Professional Documents
Culture Documents
5.1 Introduction
The relatively recent availability of high voltage power electronic switching has led
to the development of HVDC transmission and FACTS technologies. Although the
original purpose of the former was a return to d.c. transmission for large distances,
most of the recently commissioned schemes are actually back-to-back; their purpose
is the interconnection of asynchronous power systems.
The FACTS technology, on the other hand, has developed to enhance the controllability of the synchronous transmission system, thus permitting its operation closer
to the stability limits. The basically different reasons for the existences of HVDC and
FACTS have important consequences in the load flow solutions.
This chapter describes the incorporation of these two power electronic technologies in conventional load flow solutions whenever possible, and describes alternative
methods of solution when such incorporation makes the load flow convergence difficult.
130
When applied to in-phase tap changing, the following substitutions must be made
in Equation (5.1):
T , = T: = T,LO",
and
u, = u; = UVLO".
For a tap changer connected between terminals k and m, to control the voltage
magnitude at bus k, the extra linearized load flow equations required at nodes k and
m are:
Finally, at each iteration, the tap controller is updated by the following equation
(5.3)
5.2.2
Phase-shifting (PS)
A transformer admittance matrix for phase-shift use has been described in section 3.2.3.
Its incorporation in Newton's load flow solution requires an enlarged Jacobian, because
the active power flowing from node k to node m, Pkm, is not used as a control variable
in the conventional formulation of the Newton-Raphson solution. Therefore, either d,,
or &, must be added to the state variables.
131
and
- pf;l
Apt, = APf;
A4 =
- c$i
Unless the initial values of the complex tap positions are known, the phase-shifter
can be initialized with zero degree angles.
(5.7)
and the expression for X I in terms of the firing angle has been derived in section 3.2.1.
v..
II
v-
132
= - v k V m B ( I ) sin(8k - Om),
Qk
= -viB(~)
-k V k V m B ( I ) c o s ( 8 k - o m ) .
(5.9)
(5.10)
The power mismatches required to incorporate a TCSC in the load flow solution are
contained in the following matrix equation (written for a specified power flow from k
Apk
APm
(5.11)
A Qk
AQm
-Apkm
(5.12)
(5.13)
(5.14)
The presence of series compensators can cause ill-conditioning of the Jacobian matrix
if zero voltage angle initialization is adopted. To avoid ill-conditioning, it is recommended to treat the TCSC as a fixed reactance until a specified voltage angle difference
appears across the reactance. From then on, the control equations of the series compensator are included in the iterative process.
5.2
converter I1
12 converter
++
t t
V ~ R0vR
Figure 5.2
t t
vcR
ecR
133
134
where
Ykk = Gkk
Y m m = Gmm
+ jBmm = ZFi,
+ jBvR = -Z;R.I
Ideally, the UPFC neither absorbs active power from nor injects it into the a.c. system
and the d.c. link voltage, Vdc, remains constant. However, to keep v d c constant, the
)
to the d.c. power
shunt converter must supply an amount of active power ( P V ~equal
( v d c l 2 ) , which is equal to the power supplied by the series converter ( P c ~ )i.e.
,
PvR
+ PcR = 0.
(5.23)
As stated in the Introduction, the state variables representing the UPFC must be
added to the network nodal voltage magnitudes and angles to permit a unified solution
by the Newton-Raphson method. In this way, the UPFC state variables are adjusted
automatically to satisfy specified power flows and voltage magnitudes.
When the UPFC controls the voltage magnitude at node k (to which the shunt
converter is connected), as well as the active power transfer from m to k and the
reactive power injection at node m (which is specified as a P, Q bus), the following
matrix equations applies:
(5.24)
where APbb represents the mismatch introduced by the presence of Equation (5.23).
If voltage control at node k is not required, the third column of the Jacobian in
Equation (5.24) is replaced by partial derivatives of the nodal and UPFC mismatch
powers with respect to the nodal voltage magnitude vk. Correspondingly, in the variables vector the shunt source voltage magnitude increment, A V V ~ / V , is, ~replaced
,
by
the nodal voltage magnitude increment at node k, AVk/Vk. In this case, V,R is kept
constant at a value within the specified limits.
135
If a limit violation takes place in one of the voltage magnitudes of the UPFC sources,
the voltage magnitude is fixed at that limit and the regulated variable is freed.
In line with the basic load flow solution, in the absence of controlled buses or
branches, 1 p.u. voltage magnitude for all P Q buses and 0 voltage angle for all buses,
provide a suitable starting condition. However, if controllable devices are included in
the analysis, good initial estimates can be obtained by assuming a loss free UPFC and
zero voltage angles in Equations (5.15) to (5.18).
From Equations (5.17) and (5.18) for specified P m , Qm:
(5.25)
(5.26)
where
if
C1 = Qmref
vO,
= v:.
