1238 Transactions on Power Systems, Vol. 6, No.

3, August 1991

A NEW APPROACH TO AC/DC POWER FLOW

T Smed, Student Member G Andersson, Member G.B. SheblC, Senior Member L.L. Grigsby, Senior Member
Dept. of Electric Power Systems Dept. of Electrical EnBineering
Royal Institute of Technology Auburn Universit
S-100 44 Stockholm, Sweden Auburn, A1 3 6 8 3 0 , 6 S A

Abstract - A new approach to ac/dc power flow is presented Additonal insight into the interaction between the ac
in this paper. The converters are treated as voltage dependent and dc parts of the system can be gained. This is used
loads and the dc variables are eliminated from the power flow in [B], where the short term power/voltage stability of
equations. A simple and reliable method for fast decoupled dc converters is analyzed. Another example is the con-
power flow for hybrid ac/dc systems is developed. For non- cept of area-decoupling, described in this paper.
synchronous interconnections, the power flow equations can
be solved separately for each area. DC SYSTEM MODEL
The equations describing the steady state behavior of a
INTRODUCTION monopolar dc link can be summarized as 'follows [ 3 ] .
Traditionally, two different approaches have been used to solve
3
the power flow equations for hybrid ac/dc systems. The first VjR = *aR&R cos CYR - - XcId (1)
approwh is the sequential method [ 1 , 2 , 3 ] ,in which the ac and
dc equations are solved separately each iteration. The sequen- 3
vdr = *ar&r cos71 - - xcI,i (2)
tial method is easy to implement, but convergence problems
may occur in certain situations. The other approach is the VdR = +
vir RdId (3)
unified method [ 3 , 4 , 5 , 6 , 7 ] , in which the solution vector is ex- PdR = VdRId (4)
tended with the dc-variables. Because of this, earlier unified Pd, = K1rIa (5)
methods will be referred to as extended variable methods in
this paper. The drawback with the extended variable method 3Jz
S ~ R= k-aR&d,i (6)
is that it is complex to program and hard to combine with
developments in ac power flow solution techniques, such as Stir = k---ar&rId
3Jz
(7)
the fast decoupled method.
The eliminated variable method presented in this paper
overcomes these difficulties. The basic idea is to treat the
real and reactive powers consumed by the inverters as voltage
dependent loads. The dc equations are solved analytically where k is assumed constant, k zz 0.995, see Appendix A.
or numerically and the dc variables are eliminated from the
power flow equations. The method is unified, since the effect
of the dc-link is included in the Jacobian. It is, however, not
an extended variable method, since no dc variables are added
to the solution vector. The new method has the following
advantages:
A reliable method for fast decoupled power flow can
easily be developed based on the new method.
The ac and dc parts of the power flow can in principle
be handled separately. This simplifies the implementa-
tion, maintenance and modification of the program. In Figure 1: Model of dc converter.
particular, it is straightforward to include dc-links in
an existing ac power flow program.
AC/DC POWER FLOW
It is easy to handle switching between control modes
due to variables hitting limits. When the dc-link is included in the power flow equations, only
the mismatch equations at the converter terminal ac buses
have to be modified.

90 SM 399-6 PWRS A paper recommended and approved A P ~ R = P:iec - P z ( @ , V ) - P ~ R ( V ~ R , K I ,(10) X~~)

by the IEEE Power System Engineering Committee of the = p ~ ~ C - p : I C ( @ , V ) + P d I ( V I R , V l ~ , X(11) ,~c)
IEEE Power Engineering Society for presentation a t the
IEEE/PES 1990 Summer Meeting, Minneapolis, Minnesota, A Q ~ R = QiFc- Q$$(@,v) -Qd~(Vt~,Vtr,Xd~)(12)
July 15-19, 1990. Manuscript submitted September 25, A Q t r = 9::'"" - Q S ( @ , V ) - Q , i r ( V l n , V l r , x , i c ) ( 1 3 )
1989; made available for printing April 2 4 , 1990.
where xdc is a vector of internal dc-variables. The dc-variables
satisfy
R(&R,V l r , x d c ) = 0 (14)
0885-8950/91/0700-1238W1.00 Cp 1991 IEEE

I 1 1-
 1239

THE ELIMINATED VARIABLE METBOD

Table 1: Control Modes
In the eliminated variable method, the equations in (14) are,
Control Specified in princzple, solved for X d e .
Variables
xdc = f(&R, &I) (17)

The real and reactive powers consumed by the converters can

then be written as functions of &R and Vir.

