplying both sides expressed as functions of the quantities (admittances, currents) that
with V/3V, the compensating susceptance (and reactive compensating characterize the load to be compensated. These relationships, as will
power) can be expressed in terms of average real power quantities. be seen, provide useful bases for practical control strategies in those
Furthermore, since the expressions specifying the three compensating cases where the compensation of a single load (or load group) is
susceptances are identical, except for the subscripts a, b, c and ab, bc, required.
ca, which follow a natural cyclic sequence, it is enough to write only In a multi-terminal electrical transmission network tying several
load areas and generation units, these relationships may not be
the equation for B(C).ab
The other two can be obtained simply by applicable. It may not be possible, or practical, to measure the
commuting the (phase related) subscripts. Thus quantities that would meaningfully characterize the load. Similarly,
the compensation objectives usually cannot be defined in terms of the
B( (3V2) = - K1 Re(VbIb) - K2 Re (Vc* la) (27a) transmission line currents. The primary interest at the terminal of a
transmission line is generally the voltage because (1) it determines the
where again K1 = 1/ V3
R/3X, K2 = l/V/3 + R/3X and expressions
-

Re (Vbc lb) and Re (Vc* Ia) evidently define average real (in-phase)
available or transmittable power, and (2) it may have to meet some
quality requirements for consumers. For this reason, transmission-
power quantities in phases bc and ca corresponding to currents Iband network compensation requires a different approach than load
Ia respectively. compensation. In the case of load compensation the objective is to
Equation (27a) can also be written in an equivalent form in transform the current drawn by an unbalanced and generally
which the currents and voltages are expressed as real time functions, non-unity power factor load into a balanced set of (real) line currents.
namely The quantities characterizing the load are measured, and the line
currents are controlled. The impedance of the ac system plays no
B(c) (3V2)m I 1T vab(t) ib(t) dt -
K2 T l vca(t) ia(t) dt significant part in the compensation since the terminal voltage itself is
(T) (T) not controlled. By contrast, the principal objective of transmission
line compensation is regulation (support) and, in some cases, balancing
where T is the period time. (27b)
of the terminal voltages without regard to the actual cause of their
As Eqs. (26) and (27) show, each compensating susceptance is variation and unbalance. The compensating susceptances (or currents)
expressed in terms of two in-phase (line-to-line) load current compo- must therefore be expressed in terms of the terminal voltages.
nents or two real power quantities. From these relationships [or from Consider a general network compensation problem illustrated
the original ones given in Eqs. (10) and (21)] additional sets of schematically in Figure 11. A number of transmission lines are tied
equations can be derived in which each compensating susceptance is together at a terminal, the voltage of which is to be regulated and
expressed in terms of the actual quadrature components of all three balanced. Generators as well as loads may be connected to each
load currents, or the three corresponding reactive phase powers. These transmission line.
equations do not adhere to the condition of maintaining zero in-phase
voltage drop across the supply impedance [Eq. (25)], rather, they Transmission SW
satisfy the condition of keeping the positive-sequence reactive current Unes Compensator
in the supply lines zero [Eq. (18)]. Terminal Sequence Voltages Compensator Sequence Currents
/V aOt' atVt a2at
) (SW open) 1c1) I'c)
a2 (SW closed)
The first two equations given below (again for B(c))
ab
express the
VaO Val Va2 (SW closed)
compensating susceptances in terms of the reactive components of
load currents. They are analogous to, and can be derived from Eqs.
(26a) and (26b) if the R/X ratio is set to zero. Fig. 11 - A general transmission network terminal with a compensator

The voltages at the terminal are generally unbalanced, comprising

B(C) (/3V) I-
ab V1[3 m (Val + Im v lb -Im ( IC (v zero, positive, and negative-sequence components, which will be
denoted by

