Beruflich Dokumente
Kultur Dokumente
, J. Verboomen
, K. Purchala
, R. Belmans
, W. L. Kling
1
V
1
V
s
= E
s
(Y
ss
E
s
+Y
sr
E
r
)
(3)
Separating the real and imaginary parts and incorporating the
inuence of every bus in the system, gives the power equations
(using the -equivalent representation for transmission lines):
P
s
=V
s
r=1
[G
sr
V
r
cos
sr
+B
sr
V
r
sin
sr
] (4a)
Q
s
=V
s
r=1
[G
sr
V
r
sin
sr
B
sr
V
r
cos
sr
] (4b)
for every node s. Using (4), in the form of (5), the Newton-
Raphson algorithm is obtained (at the i-th iteration):
P
Q
(i)
=
P
V
V
Q
Q
V
V
(i)
V
V
(i)
(5)
This equation has to be solved iteratively.
The Newton-Raphson method has a quadratic convergence and
computing time only increases linearly with system size.
3.2 Assumptions of DC power ow
In order to reduce calculation time, the power ow problem can
be simplied in by making the system linear. A number of as-
sumptions are made in order to make his linearisation feasible:
1. Voltage angle differences are small, i.e. sin(
sr
)
sr
and
cos(
sr
) 1
2. Flat voltage prole : all voltages are put to 1 p.u.
3. Line resistance is negligible i.e. R X, thus lossless lines
4. Tap settings are ignored (and therefore also PSTs)
However, these assumptions are not always realistic. The volt-
age prole most likely is not at but may vary among busses,
causing the voltage prole to be different from the assumed
one. Moreover, the X/R ratio condition can be difcult to guar-
antee. The inuence of the resistance increases with the de-
crease of voltage, which means that only the high voltage trans-
mission networks can tolerate this condition. Each of these
assumptions has some inuence on the accuracy of the power
ow calculations. Therefore, the DC power ow is less accu-
rate compared to the full, AC power ow solution.
The forth assumption makes it impossible to include the in-
uence of PSTs. In section 4.1 this problem is overcome by
making use of power injections.
3.3 DC power ow formulation
When losses are neglected, nodal voltage magnitudes set to
1 p.u. and the sine of the angle is replaced by its argument,
equation (1) can be rewritten as:
P
Line
=
sr
X
Line
= B
line
sr
(6)
The entire system can be modelled solely using linear equa-
tions:
P
s
=
n
r=1
B
sr
(
s
r
) (7)
When using (7) in equation (5), clearly, the following holds:
P =
[] (8)
P is the vector of bus real power injections, B
1
[P] (9)
4
To check whether the both line parameters are
interdependent, their influence has been plotted against each
other in Fig. 6.
P error [%]
0
5
10
15
20
0 2 4 6 8 10 12 X/R
X=0..2 X=4..6 X=8..10 X=12..14 X=16..18
70 kV 150 kV 220 kV 380 kV
X/R ratios typical for
Figure 5. Influence of X/R ratio on active flow estimation error P
error
for a given
range of reactance X (5000 samples)
Figure 6. Influence line parameters on active flow estimation error P
error
(5000
samples)
The following conclusions can be drawn:
For low values of the resistance the size of the
reactance, and consequently X/R ratio, is almost of no
importance, as there is virtually no change of P
error
with
varying reactance.
The higher the resistance, the more important X/R
becomes. Even for the highest values of the resistance,
provided that reactance is high enough, DC load flow
does not introduce unacceptable active power
estimation error.
As a general conclusion it can be said that it is the line
resistance which is the decisive factor for the feasibility of DC
power flow approximation. The smaller the line resistance,
the better the DC approximation of power flows. Additionally,
X/R ratio greater than 4 should be enough to limit the P
error
increase in case of higher values of the line resistance.
C. Influence of voltage variations (bad voltages)
One of the assumptions of DC power flow is the flat
voltage profile meaning that, in per unit terms, all voltages are
equal. It is often asserted that these should be as close to 1
p.u. as possible. However, it is not the absolute voltage
magnitude that matters, but deviations from the predefined
value. If the nodal voltages are scattered round 1.1. p.u.
instead of 1 p.u. DC power flow gives actually a better
approximation of the power flows, as higher line voltages
decrease losses. Voltage deviations, on the other hand, lead to
line voltage differences that cannot be accounted for in DC
power flow, which in turn influences the active power
estimation error. From Fig. 7 it can be seen that the P
error
increases with the increase of voltage deviations measured by
means of standard deviation s
U
(6).