Also, in a lossfree UPFC, the active power absorbed by the shunt converter ( P v ~ )
must be equal to the power ( P c ~delivered
)
by the series converter, i.e.
PvR
= -pcR
(5.27)
combining (5.27) with (5.19) and (5.21), after some reduction leads to:
(5.28)
When the shunt converter operates as a voltage controller, the voltage magnitude of
the shunt source is initialized at the required voltage level and then updated at each
iteration. If the shunt converter is not used as a voltage controller, the voltage magnitude
of the shunt source is kept at a fixed value within its limits during the iterative process.
[V,8, KIT,
where
V is a vector of the voltage magnitudes at all a.c. system busbars;
3 is a vector of the angles at all a.c. system busbars (except the reference bus
which is assigned 8 = 0);
i is a vector of d.c. variables.
Chapter 4 has described the use of 7 and as a.c. system variables and the selection
of d.c. variables X is discussed in section 5.3.1.
The development of a Newton-Raphson based algorithm requires the formulation
of n independent equations in terms of the n variables.
136
The equations which relate to the a.c. system variables are derived from the specified
a.c. system operating conditions. The only modification required to the usual real and
reactive power mismatches occurs for those equations which relate to the converter
terminal busbars. These equations become:
(5.29)
- Qterm(aC) - Qterm(dc) = 0
(5.30)
QZm
where
Pterm(ac)is the injected power at the terminal busbar as a function of the a.c.
system variables;
Pterm(dc) is the injected power at the terminal busbar as a function of the d.c.
system variables;
P;:,,,
is the usual a.c. system load at the busbar;
and similarly for Qterm(dC) and Qterm(aC).
The injected powers Q,rm(dc) and P,rm(dC) are functions of the converter a.c.
terminal busbar voltage and of the d.c. system variables, i.e.
Pterm(dC) = f ( v t e r m 9 3
(5.31)
(5.32)
Qterm(dc) = f ( v t e r m ,
The equations derived from the specified a.c. system conditions may, therefore, be
summarized as:
(5.33)
ABterm (
7 9
8, z)
AG(v,8)
ADterm (7,8, @
Wte,,X)
=0,
(5.35)
137
(ii)
EL II/
Figure 5.3 Basic d.c. converter (angles refer to a.c. system reference)
Figure 5.4 Single-phase equivalent circuit for basic converter (angles refer to d.c. reference)
138
I,, I ,
0
a
Vd
Id
These ten variables, nine associated with the converter, plus the a.c. terminal voltage
magnitude Vterm,form a possible choice of X for the formulation of Equations (5.31),
(5.32) and (5.34).
The minimum number of variables required to define the operation of the system is
the number of independent variables. Any other system variable or parameter (e.g. P d c
and Qdc) may be written in terms of these variables.
Two independent variables are sufficient to model a d.c. converter, operating under
balanced conditions, from a known terminal voltage source. However, the control
requirements of HVDC converters are such that a range of variables, or functions
of them (e.g. constant power), are the specified conditions. If the minimum number
of variables is used, then the control specifications must be translated into equations
in terms of these two variables. These equations will often contain complex nonlinearities, and present difficulties in their derivation and program implementation. In
addition, the expressions used for Pdc and Q d c in Equations (5.29) and (5.30) may be
rather complex and this will make the programming of a unified solution more difficult.
For these reasons, a non-minimal set of variables is recommend, i.e. all variables
which are influenced by control action are retained in the model. This is in contrast
to a.c. load flows where, due to the restricted nature of control specifications, the
minimum set is normally used.
The following set of variables permits simple relationships for all the normal control
strategies:
[a] = [ V d , I d , a, cos(Y,#IT.
Variable 4 is included to ensure a simple expression for Qdc. While this is important
in the formulation of the unified solution, variable 4 may be omitted with the sequential
solution as it is not involved in the formulation of any control specification; cosa is
used as a variable rather than (Y in order to linearize the equations and thus improve
convergence.
d.c. per unit system To avoid per unit to actual value translations and to enable the
use of comparable convergence tolerances for both a s . and d.c. system mismatches, a
per unit system is also used for the d.c. quantities.
Computational simplicity is achieved by using common power and voltage base
parameters on both sides of the converter, i.e. the a.c. and d.c. sides. Consequently,
in order to preserve consistency of power in per unit, the direct current base, obtained
from ( M V A B ) / V B , has to be 4 times larger than the alternating current base [ 5 ] .
This has the effect of changing the coefficients involved in the a.c.-d.c. current
relationships. For a perfectly smooth direct current and neglecting the commutation
overlap, the r.m.s. fundamental components of the phase current is related to Id by the
139
approximation,
(5.36)
Translating Equation (5.36)to per unit yields
ls(p.u,) = -&Id(p.u.),
n
and if commutation overlap is taken into account, as described in Chapter 3, this
equation becomes:
(5.37)
where k is very close to unity. In load flow studies, Equation (5.37) can be made
sufficiently accurate in most cases by letting:
k = 0.995.
