Você está na página 1de 12

Advanced Power System Analysis

Load Flow Analysis


Hadi Suyono, ST., MT., Ph.D
1 | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University


Chapter 3: Load Flow Analysis


3.1 Introduction

The determination of the distribution of power and voltage in an integrated
power supply system consisting of generators, transmission and distribution
circuits and loads is an essential requirement in the design and operation of such
systems. It enables the loading and utilization of circuits to be assessed for actual
or predicted generation and load conditions and 'provides a knowledge of the
associated voltage levels throughout the system. The calculations required to
obtain this information are known as LOAD FLOW (or POWER FLOW studies)
and essentially provide a steady state analysis of the system. These studies are of
the utmost importance and frequently precede, in most cases provide the
starting conditions for, transient stability and fault level calculations. Such
studies are nowadays carried out almost exclusively by digital computers and,
the equations defining the problem are solved by special numerical techniques,
which have been developed to suit the special structure of the problem.

The need for load flow analysis of a power system may be envisaged as occurring
under three conditions. Firstly, during the early design stages of a power system
network, the routing of circuits and the siting of generating and switching
stations may be subject to uncertainties and as a consequence a number of
alternatives may require evaluation. Because of the, approximate nature of the
available data at this stage a d.c. or real number representation of the a.c. power
system may be sufficient and advantage of this sufficient is often taken.

Secondly, at later design stages, when firm data is available and for
reinforcement and operational studies, different operating modes of the system
require evaluation and assessment. In these circumstances reactive power
distribution is of importance and a more detailed and complex number
representation of the system is essential.

Load-flow calculation is one of the most important solutions required in the
analysis of power-system transmission networks and is used extensively in
system planning, design and operation. The object is to determine the nodal
voltages and hence the branch power flow in a transmission network for a given
set of steady-state operating conditions. Such routine solutions are nowadays
carried out almost exclusively by digital computers and the equations defining
the load-flow problem are solved by special numerical techniques, which have
been developed to suit the special structure of the problem.

Load-flow calculations are being increasingly used to solve very large systems,
with automatic control features for off- and on-line application. They are
extended for calculation of outage security assessment and with more
complicated calculations such as optimisation and stability, the basic load-flow
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
2 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

problem can be considered with the transmission network taken as linear,
balanced and represented as a lumped-parameter circuit. The settings of
network regulating devices, such as transformer tapping, can be omitted or
assumed to be fixed. However, the overall problem even with the above
assumptions is a non-linear one, because of the conditions specified at the
busbars, and therefore any numerical solution must be iterative in nature. There
are many alternative load-flow solution methods and these can be broadly
grouped into three categories:

Y-matrix iterative methods. There were the first practical methods applied to
automatic digital solution. Although they perform satisfactorily on many
problems, requiring minimal computer storage, they converge relatively slowly
and are sensitive to system parameters and conditions.

Z-matrix iterative methods. These methods are more reliable than the
Y-matrix methods but require very much more computer storage. Large systems
can also take considerable computer time for inversion and solution of resulting
matrices.

Newton-Raphson Method. This is the latest and most powerful method
generally accepted by much of industry as the most reliable method. Using
sparsity, it combines low storage with quadratic convergence characteristics of
the Newton method. It is easy to program and has wide application as a general-
purpose method which is easily extended for optimisation and on-line
application.


3.2 Basic Formulation

3.2.1 Main Symbols


B', B" approximate Jacobian matrix V VZu = e + jf = nodal voltage
E voltage behind synchronous
reactance
Y G + jB = nodal admittance matrix
G p.u. change in generation Z nodal impedance matrix
f frequency u voltage angle
H, J, L, N Jacobian submatrices of partial
derivative
N number of nodes excluding
reference
I injected net nodal current Re real part
S P + jQ = net complex power Im imaginary part
1 : T = 1 : (t+jq) = transformer ratio in pu


Subscripts Superscripts
E power exchange sp specified
d demand (load) r iteration number
g generation T transposition
i,j,k nodes min minimum
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
3 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

s synchronous values max maximum
t tap values * complex conjugate
(-) bar over a main symbol indicates a complex number
k e i subset of nodes directly connected by branches to node - i
k i subset of nodes directly connected by branches to a PQ-node i
k o i subset of nodes directly connected by branches to a PV-node i