=
n
i
i U
U U
n
s
1
2
) (
1
1
(6)
Though for most cases the average P
error
is limited to 5%,
the maximal error MAX P
error
, almost perfectly correlated to
the average, is over 8 times higher. Therefore the flat voltage
profile is of extreme importance for the accuracy of DC
power flow.
S
U
Figure 7. Influence of voltage fluctuations (standard deviation of the voltage) on
active flow estimation error (1000 samples)
In the actual power system however it is quasi impossible
to keep all voltages constant, avoiding voltage fluctuations.
To check the likelihood of a favorable voltage profile,
voltages in the Belgian high voltage network are taken as an
example. have been analyzed. Fig. 8 presents the nodal
voltage magnitudes in the Belgian high voltage transmission
grid, thus 70-380 kV, for the 13 GW winter peak scenario. As
noticed, voltage profile is not very flat, the standard deviation
being s
U
= 0.0166. As seen in Fig. 7, P
error
is very sensitive to
voltage deviations, and care should therefore be taken while
interpreting the results. Realistic example of voltages in the
actual power system shows that the assumption of perfect
voltage profile is the most critical one and voltage profile is
the biggest source of active power estimation error.
Figure 2: Inuence of voltage uctuations on the active power
ow estimation error (1000 samples on a randomly generated
30-bus test system) [7]
The DC power ow has three advantages over the standard
Newton-Raphson power ow:
1. the system matrix (B
1
has to be
calculated only once. This signicantly reduces calculation
time as triangularisation is computational heavy.
This method is called DC power ow because the linearised
model (6) can be interpreted as the model for a network of re-
sistors fed by DC currents sources, with P the vector of nodal
current intjections, the nodal vector of voltages and B
the
conductance matrix.
3.4 Accuracy of DC power ow
In [7], indexes have been identied that quantify the assump-
tions underlying the method. There are several parameters in-
uencing the accuracy of DC power ow. First of all, the volt-
age prole has to be as at as possible, meaning that there
should be as little voltage deviations as possible. The higher
they become, the higher the active power estimation error. A
standard deviation, S
V
, below 0, 015 is desirable. Secondly, the
X/R ratio should be high enough, otherwise the assumption of
negligible resistance is violated. The proposed border value
is set at X/R = 4. It should be noted that the inuence of both
parameters enforce each other as is clear from Fig. 2. The inu-
ence of linearizing the sine function is small when the voltage
angle remains below 30
r=1
[B
sr
(
s
r
)] +B
sr
PST
(11)
=
n
r=1
[B
sr
(
s
r
)] +P
shi f t
(12)
Note that in this case in fact sin(
sr
+
PST
) is replaced by
sr
+
PST
. This neglection is valid for small angles, but at an
angle of 30
.
5.2 Simulations
In order to assess the inuence of power ow controlling de-
vices on the accuracy of DC power ow, rst the correct so-
lution using AC power ow is found. The mismatch between
the obtained AC power ow and DC power ow with PSTs is
of a partially due to the incorrect phase shifter model and par-
tially to the inaccuracy of the overall DC power ow method.
By adding the difference between AC and DC solution at zero
degree phase shift to the DC solution (13), the corrected DC
voltage is found and the inuence of the phase angle on the
power ow error P
error
is seen.
P
DCcor
() = P
DC
() +(P
AC
(0
) P
DC
(0
)) (13)
P
error
() = P
AC
() P
DCcor
() (14)
Fig. 3 shows the AC, DC and corrected DC power ows
through the PST in the three considered cases (top), and the
error due to inaccurate power ow controller modelling (bot-
tom). As expected, the mismatch becomes larger when the
phase shift increases. At a phase shifting angle of 30
, the
error due to power ow controlling devices is around 5 % of
the line ow, but higher, up to 20 %, error rates were encoun-
tered during simulations. Furthermore, this error is added to
the error caused by DC power ow approximation of the sys-
tem. Fig. 4 shows the absolute value of the mismatch between
the corrected DC power ow calculation and the AC calcula-
tion using Newton-Raphson. The power mismatch caused by
the inaccurate representation of power ow controlling devices
1
The IEEE 300 system is online available http://www.ee.
washington.edu/research/pstca/pf300/pg tca300bus.htm
and within the MATPOWER program.
2
For the three cases considered in this paper, the PST is placed in the line
between nodes 4 and 16 (case 1), 129 and 130 (case 2) and between 133 and
137 (case 3)
30 20 10 0 10 20 30
800
600
400
200
0
200
400
600
800
P
h
a
s
e
s
h
i
f
t
e
r
f
l
o
w
P
f
l
o
w
(
M
W
)
30 20 10 0 10 20 30
60
40
20
0
20
40
60
E
r
r
o
r
(
M
W
)
Phase shifting angle (degree)
1
1
2
2
3
3
!
