The HK Polytechnic University
Load Flow
Load Flow Analysis
A load flow or power flow is power system jargon for the steady-state solution
of a power system network subject to certain operational constraints, such as:
Generation supplies the demand (load) plus losses.
Bus voltage magnitudes remain close to rated values.
1
Generators operate within specified real and reactive power limits.
Transformer tap settings are within limits.
Transmission lines and transformers are not overloaded.
Load flow solution gives the nodal voltages and phase angles and hence the
power injection at all buses and power flows though transmission units such as
lines, cables and transformers.
Load flow calculations are performed for power system planning, operational
planning and in connection with system operation and control.
Load flow studies are performed to investigate the following features of a power
system network:
1. Flow of MW and MVAr in the branches of the network.
2. Busbar (node) voltage.
3. Effect of the following changes on system loading:
(a) Rearranging circuits and incorporating new circuits.
2
(b) Temporary loss of generation and transmission circuits.
(c) Injecting in-phase and quadrature boost voltages.
4. Optimum system running conditions and load distribution.
5. Minimising system losses.
6. Optimum rating and tap-range of transformers.
7. Improvements from change of conductor size and system voltage.
Studies will normally be performed for various load conditions to ensure the
power network behaves properly under a wide range of operating conditions.
KWCn v1.31
The HK Polytechnic University
Load Flow
Load Flow Problem
Objectives: to determine the steady state operating conditions of (a) busbar
voltage, (b) generation, (c) branch power flows, and (d) circuit system loss.
Conventional nodal or loop analysis is not suitable for load flow studies because
loads are normally given in terms of power rather than impedance. Also,
generators are considered as power sources, not voltage or current sources.
Together with the power and voltage constraints, the load flow problem becomes
a nonlinear numerical problem formulated as a set of nonlinear algebraic
equations and the numerical solution must therefore be iterative in nature.
A load flow solution of the power system requires mainly the following steps:
1. Formulation of the network equations (load flow equations).
2. Suitable mathematical technique for solution of the equations
(Gauss-Seidel, Newton-Raphson, and Fast Decoupled methods).
Network Model Formulation
In a power system, each bus is associated with 4 quantities:
1. real and reactive powers, P & Q.
2. bus voltage magnitude and angle, |V | & .
Among these 4 quantities , only 2 can be specified and the remaining 2 are
4
obtained through the load flow solution.
Depending upon which quantities have been specified, the buses are classified
into three categories:
Bus Type
Quantities Specified
Quantities to be obtained
Load (PQ) bus
P, Q
|V |,
Generator (PV) bus
P , |V |
Q,
Slack (Swing) bus
|V |,
P, Q
KWCn v1.31
The HK Polytechnic University
Load Flow
3.1
Bus Classification
1. Load or PQ bus: The complex power (P , Q) is specified. It is desired to find out
the voltage magnitude and phase angle through the load flow solution loads.
2. Generation, PV or voltage control bus: The voltage magnitude and real
power are specified. Often limits to the value of the reactive power are giving as
well. It is required to find out the reactive power generation and the phase angle
5
of the bus voltage capacitors, synchronous compensators and generators.
3. Slack, swing or reference bus: Bus voltage magnitude and angle are
specified, typically 1.0/0o , whereas its power P , Q are obtained through the
load flow to cover any power loss, which is not known precisely in advance of
the calculation, or mismatch of load and power generation system frequency
control generators. This bus voltage angle will be taken as the reference. There
shall be only one such bus in a power system, and usually, the one with the
largest generation is assigned as the slack, swing or reference bus.
3.2
Nodal Admittance Matrix
Load flow formulation can be established by using either the loop or bus frame of
reference.
loop:
V = ZI
bus:
I =YV
where
Z : impedance matrix
V : voltage vector
Y : admittance matrix
I : current vector
Generally, bus frame of reference in admittance form is preferred as :
1. data preparation is simple
2. its formation and modification is easy
3. the bus admittance matrix is a sparse matrix (i.e. most of its elements are zero)
save computer memory and computational effort.
KWCn v1.31
The HK Polytechnic University
Load Flow
Consider the nodal current at node 1,
I1 = I11 + I12 + I13
= V1 y11 + (V1 V2 )y12 + (V1 V3 )y13
= V1 (y11 + y12 + y13 ) V2 y12 V3 y13
= V1 Y11 + V2 Y12 + V3 Y13
Three-bus system
y11 is the shunt charging admittance
I2
I1
at bus 1 and y12 is the series admittance
1
2
between bus 1 and 2, and
y12 = y21
I13
I12
I21
I 23
I11
Y11 = y11 + y12 + y13
I22
Y12 = y12
Y13 = y13
where
7
I33
Similarly for node 2 and 3,
I31
I2 = V1 Y21 + V2 Y22 + V3 Y23
I3 = V1 Y31 + V2 Y32 + V3 Y33
I 32
3
I3
Nodal current equations can be written in a matrix form:
Y11
I1
I2 = Y21
Y31
I3
Y12
Y13
Y22
Y23
Y32
Y33
V1
V2
V3
or
I =YV
or in compact form these equations can be written as:
8
Ii =
3
X
Yij Vj
for
i = 1, 2, 3
j=1
The above nodal current equations can be generalised to an n bus system:
Ii =
n
X
Yij Vj
for
i = 1, 2,. . . , n
where
Yii =
j=1
n
X
yij
j=1
Yij = yij
It can be shown that the nodal admittance matrix is a sparse matrix (only a few
number of elements are non-zero) for an actual power system.