Derivation of equations The following relationships are derived for the variables
defined in Figure 5.4. The equations are in per unit.
(i)
The fundamental current magnitude on the converter side is related to the direct
current by the equation
(5.38)
(ii)
I , = a I,.
(5.39)
(iii) The d.c. voltage may be expressed in terms of the a.c. source commutating
voltage referred to the transformer secondary, i.e.
(5.40)
The converter a.c. source commutating voltage is the busbar voltage on the
system side of the converter transformer, Vterm.
(iv)
The d.c. current and voltage are related by the d.c. system configuration,
f ( v d , Id)
= 0,
(5.41)
- I d & = 0.
(v) The assumptions listed at the beginning of this section prevent any real power
of harmonic frequencies at the primary and secondary busbars. Therefore, the
real power equation relates the d.c. power to the transformer secondary power
in terms of fundamental components only, i.e.
VdId
= Els
COS
II/.
(5.42)
140
(vi) As the transformer is lossless, the primary real power may also be equated to
the d.c. power, i.e.
(5.43)
VdId = VtermIp cos 4.
(vii) The fundamental component of current flow across the converter transformer can
be expressed as
I , = B, sin II/ - BtaVtem sin&
(5.44)
where j B , is the transformer leakage susceptance.
So far, a total of seven equations have been derived and no other independent
equation may be written relating the total set of nine converter variables.
Variables I , , I,, E and can be eliminated as they play no part in defining control
specifications.
Thus, Equations (5.38), (5.39), (5.42) and (5.43)can be combined into
Vd
- klaVterm COS 4 = 0,
(5.45)
where kl = k ( 3 a / ~ r ) .
The final two independent equations required are derived from the specified control
mode.
The d.c. model may thus be summarized as follows:
B(z,Vtermlk = 0,
(5.46)
where
- klavtem C O S a + -J3rI d x , ,
W d ?I d ) ,
R ( 5 ) = control equation,
and
= V t e d p sin 4
= Vte,kla I d sin4
(5.47)
and
(5.48)
or
(5.49)
141
(ii)
- 0.
Vd
- vsp= 0.
Id
- Isp = 0.
(v)
- P:: = 0.
These control equations are simple and are easily incorporated into the solution
algorithm. In addition to the usual control modes, non-standard modes, such as specified a.c. terminal voltage, may also be included as converter control equations (see
section 5.3.3).
During the iterative solution procedure, the uncontrolled converter variables may go
outside pre-specified limits. When this occurs, the offending variable is usually held
to its limit value and an appropriate control variable is freed.
Inverter operation All the equations presented so far are equally applicable to
inverter operation. However, during inversion, it is the extinction advance angle ( y )
which is the subject of control action and not the firing angle a. For convenience,
therefore, equation R(2) of (5.46) may be written as
(5.50)
This equation is valid for rectification or inversion. Under inversion, v d , as calculated
by Equation (5.50), will be negative.
To specify operation with constant extinction angle, the following equation is used:
cos(7r - y ) - cos(7r - y ) = 0
where ysP is usually y minimum for minimum reactive power consumption of the
inverter.
142
Unified solution [6,7] The unified method gives recognition to the interdependence
of a.c. and d.c. system equations and simultaneously solves the complete system.
Referring to Equation (5.33,the standard Newton-Raphson algorithm involves repeat
solutions of the matrix equation:
(5.51)
AQterm =
QZm
- Qterm(aC) - Qterm(dc),
(5.52)
(5.53)
and
Pterm(dc) = f(Vterm,
Qterm
(dc 1 = f ( Vterm
(5.54)
(5.55)
Applying the a.c. fast decoupled assumptions to all Jacobian elements related to the
a.c. system equations, yields:
(5.56)
where all matrix elements are zero unless otherwise indicated. The matrices [B] and
[B] are the usual single-phase fast decoupled Jacobians and are constant in value. The
other matrices indicated vary at each iteration in the solution process.
A modification is required for the element indicated as Byi in Equation (5.56). This
element is a function of the system variables and, therefore, varies at each iteration.
143
The use of an independent angle reference for the d.c. equations results in
aPterm(dc)/Wem = 0,
i.e. the diagonal Jacobian element for the real power mismatch at the converter terminal
busbar depends on the a.c. equations only and is, therefore, the usual fast decoupled
B element.
In addition,
aR/aOter, = 0,
which will help the subsequent decoupling of the equation.
