Sunteți pe pagina 1din 11

www.ietdl.

org

Published in IET Generation, Transmission & Distribution


Received on 19th April 2007
Revised on 23rd July 2007
doi: 10.1049/iet-gtd:20070165

ISSN 1751-8687

Modelling distributed generations in


three-phase distribution load flow
J.-H. Teng
Department of Electrical Engineering, I-Shou University, No. 1, Sec. 1, Syuecheng Road, Dashu Township, Kaohsiung County
840, Taiwan, Republic of China
E-mail: jhteng@isu.edu.tw

Abstract: The mathematical models of distributed generations (DGs) are integrated into three-phase distribution
load flow program to analyse and simulate the penetrations of DGs for distribution systems. DGs can utilise the
traditional energy sources such as oil and coal or renewable energy sources such as wind, solar and fuel cell and
use rotating generators or converters to transfer energy to power grids. According to the characteristics of
output power, DGs can be specified as constant power factor model, constant voltage model or variable
reactive power model in the load flow analysis. These three models are all derived and integrated into the
proposed load flow method. Test results demonstrate the validity and capability of the proposed method.

1 Introduction customer satisfaction, an efficient and robust three-


phase load flow method taking the mathematical
Distributed generations (DGs) have become an models of DGs into account should be developed first.
important part of electrical generation in many Some contributions have been developed to solve
countries and its importance is continuing to increase. parts of the problems [2– 7]. References [4, 7]
Especially, some believe that the most significant proposed some useful DG models for distribution load
changes for deregulated power systems will occur in flow.
the distribution business since the cost to serve
commercial and residential customers is higher than Conventionally, electrical power in distribution
the cost to serve industrial customers. Therefore the feeders always flow from substations to the end of
applications of DGs, in the small, self-contained, feeders in planning and operation. However, the
electric-generating plant that can provide power to a utilisation of DGs might cause reverse power flow and
single home, business or industrial facility, will be complicated voltage profiles in the feeders. Therefore
important and helpful. Moreover, if the planning and the existent power flow methods might need to be
operation of DGs can be integrated into distribution modified. Some works have been proposed to fulfill
automation (DA) and demand-side management some of those requirements [8 – 10]. Reference [8]
(DSM), a more comprehensive distribution system used the Gauss implicit Z-matrix method to solve the
with higher reliability, efficiency and power quality three-phase load flow problem. Transformer and co-
could be obtained [1 – 3]. The problems involving the generator models were also developed for rigorous
integration of DA, DGs and DSM are not easy to deal system analysis. The Gauss method is traditionally
with. Besides, when DGs were installed in distribution used for the load flow solution of general meshed
feeders and participated in system operations, the networks; however, distribution systems typically have
problems including power flow, power quality, a radial or weakly meshed structure. Therefore
ferroresonance, voltage control, loss reduction, research reveals some new ideas to deal with the
protection device coordination, voltage flicker and so special network characteristics of distribution feeders
on, needed to be carefully re-analysed. Since voltage [9, 10]. Reference [9] proposed a compensation-based
profiles are directly related to power quality and technique, where the forward/backward sweep

330 /IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340 & The Institution of Engineering and Technology 2008
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