KWCn v1.31
The HK Polytechnic University
Load Flow
Consider the nodal admittance matrix for a 5 bus system.
Y11
Y
21
Y = 0
Y41
Y51
Y12
Y14
Y22
Y23
Y32
Y33
Y34
Y43
Y44
Y54
Y15
y12 = y21
y23 = y32
Y45
y14 = y41
y34 = y43
y15 = y51
Y55
For line, cable and tapped transformer:
Yij = Yji = yij = yji
4
y45 = y54
Yij and Yij are non-zero only if there is a connection between bus i and j .
The diagonal element of each node is the sum of admittances connected to it.
The off-diagonal element is the negated admittance between the nodes.
3.3
Network Models
Power network can be operating under balanced or unbalanced conditions. The
normal procedure for a load flow study is to assume a balanced system and to
use a single-phase representation equivalent to the positive sequence network.
Shunt Branches Rectors & Capacitor:
10
Shunt admittances are add to the diagonal elements, Yii , corresponding to the
nodes at which they are connected.
Lines & Cables:
Modelled as a equivalent.
Contribute to both the diagonal and
off-diagonal matrix elements
y ij
y ii
y jj
Yii , Yjj , Yij and Yji .
KWCn v1.31
The HK Polytechnic University
Load Flow
Tapped Transformers:
1:a
Ij
Ii
ay
Ii
Ij
Vj
aVi
Vi
(a2 - a)y
(1 - a)y
11
Ij
y (Vj aVi )
Ii
aIj = y a2 Vi aVj
The equivalent circuit is just like an asymmetric network.
Ii
((a2 a)y + ay)Vi ayVj = (yii + yij )Vi yij Vj
Ij
ayVi + ((1 a)y + ay)Vj = yij Vi + (yjj + yij )Vj
3.4
Load Flow Equations
Recall the nodal current equations:
Ii
n
X
Yij Vj
for i = 1, 2, 3, . . . , n
j=1
Yii Vi +
n
X
Yij Vj
j=1,j6=i
12
or Vi
1
Yii
1
Yii
Ii
n
X
j=1,j6=i
Pi jQi
Vi
Yij Vj
n
X
j=1,j6=i
Yij Vj
with Si = Vi Ii = Pi jQi
The above load flow equations are nonlinear and can be solved by iterative methods
such as the Gauss-Seidel and Newton-Raphson methods.
KWCn v1.31
The HK Polytechnic University
Load Flow
3.5
Power Calculations
The complex power Si delivered to bus i is:
Si = Pi + jQi =
Vi Ii
n
X
= Vi
Yij Vj
j=1
Express Vi , Vj and Yij in polar coordinate (magnitude & angle):
ji
13
Pi + jQi = |Vi |e
n
X
jij
|Yij |e
jj
|Vj |e
= |Vi |
j=1
n
X
|Vj ||Yij |ej(i j ij )
j=1
The above complex equation can be expressed in the polar form:
Pi
|Vi |
n
X
|Vj ||Yij | cos(i j ij )
(1)
|Vj ||Yij | sin(i j ij )
(2)
j=1
Qi
|Vi |
n
X
j=1
where
Yij = |Yij |/ij
busbar voltage
Vi = |Vi |/i
net scheduled real power
Pi
(calculated)
Vi
Bus i
Pi(scheduled) = PGi - PLi
net scheduled reactive power
Qi
Pi
Qi
(scheduled)
(scheduled)
PGi
QGi
(calculated)
PLi
QLi
From
other
buses
Qi(scheduled) = QGi - QLi
14
where PGi & QGi is generator power
Gen
Load
PLi & QLi is load power
The power difference between the scheduled value (Pi(scheduled), Qi(scheduled)),
specified by the busbar generation and load, and the calculated value (Pi(calculated),
Qi(calculated)), derived from the best available busbar voltages and angles, is referred
to as the power mismatch (Pi, Qi) where
Pi = Pi(scheduled) - Pi(calculated)
Qi = Qi(scheduled) - Qi(calculated)
KWCn v1.31
The HK Polytechnic University
Load Flow
Gauss-Seidel (GS) Method
The GS method is an iterative algorithm for solving a set of non-linear algebraic
equations.
To start with, a solution vector is assumed. One of the equations is then used to
15
obtain the revised value of a particular variable by substituting in it the present
values of the remaining variable. The solution vector is immediately updated in
respect of this variable.
The process is then repeated for all the variables thereby completing one
iteration. The iterative process is then repeated till the solution vector converges
within prescribed accuracy.