In order to maintain the block successive iteration sequence of the usual fast decoupled a.c. load flow, it is necessary to decouple Equation (5.56). Therefore, the Jacobian
submatrices must be examined in more detail. The Jacobian submatrices are:
1
D D = -a APterml avterm
Vterm
- -(aptem(ac) /avteml+
1
( apterm (dc)/aVterm1Vterm
Vterm
and since
1 9
144
In the above formulation, the d.c. variables X are coupled to both the real and reactive
power ax. mismatches. However, Equation (5.56) may be separated to enable a block
successive iteration scheme to be used.
The d.c. mismatches and variables can be appended to the two fast decoupled a.c.
equations, in which case the following two equations result:
(5.57)
(5.58)
AZ
The iteration scheme illustrated in Figure 5.5 is referred to as -PDC, QDC- and the
significance of the mnemonic should be obvious.
The algorithm may be further simplified by recognizing the following physical characteristics of the a.c. and d.c. systems:
0
The coupling between d.c. variables and the ax. terminal voltage is strong.
Under all practical control strategies, the d.c. power is well constrained and this
implies that the changes in d.c. variables j z do not greatly affect the real power
mismatches at the terminals. This coupling, embodied in matrix AA' of Equation (5.57) can, therefore, be justifiably removed.
These features justify the removal of the d.c. equations from Equation (5.57) to
yield a -P, QDC-block successive iteration scheme, represented by the following two
equations:
[ABIV] = [B'J[AS],
(5.59)
Calculate Pterm
(dc) and d.c. residuals ( R )
Yes
b
lP=11
UDdate Hand 8
t
P=O
+=
1I
t
Figure 5.5
1 No
145
146
Initially, the a.c. fast decoupled equations are formed with the d.c. link ignored
(except for the minor addition of the filter reactance at the appropriate a.c. busbar).
The reactive power mismatch equation for the ax. system is:
(5.61)
where
AQ:,,, = Q:,
and
B is the usual constant a.c. fast decoupled Jacobian.
After triangulation down to, but excluding, the busbars to which d.c. converters are
attached, Equation (5.61) becomes
=m[
[
--
(5*62)
(5.63)
where
[ (W+
AQ;iy)]
[A:;]
(5.64)
BB
147
The most efficient technique for solving Equation (5.64) depends on the number
of converters. For six converters or more, the use of sparsity storage and solution
techniques is justified, otherwise all elements should be stored. The method suggested
here is a modified form of Gaussian elimination where all elements are stored but only
non-zero elements processed.
Sequential method [5,8] The sequential method results from a further simplification
of the unified method, i.e. the ax. system equations are solved with the d.c. system
modelled simply as a real and reactive power injection at the appropriate terminal
busbar. For a d.c. solution, the a.c. system is modelled simply as a constant voltage at
the converter ax. terminal busbar.
The following three equations are solved iteratively to convergence:
[ A F / v ] = [B][A8],
(5.65)
[AQi/v] = [B][Av],
(5.66)
[R] = [A][AT].
(5.67)
Calculate
5.3.3
(a) By local reactive power injection at the terminal. In this case, no reactive power
mismatch equation is necessary for that busbar and the relevant variable (i.e.
148
-.
i
Yes
*
) and d.c. residuals R
Converged
I Fl
t
I Solve eauation 5.67 and udate K
LTJ
P=Q=O
AV,,,) is effectively removed from the problem formulation. This is the situation
where the converter terminal busbar is a P-V busbar.
(b) The terminal voltage may be specified as a d.c. system constraint. That is, the d.c.
converter must absorb the correct amount of reactive power so that the terminal
voltage is maintained constant.
149
Vterm
- Vterm -0
(5.68)
is written as one of the two control equations. This would lead to a zero row in Equation
(5.67) and, therefore, during the solution of Equation (5.67) some other variable (e.g.
tap ratio) must be specified instead. Although d.c. convergence is marginally slower
for the PDC, QDC iteration, the d.c. system is overconverged in this iteration scheme
and the overall convergence rate is practically unaffected.
With the sequential method, Equation (5.68) cannot be written. The terminal busbar
is specified as a P-V busbar and the control equation
Q
m
r:;
is used, where Q&, (dc) is taken as the reactive power required to maintain the
voltage constant. The specified reactive power thus varies at each iteration and this
discontinuity slows the overall convergence.
5.3.4
The basic algorithm has been developed in previous sections for a single d.c. converter.
Each additional converter adds a further five d.c. variables and a corresponding set of
five equations. The number of a.c. system Jacobian elements which become modified
in the unified solutions is equal to the number of converters.
As an example, consider the system shown in Figure 5.7. The system represents the
North and South Islands of the New Zealand system, 220 kV a.c. system. Converters 1
and 2 form the original 600 MW, 500 kV d.c. link between the two islands. Converter 3
represents a 420 MW aluminium smelter. A further three-terminal d.c. interconnection
has been added (Converters 4,5 and 6 ) to illustrate the flexibility of the algorithm.
'd 6
'd 2
+
Figure 5.7
150
Normally, Converter 4 will operate in the rectifier mode with Converters 5 and 6 in
the inversion mode.