PdR = P d R ( K R ,& I , x d e )
= PdR( & R , &I , f (K R , V I ) )
= PdR(&Rr & I ) (18)

It is not needed to derive explicit functions for the real and

reactive powers, only to find a sequence of computations such
that the real and reactive powers and their partial derivatives
w.r.t. the ac terminal voltages can be computed. If all real
and reactive powers are written as functions of &R and Vir,
(15) can be replaced by (19).

where R is a set of equations given by (1)-(3) and four control

specifications. where
In the extended variable method, (15) is solved iteratively.

AV/V (15)
AQt
AR
In the sequential method, (14) is solved after each iteration
of (16).
L’ is modified analogously. Thus, in the eliminated vari-
able method, four mismatch equations and up t o eight el-
ements of the Jacobian have to be modified, but n o new
Control Modes variables are added to the solution vector, when a dc-link
is included in the power flow. The partial derivatives are
Seven variables and three independent equations, (1)-(3), are those required by (19); a P d R ( & R , b ) / a & R , for example, is
introduced when a dc-link is included. Hence, four specifica- the derivative of PdR w.1.t. VtR, assuming Vir is kept con-
tions have to be made in order to define a unique solution. It stant. The dc variables, however, are not kept constant as
is not possible to cover all conceivable control modes in this opposed to a P d R ( & R , & I , X , l e ) / a & R , which is used ’in (15).
paper, but those given in Table 1 will suffice to illustrate the Although (19) looks like (16), it is mathematically more sim-
analytical elimination procedure. ilar to (15). The Jacobian in (19) is however normally more
Control mode A is the base case, which in the well known well-conditioned than the one in (15), cf. Table 4.
current margin control corresponds to one terminal control-
ling the voltage and the other the current, or equivalently the Analytical Elimination
power. The control angles and the dc-voltage are specified,
To illustrate the procedure, the analytical elimination is car-
and the converter transformer tap positions are varied in or- ried out in detail for some representative modes. It is suffi-
der to meet these specifications. The other modes in Table 1 cient to find Pd and Sa at each converter, since Q d then can
are obtained from mode A if variables hit their limits during be computed with (8) or (9). The partial derivatives for all
the power flow computations, or if the time scale is such that
modes in Table 1 are shown in Tables 2 and 3.
the taps can be assumed to be fixed. Which modes that are Control Mode A: [ C Y R7 1 Vir P u ]
obtained when limits are encountered depend on the control
Since both the voltage and power at the inverter are specified,
strategy of the HVDC-scheme, and this must be accounted for
the direct current can be computed with (S), and P d R can then
in the computations. For modes B - D, a R determines CYR
be found by combining (3), (4) and (5)
and ar determines the direct voltage, which normally is the
case for current control in.the rectifier. For modes E - G, aR
determines the direct voltage. Subscript I refers to constant
current control. If we combine ( l ) ,(6) and (24), we obtain
The taps are assumed to be continuous variables. Discrete
tap positions can be taken into account by first assuming con- SdR = k
+ + ;xc,1:
PdI (&I
tinuous taps and subsequently fix the taps at appropriate val- COS f f R

ues. = ka(p,ir + PL + Q L )

I
1240

Analogously, for &I: QdR and its partid derivatives are computed from (25):

(39)