For numerical solution it is most convenient to assume balanced three-phase
power system with the transmission network represented by its positive
phase-sequence impedances of linear lumped series and shunt branches. Using
matrix notation, the nodal voltage equation can be stated as
Y V = I (2.1)

where I represents the equivalent nodal injected currents and is calculated from
*
*
/
i
s
i
i
V S I =
(2.2)

where S
i
*
net generation at busbar i
* s
i S = ( Pgi - Pdi) -j(Qgi - Qdi) =
s
i
s
i
jQ P
(2.3)


The load-flow equations can be stated in any number of different ways. For
example, in matrix form as equations (2.1) and (2.2). Equation (2.1) can also be
expressed in a summation form for node i as
=
e k i
ik k i Y V I (2.4)

Substituting from equation (2.2) for Ii results in
=
e k i
ik k i i Y V S V
* *
/
or as
V Y V S i
k i
ik k i
* *
=
e
(2.5)

The same problem can be stated using the polar form as

k k
ik
i k
i i
i V Y V S u u Z Z =
e
*

or as
i k k
ik
i k
i
i V Y V S u u Z =
e
*
(2-6)

Equating real and imaginary parts:

k ik ik ik ik
i k
i
sp
i
V B G V P ) sin cos ( u u + =
e
(2.7)

k ik ik ik ik
i k
i ii i
V B G V G V ) sin cos (
2
u u + + + =
=


k ik ik ik ik
i k
i ii i
V B G V G V ) sin cos (
2
u u + + + =
=

k ik ik ik ik
i k
i
sp
i
V B G V Q ) cos sin ( u u
c
=


k ik ik ik ik
i k
i ii i
V B G V B V ) cos sin (
2
u u + =
=
(2.8)
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
4 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University


where Yik = Gik + jBik is the (i,j)
th
element of nodal admittance matrix, uik = ui -
uk
Note that cos uik = cos uki and -sin uik = sin uki

3.2.2 Types of Nodes

With each node four variables are associated, Pi, Qi, Vi i of which two are
specified and the other two must be found. Depending upon which variables are
specified, the nodes can be divided into 3 basic types:

Node Specified variable Unknown variable
PQ
sp
i
sp
i
Q P ,

i i
V u ,

PV
sp
i
sp
i
V P ,

i i
Q u ,

Slack
sp
i
sp
i
V , 0 = u

i i
Q P,


For a PQ-node

-
= + = i i
sp
di
sp
gi
sp
di
sp
gi
sp
i
I V Q Q j P P S ) ( ) (
(2.9)
For a PV-node
P P P V I
i
sp
gi
sp
di
sp
i i = =
-
( ) Re( ) (2.10)

2 2
i i
sp
i
f e V + =
(2.11)

Besides the above equations, load flow also can be constrained by power
balance, i.e. input power system must be equal to the output power. In other
word, total generation system must be equal to the total load system and circuit
losses.


) , (

+ = + =

u v f P P P P
p
nb
i
di loss
nb
i
di
nb
i
gi
(2.12)

) , (

+ = + =

u v f Q Q Q Q
q
nb
i
di loss
nb
i
di
nb
i
gi
(2.13)

A circuit loss as a result of current flowing on the line is I
2
R. The current can be
found from node voltage when the load flow has converged. Therefore we must
reduce the constraint by choosing one generator bus, because the generating
bus can be changed so that fit power balancing constraint. In another word, P
can choose the reference bus as slack bus because angle phase of the reference
bus is equal to zero. We need to choose the magnitude voltage for slack bus and
our choices influence to solution.
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
5 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University


3.2.3 Mismatch

Since the numerical solution of the load-flow problem is iterative in nature it is
necessary to have some criteria for convergence. It is usual to use the power
mismatch as an accurate test. The complex power mismatch at a node i is the
difference between the specified power Si
sp
and the calculated power V I i
r
i
*

where V i
r
is the voltage evaluated at iteration r
) (
* * *
k ki
i k
r
i
sp
i
sp
i
i
r
i
sp
i i V Y V jQ P I V S S
e
+ = = A
(2.14)

Equation (2.14) can be separated into real and imaginary parts and expressed in
polar form or rectangular form as:
k ik ik ik ik
i k
i
sp
i
sp
i
V B G V P P ) sin cos ( u u + = A
e
(2.15)
k ik ik ik ik
i k
i
sp
i
sp
i
V B G V Q Q ) cos sin ( u u = A
e
(2.16)
) ( ) (
k ik k ik
i k
i k ik k ik
i k
i
sp
i
sp
i
e B f G f f B e G e P P + + = A
e e
(2.17)
) ( ) (
k ik k ik
i k
k k ik k ik
i k
i
sp
i
sp
i
e V f G e f B e G f Q Q + + = A
e e
(2.18)