The reactive power d.c. Jacobian for the unified method has the structure, shown in
Figure 5.8, where BSOD is the part of B" which becomes modified. Only the diagonal
elements become modified by the presence of the converters.
Off diagonal elements will be present in BboD if there is any a.c. connection between
converter terminal busbars. All off diagonal elements of BB" and AA" are zero.
In addition, matrix A is block diagonal in 5 x 5 blocks with the exception of the
d.c. interconnection equations.
Equation R ( 3 ) of (5.46) in each set of d.c. equations is derived from the d.c. interconnection. For the six-converter system, shown in Figure 5.7, the following equations
are applicable.
vdi
v d 2 - Idl(Rd1
R d 2 ) = 0,
vd3
Id1
= 0,
- Id2 = 0,
vd4
+ v d 6 - l d 4 R d 4 - l d 6 R d 6 = 0,
vd5
- vd6
- I d s R d ~+ l d 6 R d 6 = 0 ,
I d 4 - I d s - I d 6 = 0.
This example indicates the ease of extension of the multiple converter case.
Avterrn 6
(30 x 30)
5.3.5
151
The d.c. p.u. system is based upon the same power base as the a.c. system and on
the nominal open circuit a.c. voltage at the converter transformer secondary. The p.u.
tolerances for d.c. powers, voltages and currents are, therefore, comparable with those
adopted in the a.c. system.
In general, the control equations are of the form
where X may be the tap or cosine of the firing angle, i.e. they are linear and are
thus solved in one d.c. iteration. The question of an appropriate tolerance for these
mismatches is, therefore, irrelevant.
An acceptable tolerance for the d.c. residuals which is compatible with the a.c.
system tolerance is typically 0.001 p.u. on a 100 MVA base, i.e. the same as that
normally adopted for the ax. system.
Although the number of d.c. iterations varies for the different sequences, this is of
secondary importance and may, if required, be assessed in each case from the number
of a.c. iterations. In this respect, a unified QDC iteration is equivalent to a Q iteration
and a DC iteration executed separately.
The d.c. link data and specified controls for Case 1 are given in Table 5.2 and the
corresponding d.c. link operation is illustrated in Figure 5.9. The specified conditions
for all cases are derived from the results of Case 1. Under those conditions, the a.c.
system in isolation (with each converter terminal modelled as an equivalent a.c. load)
requires (4,3) iterations. The d.c. system in isolation (operating from fixed terminal
voltages) requires two iterations under all control strategies.
Unified cases The results in Table 5.1 show that the unified methods provide fast
and reliable convergence in all cases.
For the unified methods 1 and 2, the number of iterations did not exceed the number
required for the a.c. system alone.
152
Specified d.c.
link constraints
m-rectifier and
n-inverter end
UrnPdrnYnVdn
UmPdnian Vdn
UmPdmanVdn
UmPdniYn vdri
UmPdrnYnan
amPdmUmYn
~ m I Yn
d Vdn
amVdmYnPdn
Case 1 with initial
condition errors
50 per cent error
80 per cent error
2
3
4
5
9
10
Unified methods
(5 variables)
Sequential methods
PDC.QDC2 P, QDC
(5 variables)
(4 variables)
P.Q,DC2P.DC,Q,DC
1P,Q,DC2P,DC.Q.DC
4,3
4,3
4,3
4,3
4,3
43
4,3
4,3
4,3
43
4.3
4,3
43
4,3
4,3
4,3
4,3
494
4,4
4,4
494
4.3
5s
5,s
44
43
493
4-4
4,3
43
4,4
4.3
5,4'
4,3
44
4,3
6,5"
1,6*
5,4"
44
4,4
44
4,4
4,4
44
43
44
494
43
4-4
4,4
44
43
4-3
~~
Converter 1
Converter 2
Bus 5
Bus 4
1 0 0 kV
1 0 0 kV
0.126
0.126
0.478
0.0728
0.0728
0.629
0.334a
5 8 . 6 MW
7
10
-1 2 8 . 8 7
kV
Bus 4
V = 1.061
+Id =454.2
I
0
1
-t P = 58.60
0 = 18.79
R = 0.334
dvd
= 129.022 Vd=-128.87$.
P = -58.31
u = 10.33
Q = 16.78
All angles are in degrees. D.C. voltages and current are in kV and Amp respectively.
D.C. resistance is in ohms. A.C. powers (P,Q)are in MW and MVARs.
a = 7.0
u = 17.32
y = 10.0
494
44
43
Failed
Failed
44
153
Sequential cases The sequential method ( P , Q, DC) produces fast and reliable
convergence although the reactive power convergence is slower than for the a.c. system
alone.
With the removal of the variable 4, &,(dc)
converges faster but the convergence
pattern is more oscillatory and an overall deterioration of a.c. voltage convergence
results.
