Professional Documents
Culture Documents
EE 307 - Topic#3-Load Flow Studies
EE 307 - Topic#3-Load Flow Studies
TOPIC OUTLINE
3.1. Meaning and Objectives of Load-Flow Study
3.2. System Specification for Load-Flow Analysis
3.3. Mathematical Formulation of Load-Flow Problem
3.4. Numerical Solutions of the Load-Flow Problem
3.4.1. Gauss-Seidel (G-S) load-flow solution method
3.4.2. Newton-Raphson (N-R) load-flow solution method
3.4.3. Pθ-QV Decoupling method
3.4.4. DC Load-flow model
3.5. Methods of Load-Flow Control/Regulation
The input data for each transmission line includes the per-unit equivalent π circuit series impedance and
shunt admittance, the two buses to which the line is connected and maximum MVA ratings. Similarly,
input data for each transformer include per-unit winding impedances, the per-unit exciting branch
admittance, the buses to which the windings are connected, and maximum MVA ratings.
For the purposes of load-flow analysis and depending on the system performance variable specified
apriori, power system buses can be categorised into the following types:
• swing (slack, or floating) bus;
• P-Q bus; and
• P-V bus.
Table 3.1 presents a summary of the characteristics or parameters normally specified for these principal
bus types.
Table 3.1. Summary of principal bus types
Bus Bar Electrical Quantities
Voltage Generation Demanded Power Shunt Capacitor
Bus Type Power Power
V δ Pg Qg Pd Qd Qc
Load (P-Q) Bus x x n.a. n.a. √ √ √
Generator (P-V) Bus √ x √ x n.a. n.a. n.a.
Slack Bus √ √ x x n.a. n.a. n.a.
The detailed nature of the above types of buses is outline below.
(a) Swing (aka slack or reference) bus: This is typically as a generator bus and is only one in a
power system. For this type of bus, we know the voltage magnitude and phase angle. Physically,
there is nothing special about the swing bus; in fact, it is a mathematical artefact of the solution
procedure. The generation must supply both the load and the losses in the network. Before solving
the load flow problem, we will know all injections at P-Q buses, but we will not know what losses
will be as losses are function of the flows which are yet to be computed. So we may set the real
power injections for, at most, all but one of the generators. The one generator for which we do not
set the real power injection is the one modelled as the swing bus. Thus, this generator “swings” to
compensate for the network losses (i.e., compensate for the real power mismatches of the entire
system), or, one may say that it “takes up the slack.” The voltage magnitude of the swing bus is
chosen to correspond to the typical voltage setting of this generator, while the voltage angle may
be designated to be any angle, but normally it is set equal to zero.
(b) P-V Buses: For this type of buses, we know the net real power injection (i.e., Pg – Pd) and the
voltage magnitude. These buses fall under the category of voltage-controlled buses because of the
ability to specify the voltage magnitude of this bus. Most generator buses (i.e., buses at which
generators are connected) fall into this category, independent of whether it also has load;
exceptions are buses that have reactive power injection at either the generator’s upper limit or its
lower limit. P-V buses where generators operate at their limiting reactive power values are treated
as P-Q buses.
(c) P-Q Buses: For this type of buses, we know the net real and reactive power injections, but not the
voltage magnitude or angle. All load buses fall under this category, including buses that have not
either load or generation (i.e., null-injection buses).
Once the power system whose steady-state performance analysis is to be conducted is specified as
detailed above, the next step needed is formulation of a mathematical model that describe the
G1 G2
Sg1=Pg1+jQg1 Sg2=Pg2+jQg2
2
1 V1ejδ1 V2ejδ2
Transm. Line
Sd1=Pd1+jQd1 Sd2=Pd2+jQd2
I1 1 xs rs 2 I2
zs=rs+jxs
yp S2
S1 yp
or
where V1 = |V1|∠θ1 and I12 = current flow in the transmission line or the net current injected at busbar
“1” from some source. The expression for the line or injected current can be obtained by application of
the bus admittance equation to the appropriate bus. For our case, the application of this equation at bus
“1” gives the following expression
2
I12 = I1 = ∑ Y1iVi = Y11V1 + Y12V2 (3.3)
i =1
S1* P1 − jQ1
= = Y11V1 + Y12V2 (3.4)
V1* V1*
or
By equating the real and the imaginary parts in the equation (3.4) or (3.5), we obtain
The power flow expression given by any of the equations (3.5 – 3.7) constitutes the mathematical
model/statement/formulation of the load-flow problem for the above single-circuit two-bus system.