algorithm was adopted in the solution scheme. A Procedure 3. Repeat procedure 2 until all line sections are
sensitivity matrix, used to calculate the incremental included in the BIBC matrix.
relation between the voltage magnitude and current
injections, was derived and used for voltage-specified The relation between branch currents and bus
generator bus. A direct-approach technique for solving voltages can be written as
the three-phase distribution power flow which fully
exploits network characteristic of distribution feeder [V 0 ]  [V] ¼ [BCBV][B] (2)
was proposed in [10]. Two developed matrices, the
bus-injection to branch-current matrix (BIBC) and the where [V] and [V0] are the vectors of bus voltages and
branch-current to bus-voltage (BCBV) matrix, and no-load bus voltages, respectively.
matrix multiplications are utilised to obtain power
flow solution. However, the integration of different Equation (2) can also be rewritten as
types of DGs was not proposed in [10].
[DV] ¼ [BCBV][B] (3)
This paper tries to integrate the mathematical models
of DGs into the three-phase load flow method proposed where [DV] ¼ [V 0 ]  [V].
in [10]. According to the characteristics of output
power, DGs can be specified as constant power factor The constant BCBV matrix has non-zero entries
model, constant voltage model or variable reactive consisted by line impedance values. The building
power model in the load flow analysis. In this paper, algorithm for BCBV matrix can be described as follows:
these three models are all derived and integrated into
the proposed load flow method. Test results show that Procedure 4. For a distribution system with m branch
the proposed method can be used to analyse the sections and n buses, the dimension of BCBV matrix is
penetration of DGs to distribution feeder effectively (n– 1)xm.
and efficiently. Moreover, the matrix structures and
the performance of original load flow method can be Procedure 5. If a line section l is located between bus i and
maintained. bus j, copy the row of the ith bus of BCBV matrix to the
row of the jth bus and fill the line impedance (Zij) to the
2 Proposed three-phase load flow position of the jth bus row and the lth line column.
The proposed load flow method was developed based on Procedure 6. Repeat procedure 5 until all line sections are
the BIBC and BCBV matrices. These two matrices included in the BCBV matrix.
provide a novel viewpoint in observing the relations
between bus voltages, branch currents and bus current Note that for radial distribution networks, m ¼ n –1.
injections. The detailed derivation of these two In additional, for multiple line sections and buses, the
matrices can be found in [10]. In this paper, only the algorithm can easily be expanded. For example, if the
building algorithms of these two matrices and solution line section between bus i and bus j is a three-phase
procedure are shown. The relation between current line section, the corresponding branch current Bl will
injections and branch currents can be expressed as be a 3  1 vector, the þ1 in the BIBC matrix
becomes a 3  3 identity matrix and the Zij in the
[B] ¼ [BIBC][I] (1) BCBV matrix is a 3  3 impedance matrix as shown
in (4). Note that the effects of the neutral or ground
where [B] and [I] are the vectors of branch currents and wire are included in (4)
bus current injections, respectively.
2 3
Z aan Z abn Z acn
The constant BIBC matrix is an upper triangular
matrix and has non-zero entries of þ1 only. The [Z abc ] ¼ 4 Z ban Z bbn Z bcn 5 (4)
building algorithm for BIBC matrix can be described Z can Z cbn Z ccn
as follows:
where Zxy2n is the impedance between phases x and y
Procedure 1. For a distribution system with m branch with the ground effect included.
sections and n buses, the dimension of BIBC matrix is
mx(n–1). The BIBC and BCBV matrices are developed based on
the topological structure of distribution feeders. The
Procedure 2. If a line section l is located between bus i and BIBC matrix represents the relation between bus
bus j, copy the column of the ith bus of BIBC matrix to current injections and branch currents. The
the column of the jth bus and fill a þ1 to the position of corresponding variations at branch currents, caused by
the lth line row and the jth bus column. the variations at bus current injections, can be

& The Institution of Engineering and Technology 2008 IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340/ 331
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