The convergence is sensitive to the starting values assumed.
Eg. Consider the following simple nodal equations:
2V1 0.5V2 1.5V3 =
I1
=1
0.5V1 + 1.25V2 0.75V3 =
I2
= 1.5
1.5V1 0.75V 2 + 2.25V 3 =
I3
where V3 is 100 and V2 is assumed to be 100 initially. Find V1 and V2 .
In the iterations, the newly computed values are immediately used as soon as they are
16
obtained.
V1
1
V2
2
V1
2
V2
3
V1
4
V1
5
V1
(1 + 150 + 50)/2
100.5000
(-1.5+75+50.25)/1.25
99.0000
(1 + 150 + 49.5)/2
100.2500
(-1.5+75+50.125)/1.25
98.9000
100.2250
98.8900
100.2225
98.8890
100.22225
98.8889
V2
V2
5
V2
Hence, I3 = -1.5 100.22225 - 0.75 98.8889 + 2.25 100 = 0.49995
KWCn v1.31
The HK Polytechnic University
Load Flow
4.1
Load Flow Solution by Gauss-Seidel (GS) Method !
n
X
1
Pi jQi
Recall the load flow equation: Vi =
Yii
Yij Vj
Vi
(3)
j=1,j6=i
In GS method, the new calculated voltage Vik+1 immediately replaces Vik and
is used in the solution of the subsequent equations. Hence, eqn
! (3) becomes:
Vik+1
17
1
=
Yii
i1
X
Pi jQi
Yij Vjk+1
(Vik )
j=1
n
X
Yij Vjk
(4)
j=i+1
For PV bus, Qi is unknown but can be calculated from power eqn (2).
For slack bus, its load flow equation is excluded from the GS calculation as both
its voltage magnitude |Vi | and angle i are specified while the 2 unknown
variables Pi , Qi can be calculated from power eqn (1) and (2), i.e. there are
(n1) load flow equations in total for a n bus system.
Initial unknown voltage magnitude |Vi | and angle i can be set up 1pu and 0o .
This is referred as the flat start condition.
4.2
Gauss-Seidel Solution Steps
1. Formulate the admittance matrix Y .
2. Separate out the slack, generator and load buses.
3. Assume any unknown bus voltage to, say, 1pu and 0o .
4. Start iteration process with first bus of the system (i=1).
18
5. Update Qi using equation (2) if i bus is a slack or PV bus.
Calculate the new bus voltage, Vi , from the load flow equation (4).
6. Calculate the difference between old and new bus voltages.
Vik+1 = Vik+1 Vik
7. Using this new value of bus voltage in performing calculations for the next bus of
the system, except for the PV buses whose |Vi | should remain constant.
8. Advance for the next bus of the system and repeat steps 5 to 7 until a new set of
values of bus voltages of all buses in the system is obtained 1 GS iteration.
KWCn v1.31
The HK Polytechnic University
Load Flow
9. Repeat the iterative process from step 4 to 8 until the difference Vi for all
buses is within a specified limit or tolerance.
|Vik+1 | <
where k is the iteration count and is the tolerance level.
4.3
19
Reactive Power Limits
For a generation bus, Qi should be checked for any limit violation.
Qimin Qi Qimax
Whenever there is a limit violation, Qi will be set to the limit and the bus type will be
switched to load, i.e. PQ, as it is not possible to keep the generator terminal voltage
to the specify voltage (Vsp ) while Qi is being limited.
When bus type switched, the bus voltage is also needed to be corrected to cater for
the Qi being limited. Once, Qi becomes within the limits, the bus type and terminal
voltage can be restored.
4.4
Convergence Limits (Tolerance at Solution)
Usually 0.001, 0.0001 or 0.00001 (p.u. volts).
That is all Vik+1 must lie within this tolerance.
4.5
20
Acceleration Factors
In practice, it is found that the process of convergence due to Gauss-Seidel
method is slow. A large number of iterations is required to obtain an accurate
solution.
The rate of convergence can be increased by the use of acceleration factors.
Vik+1
= Vik + (Vik+1 Vik )
acc
where is known as acceleration factor and best values for the acceleration
factor lie in the range of 1.1 to 1.6 with 1.4 is the most common used value
in practice.
KWCn v1.31
10
The HK Polytechnic University
Load Flow
Newton-Raphson Method (NR)
The NR is a powerful method widely used for solving nonlinear equations.
the original nonlinear problem was transformed into a sequence of linear
problems whose solutions approach the solution of the original problem.
the method can be applied to one equation in one unknown or to a system of
simultaneous equations with as many unknowns as equations.
21
5.1
One-Dimensional Case
F(x)
The basic iterative procedure for solving
equation F (x)
p+1
= 0 is as follows :
F (xp )
p
=x p
F (x )
F0
F2
F1
p is the number of iterations and
F (xp ) is the derivative of F (x) at xp .
where
x2
X2
5.2
x1
x0
X1
N-Dimensional Case
The single dimensional concept of the NR method can be extended to N
dimensions. All that is needed is an N -dimensional analog to the first
derivations. This is provided by the Jacobian matrix J .