With the second sequential method (P, DC,Q, DC),convergence is good in all cases
except 2 and 3, i.e. the cases where the transformer tap and d.c. voltage are specified
at the inverter end. However, this set of specifications is not likely to occur in practice.
Initial conditions for d.c. system Initial values for the d.c. variables E are assigned
from estimates for the d.c. power and d.c. voltage and assuming a power factor of 0.9
at the converter terminal busbar. The terminal busbar voltage is set at 1.0 p.u. unless
it is a voltage controlled busbar.
This procedure gives adequate initial conditions in all practical cases as good estimates of Pter,(dc) and V d are normally obtainable.
With starting values for d.c. real and reactive powers within 35096, which are
available in all practical situations, all algorithms converged rapidly and reliably (see
Case 9).
Effect of a.c. system strength In order to investigate the performance of the algorithms with a weak a.c. system, the test system described earlier is modified by the
addition of two a.c. lines, as shown in Figure 5.10.
The reactive power compensation of the filters was adjusted to give similar d.c.
operating conditions as previously.
The number of iterations to convergence for the most promising algorithms are
shown in Table 5.3 for the control specifications corresponding to Cases 1 to 4 in the
previous results.
The different nature of the sequential and unified algorithms is clearly demonstrated.
The effect of the type of converter control is also shown. For Case 11, both the d.c.
real power and the d.c. reactive powers are well constrained by the converter control
strategy. Convergence is rapid and reliable for all methods.
In all other cases, where the control angle at one or both converters is free, an
oscillatory relationship between converter a.c. terminal voltage and the reactive power
of the converter is possible. This leads to poor performance of the sequential algorithms.
To illustrate the nature of the iteration, the convergence pattern of the converter
reactive power demand and the ax. system terminal voltage of the rectifier is plotted
in Figure 5.1 1.
Bus 5
ix
Bus 4
rl
Filters
s.
Filters
154
Table 5.3
Case specification
rn -rectifier
n -inverter
Unified
P , QDC
X;
Sequential
= 0.4
Unified
P , QDC
P , Q , DC
(i)
(ii)
Sequential
P , Q , DC
(i)
(ii)
(i) Using the five variable formulation; (ii) using the four variable formulation.
Q term
(MVAR)
$el?
28 26,24
- 0.96 1\
22
nnl
18 - o-.-0
20
Q term
(MVAR)
- 1.o
26 - 0.98
28
24
- 0.96
- 0.94
20 18 - 0.90
0.92 0
22
1
1
Figure 5.11 Convergence pattern for a.c.-d.c. load flow with weak ax. system. (a) Sequential
methods ( P , Q, DC five variable); (b) Unified method ( P , QDC)
A measure of the strength of a system in a load flow sense is the short circuit to
converter power ratio (SCR)calculated with all machine reactances set to zero. This
short circuit ratio is invariably much higher than the usual value.
In practice, converter operation has been considered down to an SCR of 3. A survey
of existing schemes shows that, almost invariably, with systems of very low SCR,some
155
form of voltage control, often synchronous condensers, is an integral part of the converter
installation. These schemes are, therefore, often very strong in a load flow sense.
It may, therefore, be concluded that the sequential integration should converge in all
practical situations although the convergence may become slow if the system is weak
in a load flow sense.
(i)
In cases where the a.c. system is strong, both the unified and sequential algorithms
may be programmed to give fast and reliable convergence.
(ii)
5.3.7
Numerical example
The complete New Zealand primary transmission system was used as a basis for a
planning study which included an extra multiterminal HVDC scheme, i.e. involving
six converter stations, as illustrated in Figure 5.7.
156
Representative input and output information obtained from the computer is given in
the following pages.
SYSTEM NO.