calculated directly by the BIBC matrix. The BCBV 3 Mathematical models of DGs
matrix represents the relation between branch
currents and bus voltages. The corresponding for load flow analysis
variations at bus voltages, caused by the variations at Fig. 1 illustrates the possible energy sources and
branch currents, can be calculated directly by the conversion devices of DGs [3]. Some DG types are
BCBV matrix. Combining (1) and (3), the relation not shown in Fig. 1; for example, the variable speed
between bus current injections and bus voltages can be wind turbine (WT), which employs the static power
expressed as converter and synchronous generator, is not shown in
Fig. 1. However, Fig. 1 can be used to categorise the
output characteristic of DGs. The energy sources of
[DV] ¼ [BCBV][BIBC][I] ¼ [DLF][I] (5) DGs can be categorised into stable energy sources,
such as fuel cell and micro-gas turbine, and unstable
and the solution for distribution load flow can be energy sources, such as wind and solar. Different
obtained by calculating (6) iteratively. energy sources combining with different energy
converters will result in special output characteristics.
  For example, if the induction generator is used to
P þ jQ convert wind energy to power grids, it will act like a
Iik ¼ Iir (Vik ) þ jIii (Vik ) ¼ i k i (6a) constant real power and variable reactive power
Vi
generator. However, if the static electronic converter
bDV kþ1 c ¼ [DLF]bI k c (6b) is used to convert wind energy to power grids, it will
mostly act like a generator with a constant power
bV kþ1 c ¼ bV 0 c þ bDV kþ1 c (6c) factor in normal operation condition. Therefore
according to the output power characteristics, DGs
can be specified as constant power factor model,
where k is the iteration number. constant voltage model or variable reactive power
model in the load flow analysis. In this paper, these
Although the DLF matrix is similar to the bus three models are all derived and integrated into the
impedance matrix which can also be built by the proposed load flow method.
conventional impedance matrix building algorithms;,
however, the BIBC and BCBV matrices provide a
novel viewpoint between bus current injections,
branch currents and bus voltages. Moreover, the DLF 3.1 Constant power factor model
matrix does not need to be built for radial feeders in The commonly used DG model is the constant power
the calculation of [DV] shown in (5). The [DV] can be factor model. The model can be used for the
calculated by (1) and (3) sequentially. The BIBC and controllable DGs, such as synchronous generator
BCBV matrices are upper and lower triangular based DGs and power electronic based DGs. For
matrices, respectively, and therefore the computational example, the output reactive power can be adjusted
time can be reduced. For weakly meshed feeders, a by controlling the exciting current and the trigger
spanning tree can be determined and other branches angles for synchronous generator based DGs and
which do not belong to the spanning tree are the power electronic based DGs, respectively. For this
meshed branches, and after that, the algorithm model, the specified values are the real power output
proposed in [10] can also be used to calculate the load and power factor of this DG. The reactive power of
flow solution. The search for a spanning tree and the the DGs can be calculated by (7a) and then the
decision of meshed branches are included in the
computer program. Besides, from the solution
techniques described above, the LU decomposition
and forward/backward substitution of the Jacobian
matrix or the Y admittance matrix is no longer
necessary for the proposed method. Only the BIBC
and BCBV matrices are needed to be built in the load
flow solution procedure. Therefore the proposed
method can save considerable computation resources
and this feature makes the proposed method suitable
for online operation. Since the proposed load flow
method can fully exploit the network structure of a
distribution feeder, it is important that the integration
of DGs into the proposed load flow does not degrade
the performance of the original algorithm. Figure 1 Possible combinations of DGs [3]

332 /IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340 & The Institution of Engineering and Technology 2008
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

equivalent current injection can be calculated by (7b).


After the equivalent current injection of the DG has
been calculated, the DG can be integrated into the
proposed load flow method.

Q i,g ¼ Pi,g tan(cos1 (pfi,g )) (7a)


! Figure 2 Equivalent circuit of induction generator based
r k i k
Pi,g þ jQ i,g WT
Ii,g ¼ Ii,g (Vi,g ) þ jIi,g (Vi,g ) ¼ k
(7b)
Vi,g
Qci,g can be calculated by
where Pi,g and pfi,g are the specified real power output  k 2
V 
and power factor for the DG installed in bus i and  i,g 
Q i,g the calculated reactive power output. Qci,g ¼ Qc,0
i,g  0  (10)
Vi,g 
If the internal circuit of the DG needs to be taken into
account, then the algorithms proposed in [8] can be used where Qc,0 c,0
i,g and Vi,g are rated reactive power and
k
to calculate the required reactive power output. voltage of this capacitor bank, respectively. Vi,g is the
voltage calculated at the kth iteration.

3.2 Variable reactive power model For multiple-step capacitor banks, the reactive power
output is dependent on the control algorithm designed
DGs employing induction generators as the power in the PLC; however, the commonly used algorithm is
conversion devices will act mostly like variable to maintain constant power factor. Therefore the
reactive power generators. By using the induction following equations can be integrated into load flow
generator based WT as an example, the real power solution for capacitor bank control
output can be calculated by the WT power curve.
Then, its reactive power output can be formulated as
a function comprising the real power output, bus
c,obj
Q i,g ¼ Pi,g tan(cos1 (pfi,g ))  Q i,g 1
(11a)
voltage, generator impedance and so on. However, the  k 2  k 2
X
n1 V  Xn V 
reactive power calculation like that is cumbersome and  i,g  obj c,0  i,g 
if Q c,0 
i,g,w 0   Q i,g  Q i,g,w  0 
difficult to calculate efficiently. From a steady-state
w¼1
Vi,g  w¼1
Vi,g 
viewpoint, reactive power consumed by a WT can be
 k 2
represented as a function of its real power [7], that is X n V 
 i,g 
thenQ ci,g ¼ Q c,0
i,g,w  0 
Vi,g 
Q1i,g ¼ Q 0  Q 1 Pi,g  Q 2 Pi,g
2
(8) w¼1
 k 2
X Nc V 
c c,0  i,g 
where Q i,g is the reactive power function consumed by elseQ i,g ¼ Q i,g,w  0  (11b)
the WT. Q 0 , Q 1 and Q 2 are experimentally obtained. w¼1
Vi,g 