Using the same procedure outlined above, we can obtain similar models for a generalised bus “k” of an
“N” bus system. That is,
PK − jQK N N
VK*
= ∑
i =1
YKiVi or PK − jQK = VK* ∑ YKiVi
i =1
(3.8)
Consequently,
* N
Pk = Re Vk ∑ YkiVi (3.9)
i =1
N
Qk = − ImVk* ∑ YkiVi (3.10)
i =1
If we denote Vk = Vk e jδ k , Vi = Vi e jδ i , and Yk ,i = Yk ,i e jδ ki , then:
N
Pk = V k ∑Y
i =1
ki Vi cos(θ ki + δ i − δ k )
N (3.11)
Qk = − Vk ∑ Yki Vi sin(θ ki + δ i − δ k )
i =1
The equations [3.8], [3.9], [3.10] and [3.11] are the various versions of the mathematical formulations
(models) of a load-flow problem. They are often referred to as the static load-flow equation (SLFE).
These equations are non-linear (because of the non-linear relation between current and voltage
variables) and therefore explicit or direct solution by inversion of the bus admittance matrix is not
possible. Solution can only be obtained by iterative numerical techniques. Some of the commonly used
numerical solution methods are discussed in the next section.
Before going to the solution of the formulated power flow problem, it is important to point out some of
the most important characteristics of these equations:
f i ( x1 , x2 , x3 ,...., xn ) = 0 (3.12)
xi = ϕ i ( x1 , x2 , x3 ,........., xn ) (3.13)
(
xi( m +1) = ϕ i x1m +1 , x2m +1 , x3m +1 ,....., xim−1+1 , xim , xim+1 ,....., xnm ) (3.14)
where the superscripts (m+1) and (m) indicates the new (improved) and initial (old) iteration.
Calculation using Equation (3.14) is continued until two consecutive iterations of xi are within some
predefined tolerance, ε, of each other. That is,
Once the test in Equation (3.15) is satisfied, the iterative process is stopped. Note that the above
criterion for acceptance of solution can also be written as
1 Si* N
Vi = * − ∑ yikVk , for k ≠ i (3.18)
yii Vi k =1
Equation (3.18) can now be solved using the method of successive approximation. Thus,
1 Si* i −1 N
*( m ) ∑ ik k ∑
m +1 m +1
Vi = − y V − yikVkm (3.19)
yii Vi k =1 k = i +1
The superscript (m) denotes the iteration number “m” and superscript (m+1) denotes the subsequent
iteration. The equation for Vk for k < i would have been solved prior to bus i. Hence, the voltage for
Vi ( m +1) − Vi ( m ) ≤ ε (3.20)
That is, the iteration process is continued until the difference between two consecutive estimates of the
voltage becomes less than or equal to the indicated (predetermined) tolerance or error of accuracy.
Convergence upon an erroneous solution may occur if the initial voltages are widely different from the
correct value. Such problems can be avoided if the initial estimates are of reasonable magnitude and do
not differ too widely in phase. Since the nominal operating voltages of a power system are usually
chosen as per unit bases, a good estimate of the magnitudes of the bus voltages is 1.0 p.u. (or nominal
rated voltage). The voltage phase angles throughout a system usually vary by a few degrees. By setting
the floating busbar voltage angle to zero, the other busbar voltage angles will probably be near zero also.
Therefore, zero degrees is a natural first approximation for all voltage angles.
Equation [3.19] refers to a busbar with P and Q specified. At a generator bus, however, voltage
magnitude and real power are specified with perhaps upper and lower limits to the reactive power.
The magnitude of the busbar voltage is fixed, but its phase depends on the reactive power at the busbar.
The values of complex voltage and reactive power from the previous iteration are not related by [3.19]
because complex voltage has been modified to give a constant value of magnitude. It is necessary at the
next iteration to calculate the value of reactive power corresponding to the complex voltage from
equation [3.19]. This value of reactive power holds for the existing value of the complex voltage and is
then substituted into equation [3.19] to obtain the new or corrected value of the voltage. Thus, for a PV
bus,
Pi − jQi( m +1) i −1 N
∑ ∑
( m +1) 1 ( m +1)
Vi = *( m )
− y ik V k − y ik V k
( m)
(3.21)
y ii Vi k =1 k = i +1
The changing value of the reactive power appearing in the above equation is determined using the
equation (3.10). That is:
* N
Qi = − Im Vi ∑YikVk
k =1
"
For evaluating the value of this reactive power, the best estimate obtained so far for the voltage and
should be used.
The real and imaginary components of the new complex voltage are then multiplied by the ratio of the
fixed (regulated) voltage and the magnitude of the new voltage, thus complying with the constant
voltage magnitude constraint. The phase of the voltage is thus obtained and the iteration can proceed to
the next busbar.
If at any iteration the required reactive power to maintain the voltage at the PV bus constant is outside
the specified limits, equation (3.19) is introduced for that bus, using the limit as the specified reactive
power value. That is, that bus is converted into a PQ (load) bus.