!
S
S
S
Z
Z
A
A
A
A
B
B
B
Figure 3: Top: Power ow through the PST calculated using
AC () and DC power ow (+); Bottom: Mismatch between
them: total (+) and corrected (). In both plots the three cases
are depicted: case 1 in blue, case 2 in red and case 3 in black.
30
15
0
15
30
0
50
100
150
200
250
300
350
0
5
10
15
20
25
30
35
Phase angle (
)
Branch number
Error (MW)
Figure 4: Absolute value of P
error
when placing a PST between
nodes 133 and 137 (case 3).
(PSTs) throughout the network is relatively small, except for
some lines that experience a power mismatch of up to 35 MW.
6 Conclusions
DC power ow is a much used tool in power system analy-
sis, especially in techno-economic studies, related to electric-
ity markets, and contingency analysis. Earlier work quantied
the main indexes of the power system for an accurate solution
when utilising DC power ow: especially a at voltage pro-
le and a high X/R ratio improve accuracy. In this paper, the
authors fundamentally extend this earlier research and exam-
ined the usefulness of this tool when the power system con-
tains power ow controlling devices, and especially PSTs. The
issues with modelling of power ow controlling devices for DC
power ow are examined. This modelling introduces an addi-
tional error caused by the approximation of a sine function by
its argument.
The DC power ow methodology is tested on a modied ver-
sion of the IEEE 300 bus system, to which a power ow con-
trolling device is added. Overall, the introduced error by in-
accurate representation of the power ow controlling device is
relatively small, increasing with higher phase shifting angles
(about 5 % of the line ow). However, this is an additional er-
ror and precaution has to be taken when drawing conclusions
based on simulations using DC power ow when PSTs are in-
volved.
Acknowledgments
The research performed at the K.U.Leuven is nancially sup-
ported by the Belgian Fund for scientic research (F.W.O.)
Vlaanderen. Dirk Van Hertem is a doctoral research assistant
of the F.W.O. The research at the TU Delft has been performed
within the framework of the intelligent power systems pro-
gram, nancially supported by SenterNovem. SenterNovem is
an agency of the Dutch ministry of Economic Affairs.
References
[1] E. Acha, C. R. Fuerte-Esquivel, Ambriz-P erez, and C. Angeles-
Camacho, FACTS Modelling and Simulation in Power Net-
works. Hoboken: John Wiley & Sons, 2004.
[2] S. Cole, D. Van Hertem, L. Meeus, and R. Belmans,
Technical developments for the future transmission grid,
in FPS, Amsterdam, the Netherlands, Nov. 1618, 2005,
p. 6. [Online]. Available: http://www.esat.kuleuven.be/electa/
publications/fulltexts/pub 1503.pdf
[3] J. J. Grainger and J. Stevenson, William D., Power System Anal-
ysis. McGraw-Hill, 1994, no. ISBN 0-07-061293-5.
[4] N. G. Hingorani and L. Gyugyi, Understanding FACTS, Con-
cepts and technology of exible AC transmission systems.
IEEE Press, 2000, no. ISBN 0-7803-3455-8.
[5] U. G. Knight, Systems in Emergencies, 1st ed. John Wiley &
Sons, Feb. 15, 2001.
[6] L. Meeus, K. Purchala, D. Van Hertem, and R. Belmans, Reg-
ulated cross-border transmission investments in europe, ETEP
European transactions on electric power, 2006, accepted for
publication.
[7] K. Purchala, L. Meeus, D. Van Dommelen, and R. Belmans,
Usefulness of dc power ow for active power ow analysis,
in IEEE PES general meeting, San Francisco, California, USA,
June 1216, 2005, p. 6. [Online]. Available: http://www.esat.
kuleuven.be/electa/publications/fulltexts/pub 1456.pdf
[8] B. Stott, Review of load-ow calculation methods, Proceed-
ings of the IEEE, vol. 2, no. 7, pp. 916 929, July 1974.
[9] D. Van Hertem, J. Verboomen, R. Belmans, and W. Kling,
Power ow control devices: An overview of their working
principles and their application range, in FPS, Amsterdam,
the Netherlands, Nov. 1618, 2005, p. 6. [Online]. Avail-
able: http://www.esat.kuleuven.be/electa/publications/fulltexts/
pub 1501.pdf
[10] R. D. Zimmerman, C. E. Murillo-S anchez, and
D. Gan, MATPOWER: Matlab Power System Simulation Pack-
age Users Manual, v3.0.0 ed., PSerc, Cornell Univer-
sity, Ithaca, NY, Feb. 15, 2005. [Online]. Available:
http://www.pserc.cornell.edu/matpower/