where V * V* and V* are conjugates of voltages Va, Vb

(28a)
and Vc.
VOa) V(at), V(t) (30)
Similarly As was previously established (refer to Eqs. (16) and Figure 9),
the compensator can generate positive-sequence reactive current com-
B ab ( XV) = - la(t) + ib(t) - ic(t) ponent, (c) and negative-sequence current component, I(c2) The
va =
0 vb =O vcO0 J
zero-sequence compensator current, I(c),
aO is zero because of the delta
configuration. Thus the compensator is unable to influence directly
1940
the zero-sequence terminal voltage (which does not generally exist in equivalently, with the positive-sequence compensating current I(al)
practice anyway, and therefore will not be considered hereafter). Therefore, the condition IVal = constant [Eq. (33)] can be satisfied.
The symmetrical components of the voltage at the network (Note, however, that the phase angle of Vai will also vary with B(p)).
terminal, with the compensator connected (SW closed), can be
expressed in terms of the current components of the compensator and Substituting Val given by Eq. (38) (with the B(C) required for the
the original terminal voltages as follows: condition IVal I = constant) into Eq. (36b), the negative-sequence
compensating current, 1a2, and the corresponding susceptances given
Val l al 12 a2 in Eqs. (21) can be determined. Thus, it can be concluded that
objectives expressed in Eqs. (32) and (33) can, theoretically, be
(31) satisfied with the compensating network considered. It is worth noting
1(c) that the compensation problem becomes considerabily simpler if the
Va2 = Va2 1 1 Ia2 -Z22 transmission network is balanced and'has no coupling between phases.
where Z1 ,1X Z 12, Z21, and Z22 are the symmetrical component imped- This occurs When the mutual sequence impedances Z12 and Z2 1 are
ance coefficients of the network. zero, so that Eqs. (36) take the simple, self-explanatory form
With Eq. (31) the objectives of the compensation may be stated
as follows: (1) eliminate the negative-sequence voltage and (2) I(c) = (V(t) V )/Z
ala/Z
stabilize the positive-sequence voltage at the terminal. Mathematically (39)
this means that
Va2 ° (32)
1a2 -
a2/Z22
IVal I V= constant (33) It has been shown that although the compensating currents
(susceptances) required to satisfy the compensation objectives can be
analytically determined, the equations obtained could not be imple-
al and l(c),
Equations (31) can be solved for I(c)- a2 and with the mented in a practical control scheme because the network coefficients
subsequent use of Eqs. (32) and (33), the sequence-currents required are neither accurately known nor measureable. This problem can,
for compensation can be specified. Thus however, be circumvented by appropriate feedback-control tech-
niques. In these, the compensating currents (susceptances) would
respond to appropnate error signals that would indicate the deviation
al 11 a) + Y12 (V Va2) (34) of the voltage-sequence components from their desired values at the
network terminal. A feedback control scheme compatible with the
analytical approach investigated above is depicted schematically in
(c2) = 21 (Val) aV) +Y22 (V(t) - Va2) (35) Figure 12. As seen, two separate control (regulator) loops are
employed. In one loop, the magnitude of the positive-sequence
where Yll, Y12, Y21, and Y22 are the symmetrical-component terminal voltage IVal I is measured and compared to a reference (Vp R
admittance coefficients corresponding to impedance coefficients ZI 1, constant). The difference of these two signals is used to vary the
Z12, Z21, and Z22. With the substitution of Eq. (32), Eqs. (34) and compensating susceptances B(C)b, B(C) and B(c) that produce the
(35) become
(V(t)- Val) + Y 12 V(t) positive-sequence compensating current Ia l) (refer also to Figure 9).
Ial= 11 (36a)
The high gain in the control loop causes IVa I to become nearly equal
to VP,R, satisfying one of the objectives of the compensation [Eq.
a2
1() (Vt -vP +
21 (Val Val) Y22 Va2( ()(36b) (33)]. In the other loop the magnitudes of the real and imaginary
components of negative-sequence terminal voltage, Va2, are measured.
Equations (36) gives a general relationship between the compensating The reference for both of these components is zero in accordance with
sequence-currents, the symmetrical-component admittance co- the compensation condition specified in Eq. (32). Thus, the control
efficients and the sequence voltages of the ac system terminal. varies the compensating susceptances, B(c)
Inspection of these equations suggests immediately that, in contrast to N,ab' B(c),
N,bc' and Bc)
N,ca so as
the relationships derived for load compensation in the previous
sections, they would not, in general, provide a useful basis for any to produce the negative-sequence compensating current la2) required
practical control approach. The primary reason for this is that the to reduce the negative-sequence terminal voltage to zero.
admittance coefficients of the ac system, in which load and generation Network
changes may occur, are not known and cannot be measured. Similar Terminal Network
difficulties exist with terminal sequence voltages V(t) al and V(t) a2'
valv a?
Despite the impracticability of Eqs. (36) they are helpful in
understanding the problem of terminal voltage compensation and, for
this purpose, they will be further investigated. The first question is
whether Eqs. (36) is soluble at all under the restriction that the
positive-sequence compensating current I(c) must, per Eqs. (16), be
purely reactive (that is Re I = 0). Let I(j) be given by
l(C) = i B(c) Va (37)
'ImV12
where, per Eqs. (16), B(C) B(c) + B(c) + B(c). From Eq. (36a)
=

