See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228364415

Modified Algorithm of Load Flow Simulation for Loss Minimization in Power

Systems

Article

CITATIONS READS

14 6,612

2 authors, including:

Trevor Blackburn
UNSW Sydney
209 PUBLICATIONS   3,076 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PD data mining View project

All content following this page was uploaded by Trevor Blackburn on 22 May 2014.

The user has requested enhancement of the downloaded file.

 MODIFIED ALGORITHM OF LOAD FLOW SIMULATION FOR LOSS
MINIMIZATION IN POWER SYSTEMS

D. Lukman T.R. Blackburn

School of Electrical and Telecommunication Engineering
University of New South Wales, Kensington
Sydney, NSW-2052, Australia
e-mail : d.lukman@student.unsw.edu.au / d_lukman@hotmail.com
Abstract
This paper focuses on the use of load flow to minimize the losses in an electrical power system. A Matlab
load flow analysis program has been developed using a modified Newton Raphson algorithm based on a
Y-bus admittance matrix to determine the voltage level of the system bus. Losses are then calculated
using B-loss coefficient formula and verified with the traditional I2R or differential power methods.
Voltage control using either switched capacitor bank or load tap changer of transformer shall be done to
improve the voltage level whilst minimizing losses. A network of 5-bus system is used as the test case.
Modeling of capacitor bank and tap changer are carried out and implemented in the program. Several
case studies with different values of capacitances and tap settings of transformers are conducted to
determine the minimum losses.
(2) addition of different values of capacitor banks to
1. INTRODUCTION control reactive power distribution
System changes can then be simulated using a Newton-
Load flow deals with the flow of electrical power
Raphson load flow computer program developed. The
from one or more sources to loads consuming energy
results of such changes are described.
through available paths as commonly shown in a one
line diagram. Electric energy flow in a network
The other method used in this work to calculate losses is
divides among branches according to their respective
using B-losses coefficients, which express the transmission
impedances until a voltage balance is reached in
losses as a function of the outputs of all the
accordance to Kirchoff’s Laws. The flow will shift
generation/power plants.[2] Hypothetically, B-losses
anytime the circuit configuration is changed or
coefficients can bias the operation of transformer tap
modified, generation is shifted or load requirements
changers and/or capacitive reactive power
change. Information about these changes are
adjustment/FACTS devices inside the traditional Newton-
important for industrial plants and electric utility
Raphson load flow algorithm. Instead of having certain
operators to ensure efficient operation, minimize
target voltages, we allow voltages to vary within a 5%
losses, maintain reliability of service and coordinate
tolerance of 1 per unit rating in order to obtain minimum
protective relaying for unexpected and emergency
losses whilst improving the voltage level.
conditions.

The losses in electrical network distribution as well as 2. NEWTON-RAPHSON LOAD FLOW

real and reactive power flows for all equipment AND MODIFIED ALGORITHM
connecting the buses can be computed by means of A power flow or load flow program computes the voltage
load flow simulation. The quantification and magnitude and angle at each bus in a power system under
minimization of losses is important because it will balanced three phase steady state conditions. Once they
determine the economic operation of the power are calculated, real and reactive power flows for all
system[5]. If we know how the overall losses occur, equipment interconnecting the buses, as well as equipment
we can take steps to minimize them. Active power losses are also computed.
losses can be determined by various methods. It can
simply be computed as I2R. The power loss in a line There are two ways to represent the bus voltage equations
can also be calculated by taking the algebraic sum of to solve the Newton-Raphson load flow problem. Most
the total power flows in either direction and the total references use rectangular coordinates of bus voltages.
loss would be the sum of all the line losses[3]. [2,3,4]. We prefer to use polar coordinates of bus voltage
as used in [1] as it will be implemented in the Matlab
Two methods to reduce the losses on the system simulation program developed.
network, which will be discussed in this paper
include: Consider first the non-linear equation y = f(x)
(1) the change of transformer tap settings

1/7
 δ 2   P2   P2 (x) 
.  .   . 
      Eq. [2.1]
.  .   . 
     
δ  δ N  P PN P(x)  PN (x)
x =   =  ; y =   =  ; f (x) =   =

V  V2  Q Q2  Q(x)  Q2 
     
.  .   . 
.  .   . 
     