Each of the N rows of J is composed of the partial derivatives of one of the
equations of the system with respect to each of the N variables.
22
Thus, the basic NR method extended to a N -dimensional system becomes:
xp+1 = xp
1
F (xp )
p
J (x )
(3)
where x and F are column vectors and J is the Jacobian matrix, of the from :
KWCn v1.31
J =
F1
x1
..
..
Fk
xk
..
FN
x1
..
FN
xN
F1
xN
(4)
11
The HK Polytechnic University
Load Flow
Instead of computing the inverse of J , eqn (3) can be written as :
F (xp )
y p
or
= J (xp )(xp+1 xp )
(5)
= J (xp )xp
(6)
where
F (x)
= F (xp ) + yp = 0
(7)
and
xp+1
= xp + xp
(8)
23
There are 4 steps for each iteration :
1. compute y p
2. compute J(xp )
3. solve for xp by Gauss elimination and back substitution
4. compute xp+1
5.3
Load Flow Solution by Newton-Raphson (NR) Method
First, rewrite the power flow equations (1) and (2) into an alternate form:
Pi
n
X
Gii |Vi | +
|Vi ||Vj ||Yij | cos(i j ij )
(9)
|Vi ||Vj ||Yij | sin(i j ij )
(10)
j=1,j6=i
Qi
n
X
Bii |Vi | +
j=1,j6=i
24
where
Gii
|Yii | cos(ii )
Bii
|Yii | sin(ii )
Then, apply the Newton-Raphson method to form the following mismatch equation:
"
Pi
Qi
"
Pi
i
Qi
i
Pi
|Vi |
Qi
|Vi |
#"
i
|Vi |
where Pi and Qi are the power mismatch at bus i and
KWCn v1.31
(11)
12
The HK Polytechnic University
Load Flow
Pi
Pi,scheduled Pi,calculated
(12)
Qi
Qi,scheduled Qi,calculated
(13)
Pi,calculated and Qi,calculated :
obtained from the power flow eqn (1) and (2).
Pi,scheduled :
Pi,scheduled = Pi,calculated and Pi = 0
(b) Otherwise (PV & PQ buses): Pi,scheduled = PGi PLi .
(a) Slack bus:
25
Qi,scheduled :
Qi,scheduled = Qi,calculated and Qi = 0
(b) Otherwise (PQ bus): Qi,scheduled = QGi QLi .
(a) Slack & PV bus:
To further improve the convergence, the mismatch equation can be rewritten as:
"
# "
#"
# "
#"
#
Pi
Qi
Pi
i
Qi
i
Pi
|Vi | |V
i|
Qi
|Vi | |V
i|
|Vi |
|Vi |
|Vi |
|Vi |
For a n bus system with m PQ buses, the mismatch equation becomes :
P1
26
H1,1
..
N1,1
H1,n1
..
N1,m
..
Ni,i
..
..
.. .. Hi,i
..
Pn1 Hn1,1 .. Hn1,n1 Nn1,1 .. Nn1,m n1
Q = J
|V1 |
1
1,1 .. J1,n1 L1,1 .. L1,m |V1 |
..
Li,i
..
..
.. .. Ji,i
..
Qm
For i
For i
= j,
6= j,
Jm,1
Hii
Pi
i
Nii
Jii
KWCn v1.31
Lm,1
Jm,n1
..
Lm,m
Qi Bii |Vi |2
|Vi | |Vi |
Pi + Gii |Vi |2
Qi
i
Pi Gii |Vi |2
Lii
|Vi | |V i|
Qi Bii |Vi |2
Hij
Pi
j
|Vi ||Vj ||Yij | sin(i j ij )
Nij
|Vj | |V i |
|Vi ||Vj ||Yij | cos(i j ij )
Qi
j
Nij
Q
|Vj | |V i|
j
Hij
Lij
(14)
|Vm |
|Vm |
Jij
..
13
The HK Polytechnic University
Load Flow
5.4
Steps in Newton-Raphson solution
1. Formulate the nodal admittance matrix Y .
2. Assume an initial set of bus voltages and set bus n as the reference bus.
3. Obtain the power injections Pi and Qi for all i
= 1, ...(n 1)
4. Obtain the power mismatches Pi and Qi for all i
27
= 1, ...(n 1)
5. Stop the iteration if all Pi and Qi are within tolerance.
6. Obtain the Jacobian matrix elements using the best available voltage values.
7. Substitute the values obtained from steps (4) & (6) in equation (14). Solve this
|V |
linear simultaneous equation by a suitable method for vectors [] and [ |V |i ].
i
8. Update i and |Vi | for all i, i.e.
ik+1 = ik + i
i|
Vik+1 = Vik (1 + |V
|Vi | )
9. Goto step (3).
5.5
Decoupled Load Flow (DFL)
An important characteristic of any practical electrical power transmission system
operating in steady state is strong interdependence between real powers and bus
voltage angles and reactive powers and voltage magnitudes.
