IEEE Transactions on Power Systems, Vol. 8, No.

2, May 1993 701

AC-DC LOAD FLOW WITH UNIT-CONNECTED GENERATOR-CONVERTER INFEEDS
J. Arrillaga, C.P. h o l d J.R. Camacho s. sankar
Fellow Sen. Member Student Member Non-member
University of Canterbury, Universidade Federal University of Canterbury
Christchurch, de Uberlandia, Christchurch,
New Zealand Brazil New Zealand

-
Abstract It is demonstrated that the conventional ac-dc
formulation is not directly applicable to the unit-connected
generator-converter. Alternative steady state models are derived
which are considered reasonable for cylindrical rotor machines
but not for salient-pole generators. F i n d an accuratealgorithm
of general a pliability, called the &quivalent Inverter, is
proposed whicg uses unit-connectioncharacteristicsderived from where 8, is the converter transformer leakage susceptance and
a time domain simulation. k in equation (1) is very close to unity.
Since the bus between the converter and the transformer
INTRODUCTION is not required, equations (l),(2), 4) and ( 5 ) can be combined
and the terminal ac busbar voltage ti',.) used as a reference. A
The connection of remote power generating plant via long reduced set of variables and e uations results[5] suitable for
distance HVdc transmission has already been proved extremely investigay the effect of mntrOP strategies in current use. The
reliable in many projects. In an attempt to simplify the set of varia les is:
conventional design, the use of unit connected enerators is
being serious1 considered at present[l-4]. C I G b Working
Group WG 1ZO9 is also invesbgating the unit connection. x
= [V, , I , , a , cosa , #IT (8)
Earlier references on this to ic have discussed the Each converter terminal is therefore representedby a five
potential cost savings and operation8 reliability of the unit- variable set of non-linear equations, which are then combined
connected generator-HVdc converters in isolation from the with the ac system e uations and solved using either
receiving end power system. However final technical and simultaneous or sequenti2 iterative algorithms[51.
economcal comparisons with the conventional configuration will
need to consider the overall system performance. In particular, -tion for use with the unit-connectioQ
the effect of the alternative scheme in ower transmission
capability and system stability must be careblly assessed. In the absence of filters the voltage at the converter
Although the ac-dc load flow is well understood[5], some terminal is not sinusoidal and can not be used as the
of the basic assumptions made in its formulation have been commutating voltage in the conventional formulation. Instead,
found inapplicable to the unit-connected configuration[6 the use of the generator internal EMF behind sub-transient
This paper establishes the modifications required )or thereactance E") is commonly accepted.
continued use of the conventional ac-dc load flow formulation
and then describes a more general and accurate algorithm,based
6
In t e case of a non-salient rotor machine the load flow
formulation described in the previous section should be
on the use of converter characteristics derived from dynamic applicable by shifting the converter interface to a fictitious
simulation. terminal (the internal EMF behind sub-transient reactance)
where the waveform is assumed sinusoidal..
3 F W
FORMULATION P
Althou E"is neither accessible nor directly controllable
and varies wit the load, its magnitude can be denved from the
generator's sub-transient phasor dia am for the nominal
With reference to fig. 1 and the per unit system[S], the operating int and then kept constant shown by dotted lines
following relationships exist between the variables involved in in fig. 2). %is is done to satisfy the generator's conventional
the static conversion process: load flow spcification (i.e. constant voltage) but at the fictitious
internal E bus. The conventional load flow is then carried out
on the assumption of erfect filtering at that point.
I, = k- 342 Id with salient-poLgenerators the conventionalformulation
x is often used by averaging the sub-transient reactances, i.e.
X = (&"+x ")/2. So far the error of such an approximation has
Ip = aI, (2) not been &cussed, due to the vast number of variables that
contribute to differences between steady state and dynamic
3a
V, = -aV,,cosa - -X,
3 Id (3)
related formulations.
In the case of unit-connected generation, ta ings are not
r x required on the transformer and hence the varialL a must be
V, I , = E I, cos0 mamtained at its nominal value.
(4)

v, I, = V-Ip cos# (5)

92 SM 593-4 PWRS A paper recommended and approved

by the IEEE Power System Engineering Committee of
the IEEE Power Engineering Society for presentation
at the IEEE/PES 1992 Summer Meeting, Seattle, WA,
July 12-16, 1992. Manuscript submitted August 26,
1991; made available for printing May 22, 1992. n
Filters
+ + Filters n
Fig. 1. Typical HVdc system.