Thus, d real and reactive powers consumed by the converters (40)

can be precomputed, and including the dc-link in the power
flow is trivial for this control mode. The same is true for any
specification of the form [ C Z R Y r X I 221, where 11 and 2 2
are any two variables of [ P ~ RP ~ IV ~ R PdI Id]. The fact Control Mode Er: [(YR aR 71 P,II]
that the real and reactive powers can be precomputed for this This mode occurs e.g. if the tap changer at the rectifier hits a
case is well known; it has e.g. previously been observed in limit in control mode A under current control in the inverter.
[2], where a sequential method is used to solve the power flow Combining (1) and (4) gives
equations for other control modes.
Contrd Mode B: [aR Y I V d I PdI] PdR
This mode occurs e.g. if the the tap changer at the rectifier
hits a limit in control mode A under current control ip the
rectifier. Since PJI and are specified, Id, and PdR
and S ~ areI computed as for mode A. Since aR is specified,
S,{R is computed with (6) instead of (25):
(44)

P,{I is computed by (24), and the partial derivatives are iden-

tical to those for P d R . Since aR is specified, S,{R is computed
with (6), which combined with (43) and (8) give

The formulas for mode Br are essentially identical; the only

difference is that P < ~ rather
I, than Id, is computed with (5).
In general, when two of the variables [ P ~ R P , ~ IV ~ R
V ~ IId ]
are specified, the other three can be computed from (3)-(5).
Control Mode C: [ a R 71 ar Pd11 (47)
These specifications are valid e.g. if the tap changer a t the
inverter hits a limit in mode A under current control in the S d I is computed with (26), and we obtain V ~ is
R given by
rectifier. Combining (2) and (5) gives

If we solve for Id, we obtain

It1 = C1 &I - J(C1 &;)' - C2 PdI (30)

Other Modes
The partial derivatives for the other control modes can be
derived analogously; if the tap changer controlling the con-
where trol angle is specified (modes B , D, F, G) , only the reactive
power at that converter will depend on corresponding ac volt-
age.
If the tap changer controlling the direct voltage is specified
(modes C,D, E, G), aU real and reactive powers will depend
on the ac voltage at that terminal. Eqs. (30) or (52) are used
Define aII as to find the direct current for constant power control, and (42)
arI = -- aid (or its counterpart for constant 71)is used to compute the
(34)
I d a% power for constant current control.
Since P,iI is specified, both its partials are zero. P ~ iR
s given If the tap changer position is specified a t a converter, S,I is
by (24), and its partial derivatives by: computed with (6) or (7), otherwise (25) or (26) are used. The
partial derivatives for all modes in Table 1 are summarized in
(35) Tables %.and 3. a I R in Table 3 is defined as

Since a1 is specified, S ~ isI computed with (7), and the partial where
derivatives of Q ~ are
I given by

(37) (53)

(38) c4 = ir +3
Rd A xc (54)

-r-[
 1241

Table 2: Partial derivatives for modes with t h e direct voltage determined by a l .