 N 
V  N 
Q QN (x)

where V (voltage), P (real power) and Q (reactive

power) terms are given in per unit and δ (phase
angle) terms are in radians. The swing bus variables
δ1 and V1 are omitted from Equation [2.1] because
they are already known. This equation shows that the
real and reactive powers at every bus except the slack
bus can be expressed as a function of voltage
magnitude and phase angles. Figure 1. (a) Off-load Tap changing transformer. (b)
On-load tap-changing transformer with S1 and S2
The outputs of Newton-Raphson load flow algorithm transfer switches, T centre-tapped reactor [4]
which give the voltage levels at each bus, power flow
in the line connecting two buses in either direction The presence of a tap changer allows manual or automatic
and line losses can be controlled by applying the change of the turn ratio, and hence of the output voltage.
modified algorithm to the original Newton-Raphson Because of the impedance of the lines, the voltage at the
algorithm by means of the followings: receiving end is slightly lower than the voltage at the
sending end for most loads. In order to get a constant and
1. Changing transformer tap changers rated voltage at the secondary of a ‘normally’ step-down
2. Additional of switched capacitor bank transformer automatically, an on load tap changer with
3. Application of B-losses formula additional S1 and S2 transfer switches and R centre-tapped
These three methods will be outlined on the following reactor is mounted at the primary side of it as shown in
sections and implemented in the program developed. Figure 1 (b).

3. APPLICATION OF TAP Assume that an automatic load tap changing transformer

CHANGERS OF TRANSFORMER (OLTC) is connected to a particular bus to keep load
Tap changing can control the reactive var flow so voltage constant. It is possible to run the load flow
optimum bus voltages can be determined and reduce program employing one tap setting and without mentioning
line losses. A method of controlling the voltages in a the magnitude of load voltage. If the voltage magnitude
network makes the use of transformers, the turns ratio determined by the load flow program run exceeds the given
of which may be changed. A schematic diagram of limits, a new tap setting is then selected for the next run.
an off-load tap changer is shown in Figure 1 (a) In general, when the automatic tap-changing feature is
which requires disconnection of the transformer when employed to represent a manual tap-changing transformer,
the tap setting is to be changed. Many transformers the output of the load flow program will specify the tap
now have on-load tap changers as can be shown in setting that gives the required bus voltage. The change of
Figure 1 (b). tap setting or turn ratio will change the system impedance
matrix. Therefore, after each tap ratio adjustment, the Ybus
admittance matrix has to be adjusted.

Another means of taking into account the LTC transformer

is to represent it by its impedance, or admittance,
connected in series with an ideal autotransformer, as shown
in Figure 2 (a). A model of a load tap changer needs to be
developed. An equivalent Π circuit, as shown in Figure 2
(b) [12], can be developed in load flow studies. The
presence of the tap changing transformer causes necessary

2/7
modifications to the Newton-Raphson power flow transfer) and to adjust the system voltage.[1,2] Shunt
technique. The elements of the equivalent Π circuit, capacitors are used to deliver reactive power and increase
can then be treated in the same manner as line the voltage magnitudes during heavy load conditions.
elements. Figure 3 shows the effect of adding a shunt capacitor bank
to a power system bus. The system is represented by its
Thevenin Equivalent at the node, where the capacitor will
be applied by closing the switch. With the switch open, the
node voltage Vt is equal to the Thevenin voltage Eth.

Figure 3. Effect of adding a shunt capacitor to a power

system bus
(b)
From the power flow standpoint, the addition of a shunt
Figure 2. LTC transformer representations: capacitor bank to a load bus corresponds to the addition of
(a) equivalent circuit; (b) equivalent Π circuit [12] a negative reactive load. The power flow program
computes the increase in bus voltage magnitude along with
The following parameters of the equivalent Π circuit the small change in phase angle.
(Figure 2 (b)) in terms of admittances and off-
nominal turns ratio T can be derived:[3] The additional capacitor is modeled with the susceptance
y ij  1 11  B. Given a required reactive power injection of Q, the
A= ; B = 1 −  y ij ; C =  − 1 y ij susceptance B can be calculated from Q = V2B. V is the
T  T T T  initial voltage of the bus where the shunt capacitor needs to
T = per unit turns ratio (i.e tap setting is +1.25% then be installed.
T = 1.0125) [10].
The addition of capacitor bank changes the bus admittance
B and C can either be an inductor or capacitor. If we matrix similar to the change of tap setting of transfomer.
want to increase the voltage of the transformer output, However, it will only affect the element of the diagonal
normally by taking positive tap, B is chosen as admittance matrix of the bus where the capacitor is added.
inductor and C is a capacitor and vice versa. The
presence of a tap changing transformer changes the
elements of both diagonal and off-diagonal of bus
5. B-LOSSES CALCULATION
admittance matrix where the transformer is connected The B matrix loss formula was originally introduced in the
between two buses. The Newton Raphson load flow early 1950s as a practical method for loss and incremental
simulation is then rerun to obtain the required output. loss calculations[7]. In this method, the results of power
flow is used to account for power transmission losses in the
power system. It is important in terms of the economic
4. APPLICATION OF SWITCHED
dispatch problem[11] to express the system losses in terms
CAPACITOR BANKS of active power generations only. This is commonly
Capacitors are used in the transmission/distribution referred to as the loss formula or B-coefficient method.
line to increase line loadability (maximum power