0885-8950/93$03.00 0 1992 IEEE

702
E characteristic with the rectifier operatin under constant current
Fig. 2. Simplified phasor diagram of the unit-connected
generator.
Modified Aleorithm for VarvinP Commutatine VoltapeS.
As already indicated, the internal EMF behind sub-
transient reactance is not directly controllable. Instead, the
generator excitation will be controlled to provide the specified
dc link power at the specified firing angle (a-).
Therefore the commutatingvoltage (E")is not known in advance
and its magnitude and phase angle (8) must be derived as part
of the iterative solution.
The conventional vector equation (8) is replaced by:
x = [V,, I d , COSQ , @ , E , E N ,B I T
Therefore seven residual equations are needed to formulate the
load flow problem at the rectifier end.
The first two equations are common to the conventional
model (with the tap ratio variable removed). Therefore:
R(1) = V, - k, E N COS@ (10)
the characteristic. The second chart is for the V,/a
control (I,, = 1.0 kA) and in this case 1 8 points have been used.
Thii is a very demanding exercise that requires
approximately 30 minutes of CPU time usin a VAX 3500
com uter for each point of the characteristic. however it only
n e e 8 to be carried out once for a particular HVdc system.
The y. of interpolation used is a variation of the
method of imded differences[7] with the de ee of the
interpolation being constrained to a second order. %is method
was chosen because it requires less arithmetic operations, points
can be simply added or subtracted from the set used to construct
the polynomial and also previous computations can be reused.
The Equivalent Inverter Model
The absence of tap changers and the irrelevance of 9 at
the generator terminals of the unit connection scheme permits
a simpler formulation, in the form of a modified Equivalent
Inverter.
In this case the vector of variables is:
X = [V,, I d , a , cosa , @ , cosy]' (17)
which contains the five variables of the conventional (inverter)
set and an extra variable (cos a) representing the rectifier end
of the link. A schematic diagram of the Equivalent Inverter is
shown in fig. 3.
In line with conventional practice the Equivalent Inverter
will normally be on extinction angle control (f . ) and the
rectifier on current control, the s ecified current g e l derived
from a constant power setting (Pg) at the inverter end.
(9)
Besides Pdq and f, a tgird control specification must
be made to match the six-variable formulation. It is su ested
that fl, is used and the inverter transformer tap free? This
provides for the highest transmission voltage. Should the tap
changer attempt to violate one of its limits, this limit becomes
the new control s ecification while freeing the value of a.
The compLte set of residual equations becomes:

R(2) = V, 3 X, I,
- k, E" cosa - - (11)
r
where k, = 3J2/rr.
The next two are derived from the generator sub-transient
phasor diagram of fig. 2. That is:
R ( 3 ) = E - E N cos8 - ( x -XI')lIP I sin(8 + @ ) (12)
R(4) = E N s i n p - ( x - x " ) IZ, 1 cos(8 + @ ) (13)
To complete the set, three control specifications are required, a
typical selection being:
R(5) = I , - Id" (14) The incorporation of the equivalent inverter in a Fast
R(6) = E -E" (15) Decoupled ac-dc load flow al orithm[S] requires modifications
to the B , AA', AA" and B b sub-matrices of the Jacobian
R(7) = COSU - cosamin (16) matrix. These sub-matrices, shown in Appendix B, are greatly
simplified as they contain only a single element each.
LOAD FLOW MODEL BASED ON DYNAMIC
SIMULATION.
Accurate Den'vation of Rectifier Characteristics.
In the absence of local load at the rectifier end, the unit-
connected scheme can be considered as an equivalent HVdc
generator and the complete HVdc link as an equivalent inverter.
A time domain solution of the differential equations
representing the generator and rectifier behaviour provides
detailed information of the required voltage and current
waveforms. Once the steady state waveforms are obtained, their
averaged values provide accurate output characteristics for the
unit-connected group. For every combination of I4 and a the
output dc voltage Vd is calculated and the resulting information Vdr = f(l,,cosa) Filters
is stored in memory for use in the load flow solution.
Two charts have been stored in this case. The first is for
the v,& characteristic with the rectifier operating as a diode
bridge (a = 00). 15 points have been used to adequately model Fig. 3. The Equivalent Inverter.

E E"
x-x))
L1
Bus 1 Bus 2
Rectifier
Fig. 4.
ORMANCE OF THE AJUX;ORITHMS
Test System
tem consists of a single unit connected
enera2: $3 dc link with a 12 ulse converter and a 3
h s b a r receiving end ac system, as stown in fig. 4. All other
relevant information is given in Appendix A.
ConventiogplLoad Flow Results (Model 1)
In order to obtain the linear vd/z,, characteristic of a
conventional HVdc scheme, the generator excitation must be
controlled to kee the commutating volta e (E") constant at
some specified v&e to provide the re uirecfnominal dc voltage
at minunum delay angle (amin). The characteristic is shown
in fig. 5.
By way of an example, the sending end of the test system
can be operated as a diode rectifier ( a = @ )with an avera ed
d
sub-transient reactance and a constant E" of 66.66 kV). h e
results of varying the constant current speci cation from 0.5 to
0.9 kA are shown in Table 1. The table includes the calculated
tap position of the inverter end transformer needed to maintain
a constant minimum extinction angle of 7- = 18". The complex
powers at the three busbars of the ac system are also shown.
The results for a current setting of 0.9 kA cannot be
included because of a violation of the maximum commutation
angle of 30". This indicates the inapplicability of the conventional
formulation in all but a limited set of circumstances.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
CURRENT (kA)
Fig. 5. Dc voltage / dc current characteristin for 3 models of the unit-connected HVdc generator- converter.
703
Bus 4
Invertor
Bus 5
The test system.
Modified Steadv State (Mode121 and Eauivalent Inverter
[Model 31 Results
In order to highlight the limitations of the steady state
formulation two examples have been chosen, one with varying dc
current and minimum firing angle (a-) and the other with
variable a and a constant link current setting. In both cases the
generator excitation is ke t constant at the value necessary to
achieve the nominal dc vo&e of 80.0 kV when the no- link
current is 1.0 kA. Therefore in all cases the specified vmables
are E, Id and a. The identical excitations are confirmed in fig. 5
for zero dc current.
Tables 2 and 3 illustrate the major differences between
the conventional steady state formulation and the accurate
solution based on dynamic simulation characteristics. At the
lower end of the current setting range (i.e. below 0.7 kA), the
tables show that the arbitrary maximum transformer tap setting
of 20% is exceeded when using the conventional formulation,
whereas this is not the case in the e uivalent inverter results.
From 0.7 to 0.9 kA,differences of 30% occur in the calculated
values of the inverter tap position and 8% in the calculated
values of complex ower in the ac system. Further increases in
current setting r e s 9 in violations of the commutation an e limit
and thus invalidate any results obtained from the stea%l y state.
On the other hand the equivalent inverter shows that minimum
extinction is still maintahed at current settings in excess
of 1.0 kA an that chan ewer from tap to 7 control takes place
at a current less than 1 6 kA. Although the commutation angle
cannot be determined in the case of the equivalent inverter
model, the results from the dynamic simulauon are valid for
angles above and below 30".
1 1.1 1.2 1.3 1.4
704
Table 1 Conventional load-flow (Model 1). Diode rectifier Table 4 Results with Id = 1.0 kA and variable a . 7,=18",
bridge ( a =V).7- = 18". With a in %, P in MW and a
, = 20%.
Q in MVAr.