If the P - and Q-V couplings are recognised to be much stronger than the P -V
and Q- couplings the sub-matrices N and J can be ignored. Then separate
equations:
28
[P ]
[Q]
[H] []
[L]
|V |
|V |
(15)
(16)
can be obtained and solved separately to give an approximate solution of |V | and .
Instead of the previous 2(n 1) 2(n 1) matrix problem, there are two
(n 1) (n 1) matrices to solve save memory and easier to solve but take
more number of iterations to converge because of the approximation.
Techniques such as these are often used in on-line (very fast) load flow solutions
and in the starting (initial stage) of conventional full length load flows.
KWCn v1.31
14
The HK Polytechnic University
Load Flow
5.6
Fast Decoupled Load Flow (FDLF)
Recall the equations for the Jacobian elements :
For i
For i
29
= j,
6= j,
where
Hii
Qi Bii |Vi |2
Lii
Qi Bii |Vi |2
Hij = Lij
|Vi ||Vj ||Yij | sin(i j ij )
|Vi ||Vj | [Gij sin(i j ) Bij cos(i j )]
Yij = Gij + jBij .
In addition to the DLF approximations, the following approximations can be make to
further speed-up the solution time and improve the convergence.
1.
cos(i j ) 1
2.
Gij sin(i j ) Bij
3.
Qi Bii |Vi |2
The previous results can be rewritten as follows:
= j, Hii = Lii Bii |Vi |2
For i 6= j, Hij = Lij |Vi ||Vj |Bij
For i
With the above approximations and taking |Vj |
1, the mismatch equations (15)
and (16) becomes:
30
Pi
|Vi |
Qi
|Vi |
[j ]
Bij
|Vj | [Bij ] [j ]
|Vj |
|Vj |
|Vj | [Bij ]
Bij
|Vj |
|Vj |
(17)
(18)
where [B ] and [B ] are made up of elements of [B] matrix and are constant
and need to be inverted or decomposed once only. Usually in building up [B ],
shunt reactors and off-nominal tap transformers are ignored while in building up
[B ], angle shifts of phase shift transformers are ignored.
The FDLF is extremely fast. The final result is exact since the iteration will only stop
when P and Q come within the specified tolerance of Pscheduled and Qscheduled .
KWCn v1.31
15
The HK Polytechnic University
Load Flow
5.7
Example for Newton Raphson and Fast Decoupled LF
V1 = 1.05/0o
V2 /
zser = 0.1 + j0.2
B1
B2
ysh = j0.15
ysh = j0.15
PL , QL
0.1 + j0.2
31
Find V2 by NR method with B1 as the slack bus and initial estimate for V2
P2
|V2 |2 G22 + |V1 ||V2 ||Y12 | cos(2 12 )
Q2
|V2 |2 B22 + |V1 ||V2 ||Y12 | sin(2 12 )
Power flow at B2:
= 1/0o .
Since B1 is the slack bus, only B2 mismatches are calculated.
"
P2
Q2
"
#"
2
|V2 |
|V2 |
"
P2
2
Q2
2
P2
|V2 | |V
2|
Q2
|V2 | |V2 |
#"
2
|V2 |
|V2 |
Admittance matrix:
A=
where
i.e.
32
yser + ysh
yser
yser
yser =
yser + ysh
1
0.1+j0.2
2 j3.85
2 + j4
2 + j4
2 j3.85
= 2 j4 and ysh = j0.15
Y12 = 2 + j4 = 4.472/116.56o
Y22 = 2 j3.85 = G22 + jB22 G22 = 2 and B22 = 3.85
H=
J=
KWCn v1.31
P2
2
Q2
2
|V1 ||V2 ||Y12 | sin(2 12 )
(1.05)(1.0)(4.472) sin(116.56o ) = 4.2
|V1 ||V2 ||Y12 | cos(2 12 )
(1.05)(1.0)(4.472) cos(116.56o ) = 2.1
16
The HK Polytechnic University
N = |V2 |
L = |V2 |
P2
|V2 |
Q2
|V2 |
"
"
34
1
2|V2 |2 G22 + |V1 ||V2 ||Y12 | cos(2 12 )
2|V2 |2 G22 +
2|V2 |2 B22 + |V1 ||V2 ||Y12 | sin(2 12 )
2|V2 |2 B22
P2
= 2(3.85) 4.2 = 3.5
2
"
4.2
1.9
2.1
3.5
Q2
= 2 2.1 = 0.1
2
P2
= 3.85 4.2 = 0.35
|V2 |2 B22
2
Q12
P21
PG PL P2 = 0.1 + 0.1 = 0
Q12
QG QL Q2 = 0.2 + 0.35 = 0.15
"
0
0.15
|V2 |2 G22 +
4.2
1.9
2.1
3.5
1
18.69
21
0.01525 rad
|V21 |
1.0337 p.u.