It is assumed in the above equations that Vi is the voltage evaluated at iteration
r. The most common convergence criterion used in practice is
APi s c for all PQ and PV-nodes
AQi s c for all PQ-nodes
where c is a tolerance typically of 0.01 MW and MVAR.
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
6 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

3.3 Newton-Raphson Method

An outstanding technique of determining real roots of polynomials,
trigonometric and other equations is known as the Newton or Newton-Raphson
method. If an approximate root x
r
of a simple algebraic non-linear equation is
known, then a better approximation can be obtained from
( ) ( )
r
x
dx
x d
n
x
x f
x
x xf x f x x f
n
n n
r r r r
!
...... ) ( ' '
! 2
) ( ' ) ( ) (
2
1
A
+ +
A
+ A + = A +
+

If an approximation function on the left side is equal zero and all series after the
second order are neglected, we can rewrite the above equation as
) ( ' ) ( 0
r r
x xf x f A + =


the above equation can be rearranged as
) ( ' ) (
r r
x f x f x = A
(3.1)

Hence the Newton-Raphson approximation is defined as

) ( '
) (
1
r
r
r r
x f
x f
x x =
+
(3.2)
or
x x x
r r
A + =
+1
(3.3)

The method can be extended to a set of simultaneous non-linear equations in an
equal number of unknowns of the form
f1 (x1,x2, - - - - - - ,xn) = 0
f2 (x1,x2, - - - - - - ,xn) = 0
. . (3.4)
. .
fn (x1,x2, - - - - - - ,xn) = 0

If a set x
r
of approximate values of the variables (x1,x2,...,xn) is known, then a
better approximation can be obtained from
( ) + A + A + =
r
r
r
r r r
x
f
x
x
f
x x x f
2
1
2
1
1
1 2 1 1
, 0
c
c
c
c

( ) + A + A + =
r
r
r
r r r
x
f
x
x
f
x x x f
2
2
1
2
1 2 1 2
2
, 0
c
c
c
c (3.5)
If an approximation function on the left side is equal zero and all series after
second order are neglected, the above equations can be rearranged as
( )
( )
) , , (x
1
1
1
1
1
1
1 1
1

, ,
, ,
r
n
r
x
r
n
r
n
n n
n
r
n
r
n
r
n
r
x
x
x
f
x
f
x
f
x
f
x x f
x x f

(
(
(

A
A
(
(
(
(
(

=
(
(
(

c
c
c
c
c
c
c
c
(3.6)
or
( )
( )
( )
( )
( )
J =
1
r
1
(
(
(

A
A
(
(
(

r
n
r
r
n
r
x
x
f
f

(3.7)
where
) (
1 r
x f J x

= A (3.8)
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
7 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

The square matrix J is the Jacobian matrix of f(x), whose (i,k)-th element is
defined as cfi/cxk.
To use the above algorithm to solve load-flow, it is necessary to write the
equations defining the load flow problem as a set F(x) = O equations. The most
convenient and most widely used form is the polar power-mismatch
formulation, equations (2.17) and (2.18). The equations become as follows:

For a PQ-node
0 ) sin cos ( = + = A
e
k ik ik ik ik
i k
i
sp
i i
V B G V P P u u (3.9a)
0 ) cos sin ( = = A
e
k ik ik ik ik
i k
i
sp
i i
V B G V Q Q u u
(3.9b)

For a PV-node only
AP P V G B V
i i
sp
i
k i
ik ik ik ik k
= + = =
c
u u ( cos sin ) 0 (3.10)

For a slack-node

0 ) , ( = +

=

u v f P P P
p
nb
i
di
nb
r i
gi gr


0 ) , ( = +

=

u v f Q Q Q
q
nb
i
di
nb
r i
gi gr
(3.11)

The general Newton-Raphson equation (3.8) can be written in terms of the
load-flow variables as

| |
(

A
A
- =
(

A
A
V
J
Q
P u
or as
(

A
A
-
(


=
(

A
A
Q
P
J
V
1
u
(3.12)

where Aui and are AVi the corrections for angle and voltage magnitude at node I.
The Jacobian matrix of partial derivatives can be divided into 4 parts as

(
(

A
A
(

' '
=
(

A
A
V
V
L J
N H
Q
P
u
(3.13a)
or

(

A
A
(
(
(

c
A c
c
A c
c
A c
c
A c
=
(

A
A
V
Q Q
V
P P
Q
P u
u u
u
(3.13b)