CONCLUSIONS
Bus 3 Bus 4 Bus 5
The conventional ac-dc steady state formulation has been
Id(W 7; I ai P, I Q3 p4 [ Q4 Ps I Q, modified to analyse HVdc load flows with unit connected
generator converter in-feeds.However the steady state algorithm
0.70 a; < -20% 1s restricted in its application to large wer conditions due to
the laqe commutation overlap caused the extra reactance of
0.80 7- -19.99 76.83 -28.57 -140. 16.50 63.17 15.57 the umt connection.
0.85 T,,,*, -17.71 78.91 -29.68 -140. 17.07 61.09 16.12 Also the results have been shown to be in considerable
error when the generators contain rotor saliency (probably the
0.90 7- -14.76 80.52 -30.67 -140. 17.58 59.48 16.61 most common case for remote generatin plant).
A new concept, called the Equivakent Inverter which uses
1.00 p, > 30" unit connection characteristics derived from time domain
simulation, has been presented as a practical alternative for
I 1.25 1
I I
p, > 30" I
I general load flow studies involving unit connected schemes.
ACKNOWLEDGEMENTS.
The authors wish to acknowledge the financial assistance
Table 3 Equivalent Inverter (Model 3). Diode rectifier bridge rovided b New Zealand Ministry of External Relations and
( a=OD). 7 . = 18", am = 20%. With 7 in degrees, frade, by 'fh e Universidade Federal de Uberlandia and b the
a in %, P E M W and Q in MVAr. Capes /Ministry of Education of Brazil. They also acknowzdge
the advice received from their collea es Dr. N.R. Watson, and
Messrs. A. Medina, M. Zavahir and %. Villablanca.
I d ( W REFERENCES.
[l] - P.C.S. Krishnayya, et "A Review of Unit Generator-
Converter Connections for HVdc T.ransmission",
LEEEKSEE Joint COnference on Hi& Vol t u
Transmission &stem , Beijing, China, 1987.-
0.90
[2] - KW. Kangiesser, "Unit Connection of Generator and
7,, -5.19 80.10 -31.13 -140. 17.99 59.90 17.04 Converter to be Integrated in HVdc or HVac Energy
Transmission", International SvmDosium on HVdc
1.25 23.06 a- 75.37 -38.62 -140. 21.82 64.63 20.89 Technology, Rio de Janeiro, Brazil, -1983.
[3] - M. Naidu and R.M. Mathur, "Evaluation of Unit
Connected, Variable Speed, Hydro wer Station for
HVdc Power Transmission",Trans. I E E , vol. PS-5, no. 2,
The second example is used to illustrate the effect of 1989, pp. 668-76.
firin angle variation on the inverter tap, extinction an e and
rectiker commutation angle; the results are given in &ble 4. [4] - J.P. Bowles, "Multiterminal HVdc Transmission Systems
The differences in the first row ( a . =V)have already been Inco orating Diode Rectifier Stations", Trans. IEEE,
discussed in the previous example. g r firing an es between 5 vol. &S-lOO, no. 4, 1981, pp. 1674-8.
and 20" both algorithms show a similar pattern fut differences
of 6.5% are observed in the values of commutation overlap. For [5] - J. Arrillaga, C.P. Arnold and BJ. Harker, ''Compuw
a=3V only Model 2 show a changeover from 7 to tap control. Modelline of Electrical Power Svstem",J. Wiley & Sons
For 45" and beyond, all models predict tap control, the actual Ltd., London, 1983.
values of 7 differing by up to 33%.
From the above results and discussion it is clear that only [6] - J. Arrillaga, S. Sankar, N.R. Watson and C.P. Arnold,
the Equivalent Inverter algorithm can be relied upon to provide "Operational Ca ability of Generator-HVdc Converter
realistic load flow information when the system contwns unit Umts", Trans. PEEE, Winter Meeting 1991, paper
connected in-feeds. 91WM 120-6 PWRD.
 705
[7]- F.C. Gerald q d P.O. Wheatley, " died N *
The elements of the dc Jacobian matrix [A]are:
AJ&&", Addson-Wesley Publishin$ko., Ne%
1985.