21
|V21 |
|V21 |
"
Similarly for the second iteration:
KWCn v1.31
#"
21
|V21 |
|V21 |
#"
3.5
1.9
2.1
4.2
0.15
"
P22
0.00037
Q22
0.00596
22 = 0.01475 rad
|V2 |
|V22 | = 1.03243 p.u.
Q2
= 2(2) 2.1 = 1.9
2
P21
33
Load Flow
J2
4.3080
2.0375
2.2367
3.9199
"
0.01525
0.0337
17
The HK Polytechnic University
Load Flow
Now find V2 again by FD method instead of the NR method.
Recall:
35
Y12 = 4.472/116.56o
G22 = 2
P21 = 0.1
P21 = 0
Q12 = 0.35
Q12 = 0.15
[B ] = [B22 ] = [3.85]
[B ] = [B22 ] = [3.85]
From (17):
From (18):
2 =
B22 = 3.85
P2
= 0 rad
B22 |V2 |
21 = 2 + 2 = 0 rad
Q2
= 0.03896 p.u.
B22
|V21 | = |V2 | + |V2 | = 1.03896 p.u.
|V2 | =
Update the calculated power injection (P2 , Q2 ) and mismatch (P2 , Q2 ) with
the latest bus voltage.
Repeat the above procedures, as shown below, until the solution converage or the
power mismatches are below the tolerance.
36
Iter
P2
Q2
P2
Q2
|V2 |
|V2 |
-0.1
-0.35
0.15
0.0390
1.0390
-0.0229
-0.2078
-0.0771
0.0078
0.0020
-0.0193
1.0410
-0.0193
-0.1025
-0.1572
-0.0026
-0.0428
-0.0111
0.0006
1.0299
-0.0186
-0.1216
-0.2010
-0.0216
0.0010
0.0003
-0.0055
1.0301
-0.0132
-0.0977
-0.2122
-0.0023
0.0122
0.0032
-0.0006
1.0333
-0.0137
-0.0940
-0.1989
-0.0060
-0.0011
-0.0003
-0.0015
1.0330
-0.0153
-0.1010
-0.1966
0.0010
-0.0034
-0.0009
0.0003
1.0321
-0.0150
-0.1017
-0.2005
0.0017
0.0005
0.0001
0.0004
1.0323
-0.0146
-0.0996
-0.2009
-0.0004
-0.0002
0.0002
-0.0001
1.0325
-0.0147
10
-0.0996
-0.1998
0.0001
-0.0003
-0.0001
0.0000
1.0324
-0.0148
KWCn v1.31
18
The HK Polytechnic University
Load Flow
Comparison of Load Flow Methods
GS method works well when programmed using rectangular coordinates,
whereas NR requires more memory when rectangular coordinates are used.
Though GS method requires the fewest number of arithmetic operations to
complete an iteration, its convergence rate is the slowest (linear convergence
characteristic).
37
NR method has quadratic convergence characteristic and is the best among all
methods from the standpoint of convergence. Typically, only 3 to 5 iterations are
needed to reach an acceptable solution for a large system. It also has the
lowest sensitivity to the choice of slack bus.
For FDLF, the convergence is geometric and it is more reliable than the formal
NR method due to the fact that the elements of [B ] and [B ] are fixed
Q
approximation to the tangents of the defining functions, P
|V | and |V | , and are
not sensitive to any humps in the defining functions.
Time units
Time units
40
Newton-Raphson
Gauss-Seidel
20
38
Gauss-Seidel
Newton-Raphson
0
0
40
80
Number of buses
(a) Time per iteration
KWCn v1.31
120
40
80
120
Number of buses
(b) Total iteration solution time
19
T HE H ONG KONG
P OLYTECHNIC U NIVERSITY
FAX : (852) 2330 1544
Department of Electrical Engineering
Hung Hom, Kowloon, Hong Kong
EE4031 Power Systems
Tutorial on Power System Load Flow
1. Fig 1 shows a 4-bus system where all the transmission line series impedances are given
to a common base of 100 MVA while the shunt admittances of the lines are neglected.
Specifications at busbars are given in Table 1 and flat start conditions are assumed.
1
S 14
2
S 23
S 21
S 12
j0.2
S 24
j0.5
j0.33
j0.1
j0.25
4
Load
Bus
1
2
3
4
Fig 1
Real
Reactive
Real
Demand Demand Generation
(MW)
(MVAr)
(MW)
100
80
60
90
50
Load
Voltage
Voltage
Magnitude Angle
(pu)
(deg)
1.04
0
1.02
Table 1
(a) Classify the type of each busbar.
(b) Determine the bus admittance matrix.
(c) Determine the initial power flows S12 , S14 , S21 , S23 and S24 .
(d) Determine the initial power generations and mismatches at the bus 1 and 2.
(e) With justification, what should be the real power generation at bus 1 ?
(f) Recommend a solution method, with justifications, which is suitable for solving
this power flow problem.
EE4031, KWCn, 8 Sept 2010
2. Fig 2 shows a single-line diagram of a 2-bus power system with parameters detailed in
Table 2. The series impedance of the line is given in per-unit on a common base of 100
MVA with shunt admittance neglected.