The elements of the Jacobian matrix can be obtained by differentiating equations
(3.9) and (3.10) and noting that
ccos uik /cu = - sin uik csin uik/cui = cos uik
ccos uik /cuk = sin uik csin uik/cuk = - cos uik

gives equations (3.14) as follows:
ii i
sp
i k ik ik ik ik
i k i k
i i i ii
k ik ik ik ik i k i ik
ii i
sp
i k ik ik ik ik
i k i k
i i i ii
G V P V B G V Q J
V B G V P H
B V Q V B G V P H
2
;
2
;
) sin cos ( /
) cos sin ( /
) cos sin ( /
+ = + = A =
= A =
+ = = A =
= e
= e
u u cu c
u u cu c
u u cu c

Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
8 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

k ik ik ik ik i k i ik
V B G V Q J ) sin cos ( / u u cu c + = A =
ik k i k ik
ii i
sp
i i i i ii
J V V N
G V P V P V N
= A =
= A =
c c
c c
/
/
2
(3.14)
ii i
sp
i i i i ii
B V Q V Q V L
2
/ + = A = c c
ik k i k ik
H V Q V L = = c c /


Iter = Iter + 1
Start with Initial
V
0
and u
0
Iter = 0
i = i + 1
Bus i = ??? Calculate A Q, Equation (3.9)
Calculate A P, Equation (3.10)
AP:c , AQ:c
OUTPUT
Iter = max
Form Jacobian Matrix, EQ (3.11)
Solve Eq (3.11) for Au and AV/V
Update u and V
PV
PQ
Yes
No

Flow diagram for a load-flow solution by Newton-Raphson method.

The iterative process of the Newton-Raphson method, as applied to the
load-flow solution, can be summarised as shown in the flow diagram.
The main advantage of the Newton-Raphson method is its quadratic rate of
convergence, which is faster than that of any other method. It is very reliable
and least sensitive to factors causing poor convergence, such as choice of slack
node or series capacitors. Both polar or rectangular formulation can be used, but
in order to use the Newton method the equations must be separated into real
and imaginary set. Generally a satisfactorily accurate solution may be obtained
in 5 to 6 iterations.

The main disadvantage of the method is the necessity to formulate and invert
the Jacobian matrix at each iteration. However, since the structure of the
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
9 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

Jacobian matrix has identical sparsity content as the Y-matrix, sparsity
techniques can be used instead of inversions. This has the advantage of
requiring only one ordering sequence and since the actual elimination and
solution is relatively less time consuming, the method presents no problem in
solution time or storage space. Another disadvantage of the Newton method is
that it is applicable only to convex functions and will converge to the nearest
solution from the initial value. It is a well-known fact that the load-flow problem
has theoretically twice as many possible solutions as there are number of nodes.
In practice the correct engineering solution is well separated from the others
and the right solution is always obtained or the solution diverges.

3.4 Fast Decoupled Newton-Raphson Method

Since the Newton method is an approximate method, alternative approximations
are possible to even further decrease the time and storage.
The basic equations of the Newton method are derived from Taylor's series
expansion and omitting second and higher order terms. The corrections at each
iteration are therefore approximations but the value of the function is calculated
exactly at each iteration, therefore the final solution can be obtained to any
desired accuracy by continuing the iterative process and is not dependent on the
accuracy of the correction itself.

This simple fact implies alternative approximation to the derivative terms of the
Jacobian matrix. A simple test of solving, say a quadratic equation by the Newton
method, can soon show that a good convergence can still be obtained, even is the
derivative is substituted by an arbitrary constant number. Of course the number
must be so chosen that the resulting equations retain the convergence property.

Since the load-flow problem is only mildly non-linear and well understood by
engineers, the physical property of the problem can be exploited to great
advantage.

3.4.1 Decoupling Characteristics

An inherent characteristic of any practical power system is the close dependence
of active power flows on bus voltage angle and the reactive power flow on bus
voltage magnitude. This characteristic may be shown from the following
consideration of a two-node system.
V
1
I
V
2
P+jQ

(a) Two node system
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
10 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

Au ~ 0
Au
AV ~ 0
+ jx P P ( ) A
V A
-jxP
-XQ
) ( Q Q X A +
1
V
V
2


(b) Vector diagram
Figure 4.1 : Q-V and P-u relation

Assuming V1 as reference (1 p.u. and u1 = 0) and Z = jx , since R << x , then
Z I V V = 1
2