Fig. F2- Schtic illustration d newerk compensation by the method of regulating

negatIve-sequence voltage components with two Independent
the positive-squence and
feedbc controls
VaiV 1 = 11val +Y1y12
Y Va2 (38)
(c) The same compensation objectives can be accomplished by the
ijBP y11
+ simpler, and therefore often preferred, technique of regulating the
three line-todline voltages at the network terminal by three essentially
independent feedback loops, each controlling one of the three
Since the numerator of the right-hand member of Eq. (38) is
compensating susceptances, Evidently, if the control succeeds in
constant (a complex number), the magnitude, IVal , of sequence making the three-line-to-line voltages equal in magnitude, the phase
voltage Val varies with the compensating susceptance B(C)
p
(or, displacement between them must also be equal; consequently, they
 1941
must be a balanced set. To achieve this, the compensating network both switching "in" and "out" an inductor is, of course, one half
must, of course, produce both the positive and negative-sequence cycle.
compensating currents in response to error signals representing the In practice it is usually necessary to connect an inductor in series
magnitude differences of the three line-to-line voltages from a given with each capacitor bank in order to limit the current in the thyristor
reference, and do so in accordance with Eq. (36) but without actual switches due to possible differences between line and capacitor
knowledge of the equation coefficients. voltages at the switching instants and to reduce the risk of establishing
Although the feedback control approach is practical for trans- resonances with the ac system impedance at those frequencies at
mission network terminal compensation, it has a number of problems which excitation by harmonic load currents is anticipated.
to contend with (which are not subjects of this paper), such as The basic scheme using a fixed capacitor in parallel with a
stability, speed of response, steady state errors, terminal-voltage thyristor-controlled inductor is shown in Figure 13a. In this arrange-
measurements and error-signal processing. These problems are inter- ment the conduction interval of the thyristor switch, and thereby the
related and their analysis is complex and usually requires modeling current flow in the inductor (L) is controlled. This is achieved by
and computer simulation. The main difficulty is due to the dynamic delaying the closure of the thyristor switch by an angle, ae, in each half
interaction between the ac system and the compensator because the cycle with respect to the peak of the applied voltage. The relationship
terminal voltage compensation must, after all, be achieved by forcing between delay angle and the amplitude of the fundamental inductor
appropriate currents through the impedances of the ac system. In current can be expressed as follows:
other words, the impedances of the ac system, which are not
accurately known and characterized, are of necessity included in the
control loops. ILF jL -° --sin 2a)
The feedback control approach becomes, of course, considerably
less involved if the ac system is balanced and the compensation (40)
objective is simply terminal-voltage (positive-sequence) regulation. °<x2
This allows the use of a relatively simple single control loop (for,/
example, the control loop shown at the top part of Pigure 12) and the
analysis and computer simulation of a single-phase model.

PRACTICAL APPROACHES TO STATIC SHUNT COMPENSATION (ia) Ig() ic+iL(a)

In the previous sections -it has been established that both load
and transmission network compensation with given objectives can be ThyristDr
accomplished by three susceptances connected in delta. Under Switc'h
steady state conditions the compensating susceptances can be ex- a ~ ~2 G a
pressed in terms of the load impedance (load compensation) or
transmission network parameters (terminal-voltage regulation). Since b
the load impedance or related quantities, such as load currents and Fig. 13 - Basic flxed capaitDr. thyristor- controlled inductor type compensator and associated waveforms
powers, are measurable, the compensating susceptances can always be
determined by the direct use of appropriate analytical expressions. By
contrast, the parameters of a transmission network in which changes
both in loads and generation can take place cannot be known
accurately. Therefore, the required compensating susceptances must Thus, the magnitude of the total fundamental current is:
be found "experimentally" by the use of a feedback control.
In practical applications neither the load nor the transmission 1 2 1
network is, in general, in steady state. Indeed, it is usually a principaf IF (a)IC-ILF(a)= (I---a--sin
F C11,17(cl) V [oC -- 0
r 7r 2a)] (41)
=
IwC