For a swing busbar, both the voltage magnitude and phase are specified. Therefore, the equation
corresponding to this busbar can be eliminated from the bus equations. So, for an N busbar system, only
The selection of the acceleration factor is based on conjecture and is often governed by experience. A
scalar acceleration factor in the range 0 < ≤ 2 will improve the convergence but cannot change a
divergent problem into one that converges. When α = 1, this becomes the basic G-S iteration. An 0<α<1
is called under-relaxation and 1<α<2 is over-relaxation. The acceleration factor chosen depends upon
the system and its value normally lies within 1.4 to 1.6.
After the power injected at the swing bus is calculated and voltage magnitudes and phases are known at
every bus in the system, the line power flows and line losses can then be easily be calculated. For
instance, the complex power flow from any bus k to m on the element connecting these two buses is:
The algebraic sum of the two equations for the line flows is the line losses in the element
interconnecting buses k and m.
Note that if we assume the π representation of transmission line, the above line flows will be expressed
as:
#&,
∗
= − = ∗
= ∗
! − . #$, + . '
2
and
#&,
∗
= − = ∗
= ∗
! − . #$, + . '
2
(c) The general algorithm for load-flow solution using G-S method
(i) Define the bus types and show the variables specified apriori and the unknown variables.
(ii) Assemble the bus admittance matrix using either the rule of self and mutual admittances, or using
the singular transformation technique.
(iii) Carry out iterative computation of unknown bus voltages (both in magnitude & phase angle)
according to the following sequence of steps:
Assume initial values of voltages for load buses and phase angles (except swing bus):
Example #1. Perform one iteration of a G-S load flow solution of the system shown below. Buses #2
and #3 are load buses with per-unit loads of 0.25 + 0.12 *+ and 0.3 + 0.15 *+, respectively. Bus #1
is the swing bus with = 1.04. /0 *+. Perform the bus calculations in numerical order. The figures
shown on the diagram are per-unit impedances.
1 0.05+j0.3 2 3
G
0.1+j0.5
0.05+j0.3
-j12.5
-j25 -j12.5
-j25
Example #2. Perform one iteration of a G-S load flow solution of the system shown below. Buses #3
and #4 are load buses with per unit loads of 0.23 + 0.19 and 0.42 + 0.13, respectively. Bus #1 is the
swing bus with = 1.03. /0 *+. Bus #2 is a voltage controlled bus with | 3 | = 1.02 *+ and a
generated power of 0.24 pu. Perform the bus voltage calculations in numerical order. The figures shown
on the diagram are per-unit impedances.
1 4 j0.5 2
j0.2
j0.5
G G
3
j0.4 j0.4
Example #3: Using the G-S method, calculate the complex bus voltages after the first iteration for the
three-bus network shown below. Busbar #2 is a load bus consuming 200 MW, 120 MVAr, busbar #3 is
a generator bus set at 70 MW and 228 kV. Busbar #1 is a swing bus with a voltage magnitude of 230
kV. The three interconnecting lines have equivalent 4-circuit parameters as:
G G
1 3
Line 1
Line 2 Line 3
The value of the function at any point can then be expressed in terms of its variables and the nominal
values as:
∂y
(xi − x0,i )
N
y = y0 + ∑ (3.25)
i =1 ∂ x0,i
For an equation written in the general form y = C , where C is a constant, linearisation by the Taylor
series gives the equation
∂y
(xi − x0,i )
N
C = y0 + ∑ (3.26)
i =1 ∂x0,i
By the method of successive approximation, any value for the function parameter can be obtained from
its nominal value and the incremental value. That is,
∂y1 m ∂y1 m ∂y
∆xi + m ∆x2 + ...... + m1 ∆xnm = C1 − y1m
∂x1
m
∂x2 ∂xn
∂y2 m ∂y2 m ∂y
∆x1 + m ∆x2 + ...... + m2 ∆xnm = C2 − y2m
∂x1
m
∂x2 ∂xn
(3.28)
....................................................................
∂yn m ∂yn m ∂y
∆x1 + m ∆x2 + ....... + mn ∆xnm = Cn − ynm
∂x1 m
∂x2 ∂xn
The above system of linear equations can be written in matrix form, in which the main matrix will be
that whose elements are the partial derivatives of the functions, and two column matrices: a column
matrix of increments of the variables =∆?@, and a column matrix of the incremental values of the
functions about the constants, =∆ @. The matrix of partial differentials is called the Jacobian matrix, and
denoted, =A@,
δ1
δ
2
..
[x] = N
δ = [δ ]
V1 [V ]
V2
..
V N
and the nonlinear function increment [∆y] ordered so that the first components correspond to real power
and the last ones to reactive power
[∆P([x])]
[∆y ] = [∆f ( x)] =
[∆Q([x ])]
The functions [∆P([x ])] and [∆Q([x])] are called the real and reactive power mismatches since they are
given by the difference between the values of the specified and the calculated net real power and
reactive power injections, respectively.