LiI
FI 0 0 0 0 0 0 o g ClI

G 2PLaIR si,( 1 +~ I R-)2PLaIRPJR

QJR 0 0 ~ I R 0 % ClI
Gr Pm-kQL QdR -~ Q L 0 0 PdR-kQL - PdR -k Q L pJ
Q

Numerical Elimination combined with a fast decoupled method. The numerical elim-
A drawback with the analytical elimination is that the for- ination does not assume anything about the model. In fact,
mulas have to be rederived for other dc system configurations the method is not restricted to dc systems; any voltage de-
or if other specifications are used. In this section, a com- pendent load can be included.
putational scheme, which avoids the analytical derivation, is Fast Decoupled Power Flow
proposed. For given terminal voltages, the real and reac-
tive powers consumed -by the converters can always be com- Previous methods for fast decoupled power flow for ac/dc sys-
puted, at least numerically. In the numerical elimination pro- tems can be classified into two groups. In the extended vari-
cedure, the required partial derivatives are computed with able approach, the dc-part is solved together with the reactive
finite differences. Let P,IR(&R, K r ) denote a function that power portion of the power flow. The drawback with this ap-
solves the dc equations and computes PdR, given the termi- proach is that the Jacobian cannot be completely preinverted.
nal voltages. Equation (55) then gives an O ( h )approximation The schemes that have been proposed to deal with this prob-
to BPJR(V;R, V t r ) / a V t ~as, required by (19). lem lead to increased complexity in the implementation [3,7].
In the sequential approach, the effect of the dc link is only
R, -
~ P , I R ( &h) & I + h ) - P ~ R ( & R&,I )
- P,~R(V;R, included in the mismatch equations. The drawback is that
a&I h (55)
convergence problems may occur.
The choice of h is a trade-off between the truncation error The eliminated variable approach overcomes the two major
and the computational precision. As a rule of thumb, h can drawbacks with previous methods. Based on (19), a fast de-
be chosen as h zz m, where to1 is the precision by which coupled power flow for the combined ac/dc system can easily
PdR is computed. For the numerical examples in this paper, be obtained. The effect of the dc-link is taken into account
to1 x lo-' and h = The errors introduced by (55) are only in the reactive power part of the approximated Jacobian,
minor as compared with the errors in the sequential method; and constant approximations of the partial derivatives of the
they are in the order of h in our example) while the reactive powers w.r.t. the terminal voltages are computed
sequential method can give significantly larger errors (S:/Qd, with some reasonable assumptions on the terminal voltages.
for example, may be greater than 2 pu, d Tables 2 and 3). Numerical experiments indicate that the selected values on
If the numerical elimination is combined with a Newton the terminal voltages are not critical. Switches between the
method, (55) is used to compute an approximation to the control modes are discussed in a later section.
derivatives each iteration, while it is used only once when

r - 1 I
1242
Area-Decoupling voltage can be set to some arbitrary value, since the analgt-
ical derivation has shown that the value of this voltage does
One great advantage with dc transmission is that power can not affect the first area’s power flow. There are no approx-
be transmitted between non-synchronous networks (this in- imations involved in the decoupling. It is only a matter of
cludes networks with different frequencies). In order to solve computing quantities in a certain sequence in order to sim-
the power flow, two ac systems have to be handled simul- plify the implementation of the program and minimize the
taneously, i.e. two slack buses and two angle references are computational effort.
needed. If the dc link connects two large networks, the inclu-
sion of the dc link in the power flow increases the number of S w i t c h i n g B e t w e e n Control Modes
variables drastically. On the other hand, if the dc link feeds
power from a large network to a weak island network, con- If a dc variable, e.g. a tap changer or a control angle, hits a
vergence problems for the island network can slow down the limit during the power flow computation, a switch to another
convergence of.the whole system considerably. Thus, a proce- control mode is neccessary. In the Newton Raphson power
dure to decouple the ac networks in the power flow is desirable flow, this is a minor problem if the eliminated variable is used,
for all non-synchronous connections. The networks are cou- since the components of the solution vector are the same.
pled via the voltage magnitudes at the converter terminal ac For the fast decoupled power flow, hitting a limit normally
buses, as shown in (56). makes a refactorization of the reactive power portion of the
approximated Jacobian neccessary. The fact that a maximum
of four elements of the reactive power part of the approxi-
mated Jacobian can change only makes a partial refactoriza-
tion neccessary. If area-decoupling can be used, the switch
is trivial if the corresponding mismatch equation is placed
last; the appropriate value of K a Q , i / a K only has to added
and subtracted to the lower right element of L in the LU-
decomposition. If area-decoupling not can be performed, we
can keep the simple refactorization by making suitable ap-
proximations, cf. the numerical example.
where There are some similarities in the handling of switches in
a P d R , etc.
N:, = --~ A P ~=R- the new method and the extended variable based method for
ah1 OK, fast decoupled power flow proposed in [3], but we want to
The only coupling between the systems are through the ’cross point out the following differences:
derivatives’. If both the real and the reactive power a t the 0 In [3], a rather large partial refactorization, including
inverter end is independent of the of the ac voltage a t the all dc equations has to be performed every iteration. In
rectifier end, N i R and LhR will both be zero, and the power the eliminated variable method, partial refactorization
flow for the ac system on the inverter side can be solved inde- is neccessary only when a variable has hit a limit.
pendently of the ac system on the rectifier side. Furthermore,
once the ac voltage on the inverter side is known, the power 0 If area-decoupling or the approximated fast decoupled
flow on the rectifier side can be solved using the known value ac/dc power flow is used, the refactorization is trivial.
of V t I . In the new method, all variables in the solution vector
In Table 2, where the cross derivatives are between the dou- and all mismatch equations remain the same before and
ble ruled vertical lines, we can see that P d I and QdI do not after a switch. In [3], either some variables or some
depend on V ~ for R any of the control modes A - D. Thus, mismatch equations has to be replaced when switching
when a1 determines the direct voltage, the ac system on the between control modes.
inverter side should be solved first. On the other hand, if
aR determines the direct voltage, the ac system on the rec- Bipolar and 12-Pulse C o n v e r t e r s
tifier side should be solved first as can be seen in Table 3.
In general, the power flow of the the system at the voltage The modeling of 12-pulse converters is straightforward. Let
determining terminal should be solved first. ng denote the number of 6-pulse bridges. Then, if Kt is re-
The advantages with the area decoupling can be summa- placed by V d / n B in (1)-(2) and the right hand sides of (6)-(7)
rised as follows. are multiplied by n B , all previous results hold.
The power flow can be broken into smaller subproblems, Bipolar converters could be considered as two independent
since there is no need to handle two non-synchronous dc links, except for the coupling via the ground return resis-
networks simultaneously. tance, cf. (3):