23 YAR 90
NUIBEL OF BUSES
NUIBEL OF LINES
NUIBER OF TUNSFUPlERS
B U S
BUS
104
108
118
127
128
NAYE
TYPE VOLTS
AVIEIOIE-220
BDINOILE-220
BUY-220
CWIl-220
CUJMZ-220
129 CLUTAA-220
1 1.0300
138 CMLDINE220 0 1.0210
0 1.0270
143 HyBS---220
D A T A
LOAD
IV
IVB
0.00
1 1.0520
1 1.0520 97.20
0 1.0030 329.60
0 1.0520
0.00
0 1.0520
0.00
0.00
0.00
95.30
LINE
114
206
19
0.OOO
0.000
0.00 600.00
0.00
0.00
80.40
0.00
0.000
0.000
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
D A T A
BUS
NAME
BUS
NAIE
104
104
104
108
118
AVIEIOIE-220
AVIEIOIE-220
AVIEMOIE-220
BENOS220
BUY-220
108
108
268
255
167
BENYORE-220
BENIOLE-220
VAITAKI-220
TVIZEL-220
ISLINGTON220
0.00330
0.00330
0.00150
0.00370
0.00210
0.01630
0.01530
0.00730
0.02610
0.01651
0.02298
0.02298
0.01052
0.06954
0.05285
118
127
127
128
128
BUY-220
CUJMl-220
CUJI1-220
cu)I2-220
CUlU2-220
181
218
255
218
255
LANILM2-220
1OXBuICH-220
TVIZEd-220
ROXBUILCH-220
TVIZEG-220
0.00110
0.00770
0.00820
0.00770
0.00820
0.00861
0.04450
0.09260
0.04450
0.09260
0.02751
0.07251
0.16746
0.07251
0.16746
LESISTANQ
LEACTANCE SUSCEPTANCE
O.Oo0
0.000
0.000
0.OOO
D A T A
BUS
NAME
BUS
NAME
LESISTANCE
6
6
6
10
10
21
BWTHOlPEl10
7
7
7
II
II
22
22
22
22
24
40
40
49
55
55
55
59
63
67
BWTHWE220
BIRITHOlPE220
BWTHOlPM20
EDGUXnBEZZO
EDCECOMBE220
0.00400
0.00400
0.00170
0.00400
0.00400
0.00170
0.00170
0.00410
0.004 I0
0.00090
BWrmOlF'ElIO
BIRITHOlPEl10
EDCECOIBEIlO
EDCECOMBEllO
HAWAIIIS-I10
21 HAYWAIDS-110
2 1 HAWALDS-I10
?I
23
39
HIWhM5-110
HFNDEISON110
IAlSDEN-IIO
MAUDEN-IIO
HEVPLYI7HlIO
OTAHIJW-I10
OTAHIJW-I10
MAHIJW-I10
39
48
54
54
54
58 PMIOSE-IIO
62 STLATSOLDIIO
66 TUUTENGAIIO
HAWANIS-220
HAWANIS-220
HAWANIS-220
HAWAMS-220
HENDEISONZ20
IAlSDW-220
IAISDM-220
NEWLTITY220
OTAWUlN-220
OTAHWU-220
OTARllWU-220
PEIROsc220
STUtfUlD220
TAIUTI)IGAZ20
157
1.OW
1.OW
0.09560
0.09560
0.04SW
0.09560
0.09560
0.05140
0.05140
0.10120
0.10120
0.01840
0.05500
0.05500
0.02480
0.04100
0.04100
0.04550
0.02750
0.05260
0.02530
o.oo0oo
0.00000
0.00080
0.00700
0 00700
0.00160
0.000w
0.00200
0.00080
I
1.ooO
1.OOo
1.ooO
1.040
1.000
0
0
0
0
1.000
1.OOo
1.053
1.OOO
0
0
1.OW
1.OW
1.000
l.OW
1.000
1.OW
1.ooO
1.OOO
0
0
0
0
0
0
0
WUhTIUNS
I r ~ ~ ~ + Y M + V D 5 r V W + Y D 7 * V D ~ W 9 + Y DIIII-ID2
l ~ l D l .III2-1D3 .III3-ID4 .IDI-IDS .IIIS-IW .LD6-1D7 .III7-ID8 .IIIB-IW .NI9-IDlO .u)lO=O
I I 0 0 0 0 0 0 0 0 25.5800 o.oo00 0 . m o.oo00 o.oo00 o.oo00 o.oo00 o.oo00 o.Ooo0 o.oo00
0 0 0
0.0000 o.Ooo0 0.0019 o.oo00 o.oo00 o.oo00 o.oo00 o.oo00 o.oo00 o.oo00
# 1 0 0 0 0
o.oo00 o.oo00 o.ooO0 1o.oooo 0.M)Oo 2o.oooo o.oo00 o.oo00 o.oo00 o.oo00
8 @ 0 1 0 1 0 0 0 0
0.0000 0.OOOO 0.0000 O.oO00 3.oo00-20.oooO 0.0000 O.oo00 O.oo00 O.Oo0o
0
0 0 0
8 @ 0 0 I -1
~,~~kID4+IDS+IDb+ID7+ID8~ID9+IDlO:O
0 0 0 0 0 0 0 0
1-1
~)~~IDkiD4+ID5rIW+I7+lD8+ID9+IDlO~O
) ( O
1 - 1 4
0 0 0 0
pC CONVEITEI WlBEl
INPUT DATA
NOMINAL DC VOLTAGE
l10.00oOo
I A X I I U I DC VOLTAGE
MINIIUI DC VOLTAGE
IAIIIUI DC CUuEm
150.00000
0.00000
O.oo00
0.08970
0.08970
10.00000
110.00000
0.00000
0.00000
0.00000
I .m
4
ComElTot P O ~ I FACIOI
Dc LINT VOLTAGE (KV)
CONVEllUl C O W L ANGLE
NNSFOUEI TAP
CowVEIlEl Dc POVEI ( I V )
T E U I N A L IEACIIVE P
O
W
10.00000
500.00000
(IVAI)
AC TEUINAL VOLTACE ( K V )
W N V U I U DC CUuEIlf (KA)
PC CONVEXTU WMBEl
co1(yElTEl
lNPWf
ATTACHED To BUS N U I B U
NOIINAL DC VOLTAGE
IAIINI DC VOLTAGE
MINIllll DC VOLTAGE
I A I I N I DC CUUENI
90 * 00000
140.00000
0.00000
0.0000
(DED)
T U N S M U U T Y : I I N I I U I (P.C.)