The consumed reactive power cannot be fully where Qc,obj


i,g is the required reactive power for constant
provided by the distribution system, and therefore power factor, Nc the number of capacitor banks and n
capacitor banks are installed for power factor the steps of capacitor banks which are necessary to
correction. For example, Fig. 2 shows the possible connect to the WT.
configuration of induction generator based WT. The
reactive power output can be expressed as 3.3 Constant voltage model
Q i,g ¼ Q1i,g þ Qci,g (9) This model is used for large-scale controllable DGs. The
specified values of this DG model are the real power
output and bus voltage magnitude. To calculate the
where Qci,g is the reactive power supplied by capacitor equivalent current injection for the proposed load
banks. flow method, a two-loop algorithm is developed in
this paper. The inner loop calculates the reactive
The capacitor banks used in the WT can be the power output of the DG which is necessary to keep
single-step bank without switching or the multiple- the bus voltage magnitude on the specified value.
step banks controlled by a programming logic After the required reactive power output has been
controller, (PLC). For a single-step capacitor bank, obtained, the outer loop calculates the equivalent

& The Institution of Engineering and Technology 2008 IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340/ 333
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

current injection and then integrates it into the power 4 Test results and discussions
flow analysis. The detailed derivation can be found in
Section 8. For a generator with constant voltage The proposed distribution load flow program was
model at bus i, the reactive power for the mth implemented by Borland Cþþ language on a
inner iteration and the kth outer iteration can be Windows-XP based Pentium-III PC. Fig. 3 shows an
calculated by 11.4 kV feeder acquired from Taiwan Power Company
 1 used in the following tests. Tables 1 and 2 are the
mis
DQ k,m
i,g ¼ Vi,g 2bX g c (12a) load data and line length for this feeder, respectively.
    Fig. 4 shows the bus voltage profiles for phases A, B
mis  spec  2  k,m  2 and C before DG installation. Fig. 4 also shows that
Vi,g ¼ Vi,g   Vi,g  (12b)
this feeder is unbalanced, and therefore a three-phase
  load flow analysis can give an exact solution. Table 3
[X g ] ¼ img bBCBVi cbBIBCi c (12c) gives the data of a WT with a rated output of 660 kW
and rated voltage 690 V [11]. The reactive power
where DQk,mi,g is the required reactive power variation output of the WT is shown in Fig. 5, and therefore
for the mth inner iteration and the kth outer iteration, the reactive power output function can be obtained
V mis
i,g the square voltage mismatch between the from Table 3 by means of polynomial interpolation.
specified voltage and calculated voltage for the mth The reactive power output function and the reactive
inner iteration and the kth outer iteration, bBIBCi c power of the capacitor bank can be expressed as (15a)
the column vector of [BIBC] corresponding to bus i and (15b), respectively
and bBCBVi c the row vector of [BCBV] corresponding
to bus i. Q1g ¼ 0:0004  Pg2 þ 0:0395  Pg  182:3465 (15a)
After DQ k,m  k 2 !2
i,g has been obtained, the generation at V 
bus i can be modified as c  g
Q g ¼ 175  0  (15b)
Vg 
k,mþ1
Pi,g þ jQ k,mþ1
i,g
k,m
¼ Pi,g þ j(Q k,m k,m
i,g þ DQ i,g ) (13)
If two WTs were installed in a new bus named Bus 29
Then the voltage at bus i can be easily calculated by (1) and connected to Bus 10, the effects of these two WTs
and (3), and therefore the voltage of bus i for the for the feeder can be analysed. For example, assuming
(m þ 1)th inner iteration can be expressed as that these two WTs are operated at their maximum
k,mþ1 k,m k,m
power output (1.32 MW), the impacts of various
Vi,g ¼ Vi,g þ DVi,g (14) capacitor banks to system voltages can be studied.
Since the single-step capacitor bank is used for power
The inner loop will be iterated until DQ k,m factor correction of the WT, this kind of generators
i,g is less
than the predetermined tolerance. In general, it may can be analysed by a variable reactive power model.
need one to three iterations. Note that when the
Q k,mþ1
i,g exceeds the reactive power capacity limitation
of this generator bus, the generator bus will be
changed to constant power factor model accordingly.