It will suffice to do the minimum number of iterations vdR1 = VdIr + RdIdl + Tg(Id1 - Ida) (57)
for each network. This can give considerable savings in
computational burden. The partial derivatives differs to those from Tables 2 and 3
in a non-trivial way only for the not very common case of
Fewer elements of the Jacobian have to be modified,
unbalanced operation in constant power control, with non-
since the terminal voltage for the first area is known
zero ground return current (this does not include the case
when the power flow for the second area is computed.
when one pole is out of operation, since the remaining pole
The area decoupling can be used with any method. If the can be treated as a monopole). The modification for this case
method uses the terminal voltage on the oppposite side to consists of taking into account the extra terms introduced by
compute the real and reactive power for the first area, that r g (Ita - Id2) in (57).

-1
 1243
Multiterminal Systems
Table 4: Suiiiiiiary of power flow results for m o d e F.
As for the bipolar converters, multiterminal dc systems can
uI/u,,denotes t h e condition number of t h e Jacobian.
be directly incorporated in the numerical elimination method;
Maxiiiiuiii iiiisinatch = 0.001 pu.
only the analytical elimination procedure has to be modi-
fied slightly. We will indicate how the analytical elimination
can handle multiterminal systems by considering the system
Newton I a Itera- kflops 1
shown in figure 2 . Assume that the current margin method
alnorithm
I
I tions I
ac Power Flow I 125 3 17
Analytical I 125 3 18
Numerical I 125 3 19 I
Id2 1+, I . +
Seauential I diverges
" I
J
Extended I 237 4 27
L
Extended I1 266 4 28
-1-

Fast decoupled -
u1 Itera- kflops
algorithm I U,& tions
ac Power Flow 1 26 4 10
Analytical Elim. I 26 4
Figure 2: Multiteriiiinal system.
Numerical Elim.
Approximated Elim.
I 26
26
4
4
11
11
I
is used to control the dc system [lo] and that converter 3 is Sequential 35 7 17
the voltage setting terminal. Then [ I d 1 I d 2 7 3 a1 7 2 v d 3 ] Extended I 215 4 12
are specified. The three latter variables are controlled with
the tap changers. As in the two-terminal system, all partial
Extended I1 1 247 4

derivatives are zero for this nominal case.