We can now write equation (3.28 or 3.29) using power system variables (for all buses except the swing
bus, which is denoted zero) to obtain equation (3.30) in, which the Jacobian matrix is partitioned into
four submatrices (JA, JB, JC, JD), the column matrix of the variables into two submatrices (∆δ, ∆V), and
column matrix of function increments into two submatrices (∆P, ∆Q).
[J B ] = ∂P — Partial derivatives of real power with respect to voltage magnitudes
∂ V
change significantly (i.e., ∂Q = 0 ). It amounts to neglecting [JB] and [JC] sub-matrices of the exact
∂δ
Newton-Raphson method.
∂P
∂δ [∆δ ] = [∆P]
∂Q
[ ∆V ] = [∆Q]
∂ V
There exists also a modification of this method, which is often referred to as fast de-coupled method. In
this method the Jacobian matrix values of the de-coupled method are made constant in value throughout
the iteration (i.e., they are not updated with the progression of iteration).
3.4.4. DC LOAD FLOW MODEL
The linearised DC power flow method greatly simplifies the power flow solution by making a number
of approximations, including:
(i) completely ignoring the reactive power balance equations;
(ii) assuming all bus voltage magnitudes are identically 1.0 per unit;
(iii) ignoring the line losses (i.e., assuming x Line >> rLine ); and
(iv) ignoring tap dependence in the transformer reactances.
The implication of these assumptions is that the DC power reduces the actual power flow problem to a
set of linear equations:
[ ]
N N
Pi = ∑ Bik Vk Vi cos 90 − (δ i − δ k ) = ∑ Bik sin(δ i − δ k )
0
k =1 k =1
where Bik is the series susceptance of the line (i.e., the inverse of the series reactance of the line, Xik), θik
= 900 and Vi = V k = 1.0 .
If the difference between the voltage angles is negligible, then the sine function will tend to the
difference itself and the above load flow equation may be written in the following linear form:
N N N
Pi = ∑ Bik ϕ ik = ∑ Bik δ i + ∑ (− Bik δ k )
k =1 k =1 k =1
V1, δ1 V2, δ2
System X System
#1 #2
P
Vi EG ,i ∠(90 − δ ) − Vi 2∠90
SG ,i = PG ,i + jQG ,i = Vi I i* = (1)
X G ,i
= i (EG ,i cos δ − Vi )
Vi EG ,i Vi 2
sin (90 − δ ) −
V (3)
QG , i =
X G ,i X G , i X G ,i
Equation (2) shows that if Vi and EG,i are held constant, the real-power output of the generator is directly
dependent on the load (power) angle, δ. Of course, in steady-state operation real electric power output
can only be increased by an increase of real mechanical power input. Therefore, the power angle must
also be directly dependent on the mechanical input. The mechanical power input is normally controlled
by use of appropriate governor settings. In Equation (3), if Vi and EG,i are held constant, Q would appear
to be affected by changes in the power angle or changes in mechanical power input. However, the
normal operating range for the power angle is less than 150. For small changes of the power angle in this
operating range, the cosine term is relatively unchanged while the sine term will vary significantly. This
implies that reactive-power output of a generator is not directly dependent on its mechanical power
input. It is primarily dependent on the magnitude of the generator voltage, EG,i if Vi and δ are held
constant. For a generator operating at a constant speed, the generator voltage is a function of the DC
excitation on the rotor.
From this discussion the reactive power output of a generator appears to be controlled by adjusting the
rotor DC excitation, and the real-power output appears to be controlled by varying the mechanical
power input. By controlling the amount of real and reactive power generated at each generating station
in a power system, the general flow of power in the system can be controlled.
3.5.2. Load-Flow Control Using kVAr Sources
Most loads in a power system are inductive. Reactive power must, therefore, be generated to supply this
inductive load. On the real power that is delivered to the load actually does real mechanical work.
However, delivering the same amount of real power to a load with a reactive part requires more current
than supplying a load with no reactive part. Larger currents mean greater voltage drops along
transmission and distribution lines and more ohmic line losses. Both of these conditions are undesirable
since they may result in violation of line thermal, voltage and stability constraints.
One way to overcome these disadvantages is to generate reactive power at the location that it is needed.
For this purpose, a variety of reactive power (VAR) generators can be used. But one source of reactive
power, which is widely used in power systems, is the capacitor banks. Capacitors consume reactive
power, which is the same as supplying VARs. Therefore, the installation of a capacitor bank at the load
bus will have the effect of reducing the amount of current required from the generating station to supply
a given amount of load. This phenomenon is illustrated in Figure 4.1.3. The result of adding the properly
Vb Regulating Vb + ∆Vb
Transformer
Vc Vc + ∆Vc
– END –