From the above derivations, it can be seen that the


additional procedure required for the interconnection
of different types of DGs to the proposed load flow
algorithm is to find the corresponding equivalent
current injections. The special structures of the
proposed matrices and solution technique do not need
significant modification. Therefore the solution
performance of the original algorithm including
computational efficiency and solution stability can be
maintained and the penetrations of DGs can be
analysed effectively and efficiently. Besides, since the
equivalent current injections for different types of
DGs can be calculated by the proposed formulations,
those models can also be used for the other load flow
algorithms such as Gauss implicit Z-matrix method
and forward/backward sweep algorithm. Figure 3 Test feeder

334 /IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340 & The Institution of Engineering and Technology 2008
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

Table 1 Load data of test feeder

Bus number Phase A Phase B Phase C


P, kW Q, kVAR P, kW Q, kVAR P, kW Q, kVAR
1 139.9 90.4 31.9 20.6 31.9 20.6
2 0 0 0 0 0 0
3 28.4 48.3 56.0 – 0.4
4 154.1 73.8 74.6 24.5 74.6 24.5
5 101.5 76.2 101.5 76.2 183.4 124.7
6 29.9 69.9 75.5 9.1
7 26.1 12.6 26.1 12.6 84.0 40.7
8 37.7 67.1 77.0 0.9
9 105.0 20.2 34.6 56.2
10 84.0 63.0 84.0 63.0 141.5 98.7
11 113.2 61.1 113.2 61.1 212.4 109.1
12 165.1 91.3 81.4 46.1 81.4 46.1
13 367.3 133.3 367.3 133.3 367.3 133.3
14 30.5 13.9 30.5 13.9 108.9 58.3
15 75.3 85.9 88.8 – 6.5
16 20.5 33.2 104.6 39.6
17 74.4 64.6 34.0 1.1
18 67 44.8 14.2 0.2
19 67.5 109.5 128.5 – 3.7
20 98.4 33.6 36.2 13.1 36.2 13.1
21 7.8 8.8 11.5 – 2.4
22 29.9 16.1 73.9 49.1 29.9 16.1
23 173.7 68.3 72.1 34.9 72.1 34.9
24 47.9 – 3.5 75.4 69.7
25 88.2 23.2 26.0 48.5
26 75.7 19.0 75.7 19.0 207.0 51.9
27 23.5 1.2 87.2 66.3
28 79.5 49.3 79.5 49.3 187.7 130.4
total demand 2053.7 1046.2 1787.1 1088.4 2350.7 1038.2

Fig. 6 shows the voltage profiles of phase C for WTs with banks on bus voltages can be clearly observed.
capacitor capacities of 175, 225, 275 and 375 kVAR, Multiple-step capacitor banks can also be integrated
respectively. Note that in order to clarify the variation into these two WTs and the impacts of wind
of bus voltages with respect to different capacitor generation can be studied. For example, if multiple-
capacities, only the bus voltages of the main feeder step capacitor banks were installed and designed to
(Buses 1 –10) are shown in Fig. 6. From Fig. 6, it can operate at a power factor of 0.99 lagging, 0.95 lagging
be seen that DGs can improve the voltage profile of and 0.95 leading, the voltage profiles are calculated
the feeder; besides, the effects of different capacitor and shown in Fig. 7. Comparing Figs. 6 and 7, the

& The Institution of Engineering and Technology 2008 IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340/ 335
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