requirement for the compensator to reduce the dynamic disturbances

on the ac system. For effective dynamic compensation the following IF(a) is positive when the total current is capacitive [IC > ILF(a)]
are required: (1) rapidly variable compensating susceptances, (2) fast and negative when the total current is inductive 1ILF(a) > IcJ
measurement of load parameters (load compensation) or ac system The control process is illustrated in Figure 1 3b, where the
voltages (terminal-voltage regulation) and (3) rapid signal processing controlled-inductor current, iL(a), the fixed-capacitor current ic, and
and/or fast-acting feedback control. These facets of static shunt the total current, i(a), with its fundamental component iF(a1), are
compensation are examined in the following sections. shown together with applied voltage, v, as the conduction interval of
the thyristor switch is reduced from maximum to zero (a increased
Realizations of Variable Susceptances from 0° to 900). In this illustration c.L is assumed to be smaller than
l/clC. That is, the rating of the inductor is assumed to be higher than
Two basic schemes 1,2,3,4,5,6,7,8,10 are presently used (al- that of the capacitor for the purpose of realizing a variable
though others6'9 are available) to provide a variable susceptance. In susceptance with a- range of control in both the inductive and
one, the susceptance is varied by synchronously switching capacitor capacitive domains.
and inductor banks to the ac lines. In the other scheme, the same From the basic operating principles it follows that, in constrast
objective is achieved with a fixed capacitor bank connected in parallel to the switched capacitor/inductor scheme, the fundamental current
with a thyristor-controlled "variable" inductor. and, consequently, the effective impedance of the thyristor-controlled
In the switched capacitor/inductor scheme the total compen- inductor compensator (Figure 13a) is continuously variable. On the
sating susceptance is split into an appropriate number of parallel other hand, as with the switched capacitor/inductor scheme, the
connected capacitor and inductor banks, each with a separate effective impedance and the compensating current can only be
bidirectional solid state (thyristor) switch. Thus, either capacitive or adjusted at discrete instants of time, not more often than once in each
inductive susceptance can be provided by closing and opening the half cycle. However, it should be noted that the current can be
switches of appropriate banks. The compensating susceptance can, changed from maximum leading to maximum lagging within one half
therefore, follow the load or terminal voltage variation in a step-like cycle.
manner. The technique of controlling the conduction intervals of the
The switching of both capacitor and inductor banks can be made thyristor switch generates harmonic current components as the
essentially transient free by choosing the instants of switching at the waveforms in Figure 13b indicate. For identical positive and negative
natural zero crossing of the current. This of course means that when current half cycles only odd harmonics are generated. The maximum
capacitor banks are switched "out", they remain charged to the amplitudes of the 3rd, 5th, 7th, 9th, 11th, and 13th are 13.8, 5.0, 2.5,
positive or negative peak of the ac voltage applied. Thus, in normal 1.6, 1.0 and 0.7 per cent respectively of the maximum inductor
operation a charged capacitor bank may be switched in when its current obtained with full condition (a = 0). These harmonics can be
voltage is equalled by the supply voltage. Consequently, the maximum kept out of the line currents by replacing the fixed capacitor with a
theoretical delay for switching "in" a capacitor bank is one cycle filter network that draws the same fundamental current at the system
(assuming that the capacitor bank is charged to the "wrong" polarity), frequency and provides low impedance shunt paths at the harmonic
and it is one half cycle for switching out. The maximum delay for frequencies.
1942
In a three-phase compensator the thyristor-controlled inductors AC
Va----
V
are delta connected. It is worth noting that under balanced operating System -
conditions (that is, with identical conduction angles) the compensat-
ing currents do not contain triplen (third, ninth, etc.,) hamonic
components since these then circulate inside the closed delta. . -t V5al Vbr vcf
LS C l Measuring, Computing, &
(cL Signal Processing Circuits
In comparing the switched capacitor/inductor scheme with the aL To Control iCI
thyristor-controlled variable inductor approach, the functional similar-
Va I
I
ity is evident. The major difference is that in the former the b V
impedance variation takes place in discrete steps, while in the latter it L c Measuring, Computing, &
Signal Processing Cimcuits
C
is continuous. Also, the thyristor-controlled inductor scheme can I C To Control I2
generally achieve a faster dynamic response. However, the practical
difference between the two schemes are considerable. The switched I
i
capacitor/inductor scheme requires a relatively large number of small I I Me~~~~~~~~~~asuring,
Computing, &
Signal Processinig Circuits
capacitor and inductor banks, each with a separate thyristor switch. I ~~Wrfa&e~~~~stancon~~~
p---- ----
upe -'
- To Control i(ci
The arrangement is therefore economical only at low voltage levels. In
medium to high voltage applications, a step-down transformer must be
used. The thyristor-controlled inductor scheme is inherently suitable Fig. 14 -A general load comipensation scheme using
for high voltage applications. Usually, only one high-voltage thyristor feedforward (computational) type control L .----
switch is used in each phase, and depending on the total rating, the