But
jQ P
V
jQ P
I =

=
*
) (
1

therefore
jxP xQ
V
jQ P j
V V =

= 1
*
) (
1
1
2


It can be seen from the phasor diagram that
(a) Change of AP in P changes the angle by Au and has a small effect on change in
voltage magnitude (AV ~ O)
(b) Change of AQ in Q changes the voltage magnitude by QV and has a small
effect on change in angle (Au ~ O)

Algebraically the above approximate relation may be stated as:
u A = A H P (4.1)
V L Q A = A (4.2)

Generalising the decoupling phenomena to a multi-busbar system, the sub-
matrices N and J in equation (5.11) can be neglected resulting in

(

A
A
(

=
(

A
A
V L
H
Q
P u
(4.3)

The obvious procedure is to perform alternate iterations of the two equations,
that is solving separately the P-u and Q-V problems with 4:1 saving on storage of
the Jacobian matrix. Such an approximation results in more iterations, but in
practice is more than compensated for by saving in inversion time and slightly
shorter individual iteration time.

3.4.2 Further Approximations

Based on similar assumptions as before, further approximations in the
decoupled Jacobian matrix (4.3) can be made:
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
11 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

ik ik ik ik
i ii i
ik
B G
V B
u u
u
u
cos sin
1 cos
2
((
((
~
(4.4)
from which the elements of the Jacobian matrix of equation (3.14) reduces to
k ik i ik k ik i k i ik
i ii i ii i ii i i i ii
V B V L V B V P H
V B V L V B V P H
= = A =
= = A =
cu c
cu c
/
/
(4.5)

or as
k ik
i k i k
i ii i i
k k ik
i k i k
i ii i i i
V B V B V Q
V B B V V P
A + A = A
A + A = A
= e
= e
,
,
/
/ u u
(4.6)
The first equation relates AP to Au, therefore any terms which predominately
affect changes in AQ can also be omitted. For example, by setting the voltages on
the right-hand side to unity results in
V B V Q
B V P
A = A
A = A
, ,
,
/
/ u
(4.7)

The matrix B' and B" is the negative of the imaginary part of the nodal
admittance matrix Y, without the column and row corresponding to the slack
node. Faster convergence can be obtained by neglecting the resistance when
calculating the terms of B', hence
ik
i k
ii ik
i k
ii
ik ik
ik
ik ik ik
B B x B
X R
x
B x B
e e
= =
+
= =
' ' '
2 2
' ' '
/ 1
/ 1
(4.8)

Further improvements in convergence can be obtained by making the following
simplification
(a) Omit from B' elements, which affect changes in AQ, for example shunt
reactors, shunt capacitors and off-nominal in-phase transformer taps.
(b) Omit from B" phase-shifters which affect changes in AP

3.4.3 Iterative Scheme

With the assumptions made, the B' and B" matrices result in being real,
symmetrical, sparse and constant since they contain only network admittances.
They represent constant approximation to the variable Jacobian matrix and
therefore need to be inverted, or factorised if sparsity techniques are used, once
only for the whole solution process.
There are many ways of further approximating the B' and B" matrices and
solving the resulting equations. From practical experience the best way is as
summarised in the flow diagram.

Although approximate Jacobian equation has been used, the mismatch equations
are exact. Therefore the solution obtained is as accurate as that given by the
Advanced Power System Analysis
Load Flow Analysis
Hadi Suyono, ST., MT., Ph.D
12 Chapter 3: Load Flow Analysis | Electric Power Engineering
Department of Electrical Engineering, Brawijaya University

Newton-Raphson method. The only difference is that in the decoupled load-flow
the gradient is constant.

The constant slope has a geometric convergence rate which is slower than the
quadratic rate of the Newton-Raphson method. Typically convergence can be
obtained in 6 to 8 iterations. The slower convergence rate is more than
compensated for by the speed of calculations. The solution time is
approximately linear with the system size.

Start with initial
Factorise X and B using Sparsity
Iter = Iter + 1
Loop i from bus i=1 to Nb
BUS ????
Calculate
Solve for
Update
Bus ???
Update
Solve to
Calculate
Output Result
Calculate Q for node PV Eq (2.8)
P and Q for slack (2.7) and (2.8)
V
r
= = 1 0 dan
r
u
(3.9) Eq. from V / P
r r
A
(4.6) Eq from
r
u A
u u u
r r r +
= +
1
A
(3.10) Eq from /
r r
V Q A
(4.6) eq from
r
V A
V V V
r r +
= +
1
A
AP : c
AQ
r
: c
converge
Slack bus
Not converged
PV
PV or PQ


Flow diagram for a load-flow solution by fast decoupled method.

Você também pode gostar