Table 2 Line length of test feeder

Line number From bus To bus Length, m


1 S/S 1 1060
2 1 2 1
3 2 3 33.7
4 3 4 31.6
5 4 5 180
6 5 6 68.4
Figure 4 Voltage profiles of test feeder without DG
7 6 7 121 installation
8 7 8 26.3
9 8 9 115.8 The proposed DG models including constant power
factor model and variable reactive power model can
10 9 10 252.5 be easily integrated into other three-phase load flow
11 1 11 50 methods such as Gauss implicit Z-matrix method and
forward/backward sweep algorithm. The constant
12 11 12 54.7 voltage model cannot be integrated into other load
13 2 13 242 flow methods directly. However, if the BIBC and
BCBV matrices and (12) – (14) were used, the
14 13 14 294.2 proposed constant-voltage DG model can also be used
15 14 15 52.6 for other load flow methods. For example, the
forward/backward sweep algorithm and the proposed
16 15 16 50 constant voltage model can be used to solve the
17 13 17 17.9 problem stated above. Fig. 9 shows the voltage
profiles. Comparing Fig. 8b with Fig. 9, it can be seen
18 17 19 21
19 18 19 47.4
Table 3 WT generator data
20 4 20 21
Rated voltage Three-phase 690 V
21 20 21 3
rated frequency 60 Hz
22 5 22 21
number of poles
23 5 23 25
4
24 6 24 63
rated speed
25 24 25 40 1890 RPM
26 7 26 50.5 rated power
output 660 kW
27 26 27 45
Generator power factor
28 9 28 121
full load
0.89
differences between single-step capacitor banks and three-fourth load
multiple-step capacitor banks can be clearly observed. 0.88
If these two WTs are replaced by power converter
two-fourth load
based fuel cells and operated at constant voltage, the
0.84
voltage profiles as shown in Fig. 8 can be obtained. In
this case, the specified voltage magnitude is 0.97 p.u. one-fourth load
Therefore from Fig. 8a, it can be seen that the 0.66
voltages at Bus 29 are 0.97 p.u. and the unbalanced
power factor
condition of this feeder is improved. Fig. 8b shows the
correction single-step capacitor bank
voltage profiles for Buses 1 – 10; it can be used to
with capacity 175 kVAR
compare the results obtained for Figs. 5–7.

336 /IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340 & The Institution of Engineering and Technology 2008
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

Figure 5 Reactive power of the WT shown in Table 1

Figure 6 Bus voltage profile with respect to different


capacitor banks (single-step capacitor bank)
Figure 8 Voltage profiles of DGs operated in constant
voltage model
that almost the same solutions are obtained. Since the a Bus voltage profiles of test feeder
characteristics of the original load flow algorithm are b Bus voltage profiles for Bus 1 to 10
maintained, the proposed method can be used to
calculate the load flow solutions under severe
operation conditions. For example, if the load at phase
C of Bus 28 is changed to 938.5 kW þ j652 kVAR,
the voltage profile of phase C as shown in Fig. 10 can
be obtained. From Fig. 10, it can be seen that the
voltage at Bus 10 has gone down to 0.9243 p.u.; it
means that the proposed method is robust even in the
heavy loading condition.

Test results show that the proposed DG models can


be integrated into the proposed load flow analysis
effectively and efficiently. Therefore the impacts of
Figure 9 Voltage profiles obtained by forward/backward
sweep algorithm

Figure 7 Bus voltage profile with respect to different power


factors (single-step capacitor bank) Figure 10 Voltage profile of phase C under heavy loading

& The Institution of Engineering and Technology 2008 IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340/ 337
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

different types of DGs on the distribution feeder can be [6] PERSAUD S. , FOX B., FLYNN D. : ‘Impact of remotely
analysed. It will be very helpful for the planning and connected wind turbines on steady state operation of
operation of DG interconnection. radial distribution networks’, IEE Proc., Gener. Transm.
Distrib., 2000, 147, (3), pp. 157– 163