3/7
The simplest form of loss equation is George’s
formula[2], which is: Knowing that the real power losses are a function of
k k generations and B-losses coefficient, varying the
PL = ∑∑P
m =1 n =1
m B mn Pn Eq. [5.1]
generations to fulfill the power demand will change the
losses accordingly. If B-losses are reduced, the losses can
where PL is the power losses
be minimized. Since B-losses coefficients are functions of
Pm, Pn is the power generation from all
resistances of every line, voltage magnitudes and power
generator sources
factors at each generation, phase angle of generator
The coefficients Bmn are commonly referred as the
currents and current distribution factors from each
loss coefficients with the units of reciprocal
generation, while resistances are physical properties of
Watt/MWatt. The B coefficients are not truly constant
electrical equipment, which tend to be constant, improving
but vary with unit loadings. A more general formula
the voltage at certain points will minimize B-losses
(Kron’s loss formula) is given by:
k k k
coefficients. Voltage control using either variable tap
PL = K L 0 + ∑ Bm0 Pm +∑∑ Pm Bmn Pn Eq. [5.2] changing transformers or capacitors as explained before are
m =1 m=1 n=1
necessary to improve the voltage levels and minimize
losses. These can be implemented in the load flow
A linear term ΣBm0Pm and a constant KL0 have been
simulation developed as explained in Section 6.
added to the original quadratic equation. This shows
that losses depend on the active power generations P
only. Bmn is called the loss coefficient and is given by 6. MATLAB SIMULATION AND
a general expression: SENSITIVITY ANALYSIS
cos(σ m − σ n )
∑ N km N kn Rk Eq. [5.3]
Matlab was chosen as the simulation tool for this research
Bmn =
Vm Vn ( pf m )( pf n ) k
because of the ease of manipulation of matrix structures
and inputs. It has in-built routines such as inverse function,
where σm, σn are phase angles of currents Im,In abs function, and so on, graphing facilities to plot
Vm, Vn are voltages at bus m and n convergence of load flow.
Nkm, Nkn are current distribution factors
pfm, pfn are power factors A single line diagram of a five bus system shown in Figure
For a simple system consisting of two generating 5 [1] was chosen to be tested. This system is selected
plants and one load as shown in Figure 4, losses PL in because it represents a typical meshed network where a
terms of power output of the plants and B-losses load bus is supplied from alternative sources. It has typical
coefficient can be derived as[5]: line ratings. The diagram has two generators of 400 MVA
PL = P12B11 + 2P1P2B12 + P22B22 Eq. [5.4] and 800 MVA power ratings, two step down transformers
where of 400 MVA and 800 MVA ratings, three long distance
Ra + Rc transmission lines and one remote load bus. Overall, this
B11 = system consists of five buses and five branches. Bus 1 is
V1
2
( pf1 )2 assigned as the slack/swing bus. Bus 3 is the voltage
controlled PV bus while Bus 2,4 and 5 are load buses. The
Rc
B12 = base of apparent power is 400 MVA. Vbase = 15 kV at
V1 V2 ( pf 1 )( pf 2 ) buses 1,3 and 345 kV at buses 2,4,5.

Rb + Rc
B22 =
V2
2
( pf 2 )2

Figure 4. A simple radial system of two Figure 5. One line diagram of 5 bus power system[1]
generators and one load bus

4/7
Input data for the simulation is shown Table 1. Voltages at bus 1-5 were calculated and the results are
shown in Table 2. Notice that the voltage magnitude at bus
2 of 0.8338 per unit is under-voltage. Voltage level at bus
2 needs to be improved by applying voltage control.
V= magV = angleV =
1.0000 1.0000 0
0.7708 - j0.3178 0.8338 -22.4063
1.0499 - j0.0109 1.0500 -0.5973
1.0181 - j0.0504 1.0193 -2.8340
0.9712 - j0.0773 0.9743 -4.5479
Table 2. Bus Voltage Outputs in pu & degrees
Real power losses were calculated using differential
powers method. Losses turn out to be 0.0871 per unit or
3.81%. This value has been verified by calculating losses
from I2R. Calculation using B-loss formula gives losses
equal to 0.0823 pu or error = 5.5%.