S G1
V1 = 1.02 0
1
V2
Z = j0.5
SL1 = 30 + j10
SL2 = 50
Fig 2
Bus
1
2
Real
Reactive
Real
Demand Demand Generation
(MW)
(MVAr)
(MW)
30
10
50
0
Voltage
Voltage
Magnitude Angle
(pu)
(deg)
1.02
0
Table 2
(a) Name the slack bus and write down the bus admittance matrix Y .
(b) Based on the load flow equation given below :
n
X
1 Pi jQi
Vi =
Yij Vj
Yii
Vi
j=1,j6=i
Use Gauss-Seidel with flat start conditions to solve the load bus voltage V2 .
(c) With justification, what should be the reactive power generation at bus 1 ?
EE4031, KWCn, 8 Sept 2010
3. Fig 3 shows a single-line diagram of a three-bus power system. All the transmission line
series impedances are given in per unit to a common base of 100 MVA while the shunt
admittances are neglected. Specifications at busbars are given in Table 3.
j0.1
200 MW
50 MVAr
j0.1
j0.1
Fig 3
Bus
1
2
3
Real
Demand
PL (MW)
0
200
0
Reactive
Real
Demand
Generation
QL (MVAr) PG (MW)
0
50
0
0
100
Reactive
Voltage
Voltage
Generation Magnitude Angle
QG (MVAr)
V (pu)
(deg)
1.01
0
0
1.02
Table 3
(a) Classify each bus type and determine which of the variables V , , P and Q should
be treated as unknown.
(b) Write down the real power generation at bus 1 by inspecting the data.
(c) Write down the Jacobian matrix in terms of partial derivatives.
(d) Determine the bus admittance matrix.
(e) Given the power flow equations at bus i as follows
Pi =
|Vi |2 Gii +
Qi = |Vi |2 Bii +
n
X
j=1,j6=i
n
X
|Vi ||Vj ||Yij | cos(i j ij )
|Vi ||Vj ||Yij | sin(i j ij )
j=1,j6=i
Derive the general equations for the diagonal coefficients of the Jacobian matrix
and hence find the diagonal coefficients of the Jacobian matrix for the first iteration when the polar form of the Newton Raphson method is used with flat start
conditions.
EE4031, KWCn, 8 Sept 2010
4. A single-line diagram of a three-bus power system is shown in Fig 4.
1
j0.4
200 MW
50 MVAr
j0.2
j0.1
Fig 4
All the transmission line series impedances are given in per unit to a common base of
100 MVA while the shunt admittances are neglected. Specifications at busbars are given
in Table 4.
Bus
1
2
3
Real
Demand
PL (MW)
0
200
0
Reactive
Real
Demand
Generation
QL (MVAr) PG (MW)
0
50
0
0
100
Reactive
Voltage
Voltage
Generation Magnitude Angle
QG (MVAr)
V (pu)
(deg)
1.05
0
0
1.02
Table 4
(a) Classify each bus type and determine which of the variables V , , P and Q should
be treated as unknown.
(b) Write down the bus admittance matrix [Y ].
(c) Using the Fast Decouple Load Flow (FDLF) convention :
P
= B []
|V |
|V |
Q
= B
|V |
|V |
write down the matrices [B ] and [B ].
(d) Given the power flow equations at bus i as follows
Pi =
|Vi | Gii +
Qi = |Vi |2 Bii +
n
X
j=1,j6=i
n
X
|Vi ||Vj ||Yij | cos(i j ij )
|Vi ||Vj ||Yij | sin(i j ij )
j=1,j6=i
carry out the first load flow iteration using the FDLF method.
EE4031, KWCn, 8 Sept 2010
5. Fig 5 shows a single-line diagram of 3-bus system with parameters detailed in Table 5.
The series impedance of each transmission line is given in per-unit on a common base of
100 MVA with shunt admittance neglected.
j0.4
80 MW
-30 MVAr
100 MW
j0.4
j0.4
100 MW
-80 MVAr
3
100 MW
60 MVAr
Fig 5
Bus
1
2
3
Load Demand
MW
MVAr
100
0
100
-80
100
60
Specified Power
MW
MVAr
80
-30
0
0
Specified Voltage
pu
degree
1.0
0
Table 5
(a) Name the slack bus and write down the bus admittance matrix Y .
(b) Based on the load flow equation given below :Vi =
n
X
1 Pi jQi
Yij Vj
Yii
Vi
j=1,j6=i
Perform one iteration of the load flow using the Gauss-Seidel method with flat start
conditions to calculate the appropriated voltages at bus 2 and 3.
(c) What should be the real power generation at bus 1 ?