5 Conclusions [7] FEIJOO A.E., CIDRAS J.: ‘Modeling of wind farms in the load
This paper integrated the mathematical models of DGs flow analysis’, IEEE Trans. Power Syst., 2000, 15, (1),
into the proposed three-phase distribution load flow pp. 110– 115
program. According to the characteristics of power
output, DGs can be specified as constant power factor [8] CHEN T.H., CHEN M.S., INOUE T., ET AL .: ‘Three-phase
model, constant voltage model or variable reactive cogenerator and transformer models for distribution
power model in the load flow analysis. In this paper, system analysis’, IEEE Trans. Power Deliv., 1991, 6, (4),
these three models were all derived and integrated pp. 1671 – 1681
into the proposed load flow method. Besides, since
the equivalent current injections for different types of [9] CHENG C.S., SHIRMOHAMMADI D.: ‘A three-phase power flow
DGs can be calculated by the proposed formulations, method for real-time distribution system analysis’, IEEE
those models can also be used for the other load flow Trans. Power Syst., 1995, 10, (2), pp. 671– 679
algorithms. Test results show that the proposed
method can be used to analyse the penetration of DGs [10] TENG J.H.: ‘A direct approach for distribution system
to distribution feeders effectively and efficiently. load flow solutions’, IEEE Trans. Power Deliv., 2003, 18,
Besides, the structures of the proposed matrices and (3), pp. 882– 887
the performance of the original load flow method are
maintained. Other applications such as asymmetrical [11] Vestas Corp, available at: http://www.vestas.dk
short-circuit current calculation [12], optimal
capacitor placement, feeder reconfiguration problem [12] TENG J.-H.: ‘Systematic short circuit analysis method for
and so on, can also be solved by the proposed BIBC unbalanced distribution systems’, IEE Proc., Gener. Transm.
and BCBV matrices, and some more applications will Distrib., 2005, 152, (4), pp. 549– 555
be derived in future research.
8 Appendix
6 Acknowledgment For a DG operated in a constant voltage model on bus i,
This work was sponsored by National Science Council, the specified voltage and the voltage after the load flow
Taiwan, under research grant NSC 91-2213-E-214-048. solution of the mth
spec
inner iteration and the kth outer
k,m
iteration are Vi,g and Vi,g , respectively. If the
reactive power of the DG has a variation of DQ k,m i,g ,
7 References then the corresponding variation of equivalent current
injection at bus i can be expressed as
[1] WILLIS H.L., SCOTT W.G.: ‘Distributed power generation-
planning and evaluation’ (Marcel Dekker Inc., New York, !
2000) k,m
jDQ k,m
i,g fi,gk,m DQ k,m k,m k,m
i,g  jei,g DQi,g
DIi,g ¼ k,m ¼  k,m 2 (16a)
Vi,g Vi,g 
[2] RABINOWITZ M.: ‘Power systems of the future 1’, IEEE
k,m
Power Eng. Rev., 2000, 20, (1), pp. 5 – 16 Vi,g ¼ ek,m k,m
i,g þ fi,g (16b)
[3] ‘Engineering guide for integration of distributed
generation and storage into power distribution systems’,
EPRI Technical Report TR-100419 Report, December 2000 Substituting (16a) into (1), the branch current variations
caused by the variation of reactive power can be
[4] NAKA S., GENJI T., FUKUYAMA Y. : ‘Practical equipment written as
models for fast distribution power flow considering
interconnection of distributed generators’. IEEE PES h i h iT
k,m
Summer Meeting, 2001, vol. 2, pp. 1007 – 1012 DBk,m
g ¼ [BIBC] 0  DIi,g 0  

[5] PECAS LOPES J.A. , MACIEL BARBOSA F.P., CIDRAS PIDRE J. : (17)
‘Simulation of MV distribution networks with
asynchronous local generation sources’. IEEE Electro- where bDBk,m
g c is the variation of the branch current
technical Conf. 6th Mediterranean, 1991, vol. 2, vector for the mth inner iteration and the kth outer
pp. 1453 – 1456 iteration.

338 /IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340 & The Institution of Engineering and Technology 2008
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

Equation (17) can be written as Equation (22) can also be expressed in matrix form as
2 3
h i  k,m fi,gk,m
" k,m # " #6  2 7
DBk,m
g ¼ BIBCi DIi,g (18) Dei,g Rg X g 6 Vi,g k,m  7
¼ 6 7DQ k,m
k,m
Dfi,g X g Rg 4 ei6 k,m 7 i,g (23a)
5
 k,m 2
where [BIBCi] is the column vector of [BIBC] corresponding Vi,g 
to bus i.
Substituting (18) into (3), the bus voltage variation k,m
DVi,g ¼ Dek,m k,m
caused by the variation of reactive power can be i,g þ jDfi,g (23b)
written as
k,m k,m
The magnitude summation of DVi,g and Vi,g is equal to
h i  k,m the specified bus voltage, that is
DV k,m
g ¼ [BCBV] BIBCi DIi,g (19) k,m k,m 2 spec 2
(Vi,g þ DVi,g )  (Vi,g ) ¼0 (24)