thyristor switch with the series inductor can be connected directly to
the ac lines. Consider, as an example, the use of Eqs. (26b) for control of the
Control Approaches compensating current, i(c) in phase ab. The relationship between the
amplitude of the compensating current, I(c)
ab = +/3V B(c), [first line
abE is [frs)ln
It has been seen that compensating current can be adjusted with
a practical "variable susceptance" only at discrete instants of time, of Eqs. (26b)].
once in each half cycle of the applied voltage. The means of control is
a solid-state (thyristor) switch that can be turned on ("fired") at will, I(C) = -K1ib(t) bcdt - 2i(t) vca
K (42)
but it can only be turned off when the current reaches zero (natural
commutation). However, to ensure transient-free switching, the firing
ab b()dvbc/d - 0-2a(t dvca/dt = 0
of the thyristor switch must be synchronized to the ac system voltage.
Using the switched capacitor/inductor scheme, the firing of the which, for simplicity, may be written in the form
thyristor switch must coincide with the peaks of the ac voltage. In the
thyristor-controlled inductor scheme the conduction of the thyristor
switch can be initiated anywhere in the "firing interval" stretching
Ia=b K1'b,bc K2 Ia,ca (43)
over 90 degrees from the peak to the subsequent zero crossing of the where lb,bc means the value of ib(t) at the time when dvbc/dt = 0
applied voltage. The final information in the switched capacitor/ (and vbc > 0), and Iaca means the value of ia(t) when dvca/dt = 0
inductor scheme is the number of capacitor (or inductor) banks to be (and vca >O). ac
connected to the ac line, and that in the thyristor controlled inductor
scheme is the delay angle at which the firing of the thyristor switch is The net compensating current provided by the fixed-capacitor
to be initiated. In both cases the operation of the control must be and the thyristor-controlled inductor is expressed as a function of
rigidly synchronized to the ac system voltages. delay (firing) angle a by Eq. (41). This in the present case is
Three basic control approaches are generally employed. The first
is a direct computational, so called "feedforward" method, which
repeatedly solves a set of appropriate steady state equations to find Iab(aab) = 'C,ab 'LF,ab((%ab) - (44)
the susceptances (or currents) required for compensation. The second where IC,ab is the current in the fixed-capacitor, and ILF ab(°ab) is
is a feedback control method, in which the compensating susceptances the fundamental current in the thyristor-controlled inducto'r. In order
are closed-loop controlled so as to reduce certain error signals. The
third approach may be termed hybrid; it uses a combination of the
feedforward and feedback control techniques. As was already indi-
to provide the current labc
the delay angle '%b has to be set so as to
satisfy the equality
cated the first approach is normally suitable only to control load
compensators. The second one is usually used to control compensators (45)
regulating terminal voltage. The third approach may be used to ab ab ab
control either type of compensator. These basic control approaches or, with the substitution of Eqs. (43) and (44)
are discussed below in connection with the compensator which uses
the thyristor-controlled inductor. They are, of course, equally
applicable to a switched capacitor/inductor type compensator. ILF,ab(atab) = 'C,ab + K1 'b,bc + K2 la,ca (46)
All feedforward (computational type) control approaches are Thus, in Eq. (46) the current of the thyristor-controlled inductor is
based on the fundamental presumption that the load (and the ac defined in terms of the fixed-capacitor current and components of the
system) is in "steady state" between any two consecutive instants of load currents flowing in phases a and b. The remaining problem is
time at which the current in the compensating susceptance is changed. simply to find angle aab that corresponds to the required inductor
Thus, between these time instants the relevant load currents (or current ILF,ab. In other words, the following equation
powers, or impedances) can be measured and, from these, the required
compensating current can be determined (computed) using appro-
priate steady state equations. Since the operation of each thyristor-
controlled inductor (representing, with the corresponding fixed
X(ab -
+sin 2aab)IC,a b +K +K2I a,c (47)
capacitor, one of the three variable susceptances) is synchronized to obtained with expression (40) from (46), has to be solved for caab to
the line-to-line voltage to which it is connected, it follows that the know at what time instant the thyristor switch is to be fired.
processing of each of the three steady-state equations, expressing the
required compensating current (susceptance) in terms of the quantities The above procedure indicates that the feedforward control
characterizing the load (currents, powers, or impedances), is carried scheme requires two main functional elements as illustrated by the
out over time intervals which are mutually displaced by a third of the block diagram of Figure 15: one to compute the desired inductor
period time of the ac system voltage. The three compensating currents current, the other to convert the desired inductor current to a
are essentially controlled independently of each other as shown corresponding firing angle.
schematically in Figure 14. Computation of the compensating currents The various steps in the control process discussed are illustrated
can be based on any relevant set of steady state relationships, such as in Figure 16. The complete cycle of control events are carried out in
those given by expressions (10), (21), (26), (27), (28) and (29). each half cycle of voltage wave vab which provides the timing
 1943
AC System AC System