If a1 hits its limit, a 1 is no longer specified and, as a con-
sequence, only a Q d l / a v t , will be nonzero, cf. mode B . Sim-
The number of iterations given for area-decoupled algo-
ilarly, if a2 hits its limit, V t 2 a Q d z / a & z = S j , / Q d 2 . If a3 hits
rithms is for the larger area, since it carries the major part
its limit, 11,s and consequently all real and reactive powers
of the computational burden. The advantage of the area-
will depend on V;3.
decoupling is illustrated in the power flow for mode Dr;when
Clearly, the pattern from Tables 2 and 3 is similar for mul-
area-decoupling is used, the mismatch for the larger area was
titerminal systems. If the ac systems are non-synchronous,
5.10-' after 3 iterations, otherwise 5 iterations were needed
the power flow for the system at the voltage setting terminal
to obtain a mismatch of T l O - ' .
should be solved first.
It is clear from Tables 4 and 5 that the methods presented
NUMERICAL EXAMPLE in this paper can decrease the computational effort without
complicating the implementation of the ac/dc power flow.
The IEEE 14-bus test system was used in a numerical exam-
ple. Since convergence problems for weak receiving systems CONCLUSIONS
have been observed, a two bus system with a short circuit ratio
of 3.3 was connected to bus 14 with a dc link. The power flow A new method to include dc systems in power flow calcula-
was run for all control modes and for all implemented algo- tions has been proposed. The method is more efficient than
rithms, including variations in terms of fast decoupled power previous methods and more importantly, the method is easy
flow and area decoupling. If sparsity techniques are used, to implement and developments of ac power flow solution
the computational burden per iteration is approximately the techniques can be combined with the method. In particular,
same for all algorithms. The number of iterations and the fast decoupled power flow can easily be implemented.
approximate number of floating point operations for mode G The main idea of the new approach is to treat the convert-
are given in Table 4 for each algorithm. An ac power flow for ers as voltage dependent loads and eliminate the dc variables,
area 1 has been included for reference. The average number either analytically or numerically. Further insight into the in-
of iterations for each algorithm when solving the power flow teraction between the ac and the dc networks can be gained
for all control modes is shown in Table 5 . The variables used through the analytical elimination. One example is the con-
in extended I [3] and extended I1 [ 9 ] are given by X I and x 2 , cept of area-decoupling, which allows the power flow for the
respectively. Extended I1 is similar to the schemes in [5,6] in two ac networks in a non-synchronous interconnections to be
the sense that the overlap angles are used as variables. In [ 5 ] , solved independently.
the problem is solved only for constant power factor (cos 4 ) . The numerical elimination is completely general and di-
In [ 6 ] , a, U and a are specified at each converter to give six rectly applicable to any dc system configuration. Further-
specified variables in total, which are two too many. more, a sequential method can be modified to the numerical
elimination scheme with a few lines of code.

ACKNOWLEDGEMENT

In Table 5 , no distinction is made between analytical and This work was sponsored by ABB Power Systems, the Ful-
numerical elimination, since the maximum mismatch differed bright Commission and Vattenfall, all of Sweden. Thanks to
only in the fourth or fifth decimal for all tested cases. The T. Adielson and G. Liss, ABB Power Systems, for valuable
approximated fast decoupled power flow method only includes comments on dc system characteristics.
the ' S j / Q d terms', cf. Tables 2 and 3 .

1 7- I
1244
APPENDIX A:LIST OF S Y M B O L S
Table 5: Average number of iterations.
vd Direct voltage
Direct current

I 1 :3!* 1
Newton Method I Iter fl Id
Converter transformer tap ratio
Eliminated, Area-dec. 1 3.0’
a
a! Firing angle
Eliminated variable I 3.4’ Overlap angle
U
Sequential,+Area-dec.+
Y Extinction angle
Extended I1
pd Real power consumed by the converter
Extended I
Qd Reactive power consumed by converter and
Sequential transformer