where bDV k,m


g c is the the variation of bus voltage vector
Equation (24) can be rewritten as
for the mth inner iteration and the kth outer iteration. k,m k,m 2 spec 2 k,m 2 k,m 2
(Vi,g þ DVi,g )  (Vi,g ) ¼ (Vi,g ) þ (DVi,g )
The voltage variation for bus i can therefore be spec 2
expressed as þ 2(ek,m k,m k,m k,m
i,g Dei,g þ j fi,g Dfi,g )  (Vi,g ) (25)

k,m 2
k,m k,m The term of (DVi,g ) can be omitted because of
DVi,g ¼ bBCBVi cbBIBCi cDIi,g its minor effect on (25), and therefore (10) can be
h i (20) rewritten as
k,m
¼ Z g DIi,g
spec 2 k,m 2
2(ek,m k,m k,m k,m
i,g Dei,g þ j fi,g Dfi,g ) ¼ (Vi,g )  (Vi,g ) (26)

where [BCBVi] is the row vector of [BCBV] Substituting (23) into (26), the required variation of
corresponding to bus I and [Zg] the relationship reactive power for the DG bus i in the (m þ 1)th
matrix between the voltages and current injections of inner iteration and the kth outer iteration can be
the DG buses. The matrix dimension of [Zg] is equal calculated by
to the number of the DG buses.
spec 2 k,m 2
(Vi,g )  (Vi,g ) ¼ 2bX g cDQ k,m
i,g (27)
Equation (20) can be further rewritten as
spec
fi,gk,m DQ k,m k,m k,m Let (Vi,g )2  (Vi,g
k,m 2 mis
) ¼ Vi,g , (27) can be expressed as
k,m i,g  jei,g DQ i,g
DVi,g ¼ ð[Rg ] þ j[X g ])  k,m 2 (21a) mis 1
Vi,g  DQ k,m
i,g ¼ Vi,g (2bX g c) (28)

bZ g c ¼ bRg c þ jbX g c (21b) Therefore the generator power at the (m þ 1)th inner
iteration and the kth outer iteration can be expressed as
k,mþ1
Pi,g þ jQ k,mþ1
i,g
k,m
¼ Pi,g þ j(Q k,m k,m
i,g þ DQ i,g ) (29)
The real and imaginary part of (21) can be expressed as
Bus voltage can be updated by using (10) and
fi,gk,m DQ k,m ek,m k,m expressed as
i,g i,g DQ i,g
Dek,m
i,g ¼ ½Rg   k,m 2 þ ½X g   k,m 2 (22a)
Vi,g  Vi,g  k,mþ1
Vi,g k,m
¼ Vi,g k,m
þ DVi,g (30)
ek,m k,m
i,g DQ i,g fi,gk,m DQ k,m
i,g The inner loop will be iterated until the variation of
Dfi,gk,m ¼ ½Rg   k,m 2 þ ½X g   k,m 2 (22b)
Vi,g  Vi,g  DQ k,m
i,g is less than a predetermined tolerance. After
the required reactive power output has been obtained,

& The Institution of Engineering and Technology 2008 IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340/ 339
doi: 10.1049/iet-gtd:20070165
www.ietdl.org

the outer loop calculates the equivalent current injection iteration can be calculated by the following equations
and then integrates it into the power flow analysis. " #
BCBVi 
[Z g ] ¼ BIBCi BIBCj (31a)
The derivations shown above can be easily extended BCBVj
to multiple DG buses operated in a constant voltage " # " mis #
model. For example, if two DGs operated in a DQ k,m V 0  h i1
i,g i,g
constant voltage model were installed on buses i ¼ mis 2 Xg (31b)
and j, the reactive power variations in each inner loop DQ k,m
j,g 0 V j,g

340 /IET Gener. Transm. Distrib., 2008, Vol. 2, No. 3, pp. 330 – 340 & The Institution of Engineering and Technology 2008
doi: 10.1049/iet-gtd:20070165

S-ar putea să vă placă și