Thyristor- a
Inductor Current IL Desired
Controlled To Firing Angle Inductor Current
Susceptance Computer ErrTo
Erronq Thyristor
PErroEsr
Converter
i Generator F ring Angle Controlled Load
Converter Susceptances

Hg. 17 - Major functional elements In a general feedback control scheme

Fig. 15- Major functional elementsin a general feedforward control scheme

nents, reactive power, etc.) and, second, to generate an error signal,

e by comparing the parametric value of the variable derived with the
rab reference signal Rq. The Error Processor provides an output signal, Eq,
which is generally a sum of two terms. The first is proportional to the
~~~~~~~~~~~~~~~ut error signal e and the second is proportional to the integral of eq.
(k-i) th The Error to riAng Angle Converter converts the processed error
"steady state" kth "steady state"
interval br
computation
interval fr
computation
signal, Eq, to the firing angle of the thyristor switch. The firing-angle
Ik-lith firing kth firing
generation must, of course, be synchronized to the appropriate
inter nterval line-to-line voltage and repeated in every half cycle.
lica
* . dt c_a The objective of the feedback control is to maintain some
parametric value of the variable at a desired level. However, some of
these variables, such as, line voltages, currents, and their relevant
-N constituent components, are periodic waves with fluctuating ampli-
tude and, often, considerable harmonic content. For these reasons the
choice of parametric characterization, and method of measurement is
important. For example, the amplitude or a sampled value of a voltage
or current wave would not be a suitable parameter for error signal
generation; the harmonics (and electrical noises) could cause consider-
able erratic variation in the variable controlled. This problem can be
minimized by calculating the amplitude of the variable from the
1-3 V waveform integrated over a suitable time interval. Another approach is
ab- 17sin
2
A
LF, ab (a)= -i-L (I 71
,
2
aab) to characterize the waveform by half-period average or rms values.
Average power quantities can be used readily for error-signal genera-
1.11 tion. Since the thyristor-controlled susceptance cannot respond
continuously to a control signal, only once in each half cycle, the
practical implementation of these averaging methods of error signal
generation is relatively simple and convenient.
Fig. 16 - Graphical Illustration of the feedtorward contro process using equatlon (47)
The feedforward and feedback control approaches are comple-
reference. Thus during the kth half cycle, considered in the Figure, mentary. The feedforward control is inherently stable and can be
current ia is sampled when vca reached its positive peak (dvca/dt = 0, made fast. The feedback control is inherently tolerant to changes in
vca > 0). The value la,ca so obtained is multiplied by K2 and, per Eq. the control elements as well as the controlled system and can be made
(47), added to a constant, IC,ab' that represents the magnitude of the very accurate. For these reasons, they are frequently used together in
current in the capacitor. When vbc reaches its negative peak (dvbc/dt = practical control schemes. For example, the accuracy of the desired
0, vbc < 0), current ib is sampled, the value, - Ib,bc obtained is inductor current to firing angle conversion can be improved and made
multiplied by K1 and added to the sum IC,ab + K2 Ia,ca to get the independent of circuit parameters by providing a negative feedback
from the actual inductor current as indicated in Figure 18. In another
desired magnitude of current, I(c) ab in the thyristor-controlled hybrid control scheme shown in Figure 19, a feedback control is used
inductor. At the time when vab reaches its peak (which defines the
earliest permissible firing) thea-function ILF,ab (<ab), as defined in
equation (40), is generated in real time. The intersection of l(c)
LF,ab
with ILF,ab(a) defines the delay angle, acb) at which the thyristor
switch needs to be fired.
The possibility (indeed, the necessity) of using a feedback
control approach for regulating the terminal voltage of an ac
transmission network by a shunt compensator has already been briefly Fig. 18- Use of negative feedback in an overall feedforward control scheme to increase the
discussed. Similar "closed-loop" approaches can, of course, also be accuracy of the" Inductor Current to Firing Angle Converter"
used to accomplish the objectives of load compensation (balancing
and power factor correction). In principle all feedback controls work
similarly. The thyristor-controlled inductor (providing with the fixed
capacitor the variable susceptance) is made to respond to an
appropriate error signal. The error signal represents the difference
between a chosen reference and a corresponding parametric value of
the variable to be controlled. Any change in the error signal results in
an opposing change in the effective susceptance value of the
thyristor-controlled inductor. This tends to keep the error constant
and close to zero. The block diagram of a typical feedback control
scheme is shown in Figure 17. There are three basic functional
elements in the control loop: the Error Generator, Error Processor,
and Error to Firing Angle Converter. The input signals to the control
are the line currents and/or voltages of the ac system. The function of
the Error Generator is, first, to derive a parametric value of the Fig. 19 - Use of fast feedforward load compensation in an overall feedback controlled
variable, q, to be controlled (e.g., the positive-sequence and negative- voltage regulation scheme
sequence current or voltage components, reactive line current compo-
1944
for overall terminal voltage regulation, while a feedforward control is REFERENCES
employed to provide fast compensation for a rapidly varying load
(e.g., an electric arc furnace). In a similar manner, it would of course 1. Frank, H. and Landstrom, B., "Power-Factor Correction With
also be possible to regulate the power factor, reactive power, etc. Thyristor-Controlled Capacitors", ASEA Journal, 1971, Vol. 44.
instead of the terminal voltage by accurate feedback control, while No. 6.
providing compensation for rapid dynamic load (or network) varia-
tions by feedforward control. 2. Jager, S. and Knuth, D., "Unterdruckung der Netzruckwirkungen
eines Lichtbogenofens durch eine Kompensationsanlage mit thy-
CONCLUSIONS ristorgeschalteten Drosseln" Electrowirme Int., 1972, 30 1972
(B5) SB 267-274.
Static, thyristor-controlled shunt compensators operate essen-
tially as variable reactances (capacitive and inductive impedances). 3. Johnson, E. R. et al, "Static High Speed Var Control for Arc