All parameters quoted are in per unit on a 100 MVA w=

base except where otherwise specified.
nt Pole Generator:
Rating: 100 MVA
T e r n Voltage: 13.8 kV
D-axisreactan&: 1.2
Q-axis reactance: 0.8
D-axissub-transient reactance: 0.2
Q-axis sub-transient reactance: 0.367
DEbdBk: where Vd, = f(id, cosa)
12 pulse
2ZL arrent:
Nominal Voltaee:
1.0
80.0 kV
and with rows ordered as in equations (18)-(23) and columnS
ordered as in equation (17).
Resistance: " 1.0 n
Converter Transformer:
Rating: 50 MVA
Reactance: 0.1
Voltage: 13.8/30.36 kV

L1: i0.9165
rn 40.04
U: 10.03
U: j0.03
..
mtial conditions:
E -
E" -
internal EMF,
commutating voltage,
Bus 1: Slack V = 1.548 pu X - generator contribution to the commutation
Bus 2 P,Q voltage,
Bus 3: Slack V = 1.0 pu
P,V V=1.0 PU P=-140.0MW
r - generator sub-transient reactance,
Bus 4: X - converter transformer reactance,
Bus 5: P,Q $- commutating reactance of the converter,
Dc current of the link,
4: Ac current of the primary of the converter
transformer,
!llmNmB- ivalent Inverter ac-dc Jacob= 1,- Ac current of the secondary of the converter
transformer,
a - tap of the converter transformer,
V Dc voltage of the converter,
fi- Ac voltage of the primary of the converter
transformer,
Er,i - Ac voltage of the secondary of the converter
transformer,
Q - firing angle of the converter,
7 - extinction angle of the converter,
U - commutation angle of the converter,
9 - power factor angle in the primary of converter
transformer,
9- power factor angle in the secondary of converter
transformer,
1- fectifier,
i- inverter.
706
Jos Arrillau received his BE degree of Spain and his
MSc, PhD and I& of Manchester, where he led the power
systems group of UMIST between 1970-74. He has been a
Professor at the Universi of Canterbury since 1975. He is a
Fellow of the IEE, IEE%, the New Zealand Institution of
Engineers F d the Ro al Society of New Zealand.
Chns Arnold oitained his Msc and PhD from UMIST in
1970 and 1976 and between these events worked with G.E. in
Schenectady, NY, USA Since 1976 he has worked as a Lecturer
and Senior Lecturer at the University of Canterbury,
Christchurch, New Zealand. Dr. Arnold is a Member of IEE, a
Senior Member of the IEEE, a Chartered Engineer of the U.K.
and a Re istered Engineer of N.Z.
If received his BE de ee from Universidade
Federafdd Z z d i a (UFUb) - BraziRn 1978, his MSc from
Universidade Federal de Santa Catarina (UFSC) - Brazil in
1987. Current1 he is on leave as a Lecturer at UFUb-Brazil
workin on {is PhD at the University of Canterbury,
Christcfurch, N.Z. Mr. Camacho is a Student Member of the
IEEE and a Registered Engineer of Brazil.
S. Sankar received his BE degree from Madras University
in 1982, his M.Tech. from the Indian Institute of Technology,
Bombay of 1984 and his PhD from the University of Canterbury,
N.Z. in 1991. Currently he is a Post-Doctoral Fellow at the
University of Canterbury, N.Z.