IAXIIUI (P.C.)
INCWEW
F I L T U IEACTANCE ( P . U . )
WIBU OF WIDGES IN SUlEs
0.07000
0,07000
8.o0000
150.00000
0.00000
0.00000
o.Ooo00
0.70000
2
-220.00000
8.00000
158
WAD
IV
CE)IEUTION
IVAL
4496.80
AC W S S S
IV
IVN
IVAl
1518.60
5226.58
ITERATIONS
194.91 -306.06
791.80
IISIATCH
#VAL
IV
SWVMS
IVM
534.87 -182.89
37.85
7 (BUNTHOlPE220)
IV
TUNSMUE1 TAP
(PEL CENT)
CONTML ANGLE
(DW
COIImATION ANGLE
-8.26
8.00
22.45
87.94
W
LIM TUIINAL POVEL =
-309.23 IV
no1 TUNSFDUEL M COmFITEI = -309.23 IV
REACTIVE P D V U OF FILTElS =
DC C W T
(K-ANPS)
(DGS)
DC VOLTAGE
(I(-VOLTS)
I .406
-220.00
L TUNSFEQ
207.61 IVAL
125.89 IVAt
142.86 IVAL
BUS DATA
BUS
NAlE
104 AVIEIOU-220
VOLTS
ANGLE
CENEMTION
IV
IVAl
1.052
4.78
220.00 -33.87
MAD
IV
IVAl
SHUNT
IVAL
0.00
0.00
0.00
BUS
108
NAIE
4.43
540.00 -88.04
97.20
0.00
IVN
~ ~ ~ 1 0 ~ 41.89
~ 2 2 -10.17
0
108 BENIOLC220
268 VAITAKI-220
IISIATCH
108 BMIOLC-220 1.052
41.89
136.22
o.oO0
-10.17
-13.53
o.Oo0
0.00
104 AVIEIDlE-220
41.83
7.88
104 AVIEIOU-220 -41.83
7.88
255 W 1 2 E 6 2 2 0
26.47
6.87
IISIAtCW
5OO.oO0 -110.672
118
BUY-220
0.968 -12.95
0.00
0.00 329.60
95.80
0.00
167 ISLINCMN220 -120.18
181 LAhDf02-220 -209.41
IISIATCU
5.4
4.019
-76.16
-19.64
4.002
References
1. Fuertes-Esquivel, C R and Acha, E, (1996). Newton-Raphson algorithm for the reliable solution of large power networks with embedded FACTS devices, Proceedings of the IEE on
Generation, Transmission and Distribution, 143 (3,pp. 447-453.
2. Fuertes-Esquivel, C R, Acha, E and Ambriz-Perez, H, (1998). A thyristor controlled series
compensator model for the power flow solution of practical power networks, IEEE Trunsuctions Winter Power Meeting, Paper 98WM 162.
3. Gyugyi, L, (1992). A unified power flow control concept for flexible a.c. transmission
systems, Proceedings of the IEE, 139, (4), pp. 323-333.
4. Fuertes-Esquivel, C R and Acha, E, (1998). The unified power flow controller: A critical
comparison of Newton-Raphson UPFC algorithms in power flow studies, Proceedings of the
IEE on Generation, Transmission and Distribution, 144 (9,pp. 437-444.
5. Sato, H and Amllaga, J, (1969). Improved load-flow techniques for integrated a.c.-d.c.
systems, Proc. IEE, 116. (4), pp. 525-532.
5.4 REFERENCES
159
6. Braunayal, D A, Kraft, L A and Whysong, J L, (1976). Inclusion of the converter and transmission equations directly in a Newton power flow, IEEE Transactions on Power Apparatus
and Systems, PAS-75, (1). pp. 76-88.
7. Arrillaga, J, Harker, B J and Turner, K S, (1980). Clarifying an ambiguity in recent a.c.-d.c.
load-flow formulations, Proceedings of the IEE, 127, Pt.C(5), pp. 324-325.
8. Reeve, J, Fahmy, G and Stott, B, (1976). Versatile load-flow method for multi-terminal
HVDC systems, IEEE PES Summer Meeting, Paper F76-354- 1. Portland.