Functionally they can be divided into two classes: those which Furnace Flicker Reduction", Proceedings of the American Power
compensate for specific load disturbances and those which regulate Conference, 1972.
(support) transmission line terminal voltage. The general objectives in
the first class are dynamic load balancing and power-factor correction; 4. Reichert, K., Kauferle, J., and Glavitsch, H., "Controllable
those in the second class are terminal voltage balancing and regulation. Reactor Compensator for More Extensive Utilization of High
Thus, a load compensator is required to supply the negative-sequence Voltage Transmission Systems", CIGRE Paper No. 31-04, 1974.
load current and the reactive part of the positive-sequence load
current. A shunt compensator regulating terminal voltage is required 5. Barthold, L. 0. et al, "Static Shunt Devices for Reactive Power
to provide negative and positive-sequence currents to cause appropri- Control", CIGRE Report 31-08, 1974.
ate negative and positive-sequence voltages across the sequence
impedances of the ac system which will make the negative-sequence 6. Gyugyi, L., Pelly, B. R., "Static Power Frequency Changers",
terminal voltage zero and the magnitude of the positive-sequence John Wiley & Sons, Inc. 1976, pp 406-423.
voltage constant. 0 0
7. Olwegard, A. and Ahlgren, L., "Thyristor-Controlled Shunt
There are three basic approaches to the control of static shunt Capacitors for Improving System Stability", CIGRE Paper No.
compensators. The first is a direct computational ("feedforward") 32-20, 1976.
approach in which a set of equations defining the compensating
currents in terms of load and/or ac system system parameters are 8. Gyugyi, L., Otto, R. A., "Static Shunt Compensation for Voltage
solved at regular time intervals. The second is a feedback control Flicker Reduction and Power Factor Correction", Proceedings of
approach in which the compensating currents are closed-loop con- the American Power Conference, 1976.
trolled to maintain some parametric values of variables of interest at
desired levels. The third approach uses a combination of feedforward 9. Gyugyi, L., "Reactive Power Generation and Control by Thyris-
and feedback techniques. Feedforward controls are fast and inherently tor Circuits" IEEE Power Electronics Specialist Conference
stable. They are particularly advantageous in the compensation of Record, 1976.
large rapidly fluctuating loads which can be characterized by
(repeated) measurements. Feedback controls are accurate and tolerant 10. Meusel, W., and Waldmann, H., "Coordinate Transformations of
of changes in the control elements and the system controlled. They Multi-Term Regulation Systems for the Compensation and
provide a viable solution to the control problem for transmission line Symmetrization of Three-Phase Supplies", Siemens Forsch-U.
terminal voltage support, where the parameters of the ac system are Entwickl.-Ber. Bd. 6 (1977) Nr. 1.
neither accurately known nor measurable.

Laszlo, Gyugyi was born in Tamasi, Hungary on January 16, Semiconductor Division. In 1972, he joined the Switchgear Division as
1933. He received his basic technical education at the University of Sales Manager of Regulator Products and Power Electronics. In 1974,
Technology, Budapest, further studied mathematics at the University he was named Product Manager for that department, the position he
of London and electrical engineering both at the University of presently holds.
Pittsburgh, Pennsylvania, where he received an M.S.E.E. (1967), and
at the University of Salford, England, where he received a Ph.D. Mr. Otto is a member of the IEEE. He is also a member of the
(1970). Power Generation Excitation Systems Subcommittee.
He began his career at the Epsylon Research and Development Thomas H. Putman (SM '68) was born in Pittsburgh, Pennsylva-
Co., Ltd., England, in 1958, then joined the Westinghouse Research nia on Nov. 22, 1930. He received the B.S. degree from Union College,
Laboratories in 1963, where he is now Manager of the Solid-State Schenectady, New York in 1952, and the S.M. and ScD. degree from
Power Controls Section. In this position he has been responsible for Massachusetts Institute of Technology, Cambridge, Massachusettt in
the research and development of a variety of solid-state power control 1954 and 1958 respectively.
and conversion equipment, including static VAR compensation
systems for both industrial and utility applications. Upon completion of his work at M.I.T., Dr. Putman joined the
Westinghouse Research Laboratories as a Senior Engineer in the
Dr. Gyugyi is co-author of the book "Static Power Frequency Electromechanics Department and subsequently became manager of
Changers" (Wiley, 1976), and author of the section "Power Frequency that activity in 1967. In 1974 Dr. Putman was appointed Consulting
Changers" in the Electronics Engineer's Handbook (McGraw Hill, Engineer in the Electrical Sciences Division of the Westinghouse
1975). He has written several technical papers, holder of 32 patents, Research Laboratories.
and is a Fellow in the Institution of Electrical Engineers.
Dr. Putman's technical activities encompass the analysis and
Richard Otto was born July 15, 1932, in Watertown, Wisconsin. design of electrical, mechanical and hydraulic systems. He serves on
He received both a Bachelor of Business Administration degree in the Industrial Drives Committee of the IEEE and is also a member of
1954, and a BS degree in Electrical Engineering in 1959 from the the Society of Naval Architects and Marine Engineers serving on the
University of Wisconsin at Madison. Ships' Machinery Committee Panel M-20. Dr. Putman is a member of
Sigma Xi and is a Registered Professional Engineer in the State of New
He joined Westinghouse in 1959 and after several field sales York.
assignments, he became Applications Manager at that company's
 1945

Discussion defined in terms of the fundamental components of the load currents

and ac supply voltages, the measurement of these quantities over a time
interval shorter than one cycle is generally not possible because of the
J. D. Ainsworth (Power Transmission Division, GEC Switchgear Ltd., distortion of ac current and voltage waves encountered in a practical
Stafford, England): The paper demonstrates that by independent VAr system. There are a number of approaches available to overcome this
control at fundamental frequency, an unbalanced load can be made to problem, which include filtering, different estimation techniques, and,
look like a balanced constant-voltage load, viewed from the supply in particular, measurement of some parametric values, such as the
system. As the authors point out, this has been known for many years, average or rms value of the ac wave instead of certain instantaneous