EE4031, KWCn, 8 Sept 2010
EE4031 Power Systems
Tutorial Solution on Power System Load Flow
1.
a) Bus
1
2
3
4
Type
Slack
Generator
Load
Load
b)
j5
0
j 3.03
j8.03
j5
j 2
j17 j10
Y =
0
j10 j14
j4
j2
j 4 j 9.03
j 3.03
c)
S12 = V1[Yser12 (V1* V2* )]
= 1.04[ j 5(1.04 1.02)]
= j 0.104 pu = j10.4 MVAr
S14 = j 0.126 pu = j12.6 MVAr
S 21 = j 0.102 pu
S 23 = j 0.204 pu
S 24 = j 0.0408 pu
d)
S1 = S12 + S14 = j 23 MVAr
S 2 = S 21 + S23 + S24 = j14.28 MVAr
Bus 1 Slack mismatch = 0 MW
Bus 2 PV bus P mismatch = Pg2 Pl2 Re(S2) = 100 MW
Q mismatch = 0 MW
e) No transmission loss Pg1 = 80 + 90 100 = 70 MW
f)
2.
GS: simple, low memory usage, easy to implement
NR: faster, better convergence, more reliable
FD: even faster & more reliable
a) slack bus : BUS1
j2
j 2
Y =
j 2 j 2
*
1 S2
b) V 2 =
k 1* Y 21V 1 ,
Y 22 V 2
EE4031, KWCn, 8 Oct 2012
= S L 2 = 0.5 ,
22
= j2 ,
21
= j2
Re(V2k )
Im(V2k )
0
1
2
3
4
5
6
7
1.000000
1.020000
0.963331
0.961106
0.957653
0.957496
0.957271
0.957260
0.000000
-0.250000
-0.231209
-0.245381
-0.244199
-0.245117
-0.245039
-0.245099
| V2k |
V2k
1.000000
1.050190
0.990689
0.991936
0.988298
0.988373
0.988136
0.988140
0.0000
-13.7716
-13.4963
-14.3223
-14.3054
-14.3592
-14.3581
-14.3616
Solution with accuracy of 2 decimal points (4 iterations) will be adequate.
c)
3.
10 MVAr
a) Bus
1
2
3
Type
Known
Unknown
Slack
Load
Generator
V,
P, Q
P, V
P, Q
V,
Q,
b) No transmission loss Pg1 = 200 100 = 100 MW
c)
P2
2
P
3
2
Q
2
2
P2
3
P3
3
Q2
3
P2
V2
P3
V2
V2
Q2
V2
V2
V2
d)
j10
j 20 j10
j10 j 20 j10
j10
j10 j 20
e)
P2
= (1.01)(10) sin 90 + (1.02)(10) sin 90 = 20.3
2
P3
= (1.02)(1.01)(10) sin 90 + (1.02)(1)(10) sin 90 = 20.502
3
V2
Q2
= 20.3 2(1) 2 (20) = 19.7
V2
EE4031, KWCn, 15 Mar 2011
4.
a) Bus
1
2
3
b)
Type
Slack
PQ
PV
j 2.5
j5
j 7.5
[Y ] = j 2.5 j12.5 j10 = [G + jB ]
j 5
j15
j10
c)
P2
V B
2 = 22
P3 B32
V
3
B23 2
=
B33 3
12.5 10
B ' =
10 15
d)
Q2
V2
= [ B 22] =
V2
V2
[ B "] = [12.5]
Flat start conditions all angles are zero
i.e. no angle difference P2 = 0, P3 = 0
200
P2 = P2schedule P2 =
0 = 2 pu
100
100
P3 = P3schedule P2 =
0 = 1 pu
100
2
1.0 12.5 10 2
1 10 15 3
1.02
1 15 10 2 0.2308
2 =
=
3 87.5 10 12.5 0.98 0.0886
0.2308
2 =
rad
3 0.0886
Q2 = V2 B22 + V2 V1 Y21 sin( 2 1 21 ) + V2 V3 Y23 sin( 2 3 23 )
= 12.5 + (1.0)(1.05)(2.5) sin 0.2308 + (1.0 )(1.02 )(10 ) sin 0.1422
2
2
= 0.152 pu
Q2 =
50
(0.152) = 0.348 pu
100
Q2
V
= [12.5]
V2
V2
V2 =
0.348
= 0.028 pu
12.5
i.e. V2 = 1 + V2 = 0.972 pu
EE4031, KWCn, 15 Mar 2011
5.
a) Slack bus - Bus 1
Z L = j 0.4,
b)
YL = j 2.5,
j5
Y = j 2.5
j 2.5
j 2.5
j5
j 2.5
j 2.5
j 2.5
j5
Flat start: V10 = 10, V20 = V30 = 10
S2 = 1 + j 0.8 + 0.8 j 0.3 = 0.2 + j 0.5 pu
S3 = 1 j 0.6 pu
c)
V21 =
1 0.2 j 0.5
0.2 + j5.5
j 2.5 j 2.5 =
= 1.1 j 0.04 = 1.1 2.08 pu
1
j5
j5
V31 =
1 1 + j 0.6
1 + j 4.4
j 2.5 j 2.5 =
= 0.88 j 0.2 = 0.9 12.8 pu
1
j5
j5
100 + 100 + 100 - 80 = 220 MW
EE4031, KWCn, 15 Mar 2011