I 11
Fast Decoupled Method sd Magnitude of complex power consumed by
converter and transformer
PL pL = RJ;
11 Eliminated. I 4.2* 11 QL QL = $XCI;
Sequential, Area-dec. 42:
Approx. Elim. k I,, =q k I d , k z 0.995, see [3].
Extended I1 ka k, = k / c o s a
Extended I k, k, = k / cos y
Sequential XC Commutating reactance
Rd dc line resistance
’ Did not converge for all modes. The listed number is based V
0
Vector of nodal voltage magnitudes
Vector of nodal voltage angles
on the cases, for which the sequential method converged.
New computational schemes based on methods in this paper. AP Vector of real power mismatches
AQ Vector of reactive power mismatches
Specified real power at converter terminal
References QiyeC Specified reactive power at converter terminal
Pp“ Real power transmitted by the ac network
J. Reeve, G. Fahmy and B. Stott, ”Versatile Load
Flow method for Multiterminal HVDC Systems”, IEEE
Q:” Reactive power transmitted by the ac network
a11 is defined by (34) and ~ I isR defined by (51)
Transactions on Power Apparatus and Systems, Vol. All quantities are given in pu. Subscript t refers to the con-
PAS-96, NO 3, 1977, pp. 925-933. verter ac terminals. Subscript R refers to the rectifier and I
H. Fudeh and C.M Ong, ”A Simple and Efficient AC- to the inverter.
DC Load-Flow Method for Multiterminal DC Systems”,
IEEE h n s a c t i o n s on Power Apparatus and Systems, APPENDIX B: P U SYSTEM
Vol. PAS-100, NO 7, 1981, pp. 4389-4396. The ac system pu base values are:
J. Arrillaga, C.P. Arnold and B.J. Harker, ”Computer SE Three-phase power base
Yodelling of Electrical Power Systems”, John Wiley and VB Line-to-line voltage base
Sons, 1983. IE Current base, SE = & VBIB
T. Adielson, ”Modelling of an HVDC System for Dig-
ital Simulation of AC/DC Transmission Interactions”, The dc system pu base values are:
CIGRd Symposium 09-87, Boston, 1987.
s$ = SE
D.A. Braunagel, L.A. Kraft and J.L. Whysong, ”Inclu-
sion of dc Converter and Transmission Equations Direct-
v& = VB
ly in a Newton Power Flow”,IEEE Trans. on Power Ap- I$ = &I,
paratus and Systems, PAS-95, No 1, 1976, pp. 76-88.
BIOGRAPHY
G.B SheblC and G.T. Heydt, ”Power flow studies for Sys-
Thomas Smed (S’88) was born in in 1961 and received his
tems with HVDC Transmission”, PICA Conference,
Civ. Ing and Tekn. Lic. degrees from Uppsala University,
pp. 223-228, New Orleans, La, 1975.
Sweden and the M.S.E.E. degree from Auburn University. He
M.M. ELMarsafawy and R.M. Mathur, ”A New, Fast is a graduate student at the Royal Inst. of Tech., Stockholm.
Technique for Load-Flow Solution of Integrated Multi-
terminal dc/ac Systems”, IEEE Trans. on Power Appa-
GZiran Andersson (M’86) was born in 1951 and received his
ratus and Systems, PAS-99, No l, 1980, pp. 246-255.
Civ. Ing and Ph.D. degrees from Lund Institute of Technol-
T. Smed, G .Andersson and B. Franken, ”Analysis of ogy, Sweden. He is Professor and Head of Dept. of Electric
Power/Voltage Stability of dc Converters in ac Systems”, Power Systems, Royal Inst. of Tech., Stockholm.
submitted to the IOth Power System Computation Con- Gerald B. Sheblk (M’71-SM’85) received his B.S. and M.S.
ference, Graz, Austria, 1990. degrees from Purdue University and his Ph.D. from Virginia
T. Smed, “A New Approach to ac/dc Power Flow”, Mas- Tech. He is an Associate Professor of Electrical Engineering
ter’s Thesis, Auburn ,University, 1989. a t Auburn University.
G. Liss, ”Multiterminal HVDC Design with Emphasis L.L. Grigsby (M’57-SM’76) received the B.S.E.E. and
on the Need for Special Control Functions”, 11 Sympo- M.S.E.E degrees from Texas Tech. , and the Ph.D. from Ok-
sium of Specialists in Electric Operation and Expansion lahoma State Univ. He is the Georgia Power Distinguished
Planning, Sao Paulo, Brazil, 1989. Professor of Electrical Engineering at Auburn University.

T‘ -1 -