but they give a good summary of this in their items 1-5, p. 3 of the values of its fundamental component. The success of these techniques is
paper. greatly dependent on the actual implementation, which is usually a
I found it disappointing that the authors only deal with fundamen- guarded secret of the manufacturer involved.
tal frequency for which, as they state, control is relatively easy. The All static shunt compensators presently used (including the
most interesting and important problems in control lie in the perfor- saturating reactors) have an inherent time delay (which creates an
mance at other frequencies; this affects e.g. sub-synchronous stability ultimate limit on the overall response time) due to the fact that their
when used with a transmission line, and also the associated subject of VAR output can only be changed at discrete instants of time. This time
transient response, e.g. when used with arc-furnaces. We have found it delay is a function of the circuit complexity of the compensator and is
necessary in practice to develop design and analytical methods effective- thus subject to economic trade-off.
ly covering all frequencies, including sub-synchronous, both for the Since static shunt compensators are basically controllable VAR
thyristor-controlled-reactor and the multiphase saturated reactor (e.g. sources, whereof the reactive output current is primarily controlled, the
treble-tripler) type of static VAr compensator, allowing for the discon- ac supply network (to which the compensator is connected) inevitably
tinuous nature of the devices. The saturated reactor, due to its extreme- plays a significant part in the transient behavior and stability of the
ly fast response, can be very satisfactorily represented for analysis by an overall ac system. The impedance characteristic of the ac supply net-
equivalent even simpler than the approximation given by the authors work can affect the sub- and super-synchronous behavior of the com-
for the thyristor-reactor; the latter presents considerably more difficul- pensator considerably, thus influencing both the control and filter net-
ty, and we have not so far found any simplification possible in its work design.
analysis for sub-synchronous frequencies, other than full representa- Due to the complex interactions between the ac supply network
tion. What are the authors' views on this? and the compensator, particularly when unbalanced system conditions
Hardware models are an alternative, though with limitations. As ex- are also considered, a full representation of the total ac system (com-
amples, we have found that for extreme system conditions of high pensator and supply network) is usually necessary, as Dr. Ainsworth in-
source impedance and shunt capacitance, the saturated reactor may re- dicates, for reliable analysis. For this purpose we use digital and hybrid
quire extra main filters, but with the thyristor-reactor, stability can computer simulations extensively, together with actual hardware
sometimes be obtained (at zero cost) by control system modification, modeling. The simulation and modeling normally includes all three
though at the expense of response time. We have found in some model phases of the ac system. In our opinion, simplified single phase
tests in conjunction with an h.v.-d.c. invertor on a particularly weak representation and modeling may, at times, be satisfactory to evaluate
system, that with a saturated reactor the system can be stable, but with a transmission line compensation; however, the same approach usually
thyristor-reactor it was not stable with any control changes that we gives misleading and generally optimistic results when used to study arc
could find. We admit we do not understand this yet, but we do not furnace compensation, regardless of whether the compensator is the
think it is particular to the controls used. I would conclude that we all saturating reactor type or thyristor-controlled. We wish to point out
require good engineering rules of thumb for static compensator design, here that we did not intend to give any approximate representation for
and much remains to be done. the thryistor-controlled compensator in the paper. We showed a
number of possible control approaches and illustrated (pages 8 and 9),
Manuscript received February 22, 1978. as an example, the major steps required to implement a typical "feed
forward" type control using a particular set of equations derived in the
paper.
L. Gyugyi, R. A. Otto, and T. H. Putman: The purpose of the paper is We cannot concur with Dr. Ainsworth's experimental findings
to provide a broad theoretical basis for, and define the general objec- which appear to indicate that a particular ac system (with HVDC in-
tives of static shunt compensation in different applications and to show verter) can be made stable with a saturating reactor, but not with a
that these objectives can be accomplished analytically. The thyristor-controlled reactor compensator. The inherent control
mathematical relationships derived can serve as a basis to formulate mechanism of the saturating reactor (which attempts to maintain cons-
overall control strategies for thyristor-controlled shunt compensators. tant volt-seconds for each half cycle of the ac terminal voltage) can be
However, it has been pointed out (paragraph 2, page 7) and is restated imitated by the thyristor-controlled rector. Assuming similar ratings
here, without apology, that the numerous design problems of a prac- and control ranges for both types of compensators, there is no apparent
tical control scheme, such as stability, speed of response, measurement reason to assume that their effect on the stability of the ac system would
of the variables, signal processing, etc., are not subjects of the present be different.
paper. It may be appropriate to mention here that considerable work on
The control difficulties of a static shunt compensator are greatly static shunt compensation of transmission lines subject to subsyn-
influenced by the following factors: (1) measurement of. the variable chronous resonance has been carried out at the Research Laboratories
controlled, (2) the overall time delay of the compensator, and (3) the with which the authors are associated. Although the studies are not ful-
impedance characteristic of the ac power transmission network com- ly completed yet, the results so far clearly indicate that a properly con-
pensated. trolled static shunt compensator always tends to damp sub-synchronous
The difficulties in measuring a changing variable (current or oscillation.
voltage) are due to the short time interval over which the measurement
usually must be executed. Although the compensation requirements are Manuscript received April 24, 1978.