6.1 Applying voltage control by changing tap setting of

Transformer between bus 1 and 5
Using Matlab, parameters of the equivalent Π circuit of the
tap changing transformer between bus 1 to 5 are calculated
to be A = 0.9208 – j12.2766, B = 0.0115 – j0.1535 and C =
-0.01137 + j0.1516. This shows that B is an inductor and
C is a capacitor.
Table 1. Bus input data, line input data and
transformer input data of Figure 5.[1] The tap setting is increased in step of 1.25 % and the load
flow is rerun to obtain an acceptable voltage level within ±
A Newton-Raphson load flow algorithm has been 5 % of unity which gives the minimum losses. Two ways
implemented in Matlab and the five bus power of calculating real power losses using traditional I2R and
system was simulated. The load flow program B-loss formula are compared. The minimum losses of
converges after 5 iterations. 0.0704 pu using I2R as shown in Figure 7 occurs using tap
setting = 15% with voltage of bus 2 of 0.941 pu. Minimum
Plot of convergence of Newton-Raphson in terms of losses of 0.0676 pu using B-losses as shown in Figure 7
power mismatches errors versus number of iterations occur at tap = 10% with voltage of bus 2 of 0.9098 pu.
are shown in Figure 6. The error drops rapidly from These two voltage levels are still not acceptable. Minimum
iteration 1 to iteration 2 and then reduces at slower losses with acceptable voltage magnitude of 0.9553 pu is
rate and finally reaches zero sluggishly after five obtained with tap = 17.5% where the losses are slightly
iterations. The tolerable power mismatch error was different using both methods.
assigned to be 5e-5. 2
Real Power Losses using B-losses and I R vs Tap Setting

0.09

0.085
Real Power Losses (pu)

0.08

I2R
0.075
B-losses

0.07

0.065

0.06
1.0125 1.025 1.0375 1.05 1.0625 1.075 1.0875 1.1 1.1125 1.125 1.1375 1.15 1.1625 1.175 1.1875 1.2
Tap Setting of Transformer (1.25 - 20 %)

Figure 7. Graph of Real Power Losses based on I2R and

B-losses vs Tap Setting of Transformers.
Figure 6. Convergence of Newton Raphson

5/7
 7. CONCLUSION
The two curves are slightly different. They intersect
at tap = 15% where the losses are equal to 0.0704 pu.
A Matlab load flow simulation program has been
developed using a modified Newton-Raphson algorithm to
6.2 Applying voltage control by adding shunt calculate and control the voltage, determine real and
capacitor at bus 2 reactive power flows and compute real power losses.
Now we want to improve the voltage level at bus 2
whilst minimizing the total real power losses. The
Voltage control using tap changers can be implemented in
capacitor has been modeled with susceptance B. The
the load flow analysis by using a Π equivalent circuit.
first run of load flow program gave a voltage at bus 2
Optimum tap setting can be determined by load flow
of 0.8338 magnitude. The sensitivity analysis is
simulation, which gives minimum real power losses and
carried out by adding 5 steps of 0.2 pu reactive power
acceptable voltage level. Optimum voltage control by
injection from 0.2 to 1 pu. Susceptance can then be
means of switched capacitor bank can also improve the
calculated from B = Q/V2.
voltage level at a bus to result in minimum power losses.
The shunt capacitor is added at the bus where the bus
Minimum losses of 0.0583 pu calculated from I2R as
voltage is under voltage before load flow simulation is
shown in Figure 8 is attained by adding 0.6 pu or 300
rerun. Optimum value of capacitance is obtained from
Mvar with acceptable bus 2 voltage of 1.0595 pu.
simulation, which gives minimum losses at acceptable
Using B-loss formula, minimum losses 0.0733 as
voltage level.
shown in Figure 8 is achieved by adding 0.4 pu of
reactive power with acceptable voltage of bus 2 of
Results of losses using I2R are slightly different than the
0.9791 pu.
results calculated using B-losses formula. At this stage it
Real Power Losses using B-losses and I2R vs Capacitor Bank
can be said that the calculation of losses using I2R is more
0.1
accurate because it has been verified with differential
0.095
power method which gives the same results. Although it is
0.09
not so accurate, calculation of losses based on B-losses
0.085
coefficients is useful because it allows optimization
Real Power Losses (pu)

0.08

0.075
B-losses
configuration to achieve minimum losses.
I2R

0.07

0.065
Future research will explore the use of B-losses
0.06
coefficients to determine the tap settings of the transformer
0.055
or the capacitance of capacitor banks to satisfy the required
0.05
voltage level whilst minimizing losses.
0.2 0.4 0.6 0.8 1
Capacitor Bank Reactive Power (pu)

Figure 8. Graph of Real Power Losses using I2R 8. REFERENCES

and B-losses vs Capacitor Bank Reactive Power 1. Glover, J.D. and Sarma, M. 1994, Power System
Analysis and Design, 2nd ed., PWS Publishing
The two curves have similar shapes although the Company, Boston
losses calculated using I2R are smaller than the losses 2. Stevenson, W.D. 1975, Elements of Power System
calculated using B-loss coefficients. Analysis, 3rd ed., McGraw-Hill Kogakusha, Ltd.,
Tokyo.
It has been shown that the real power losses vary 3. Stagg, G.W. and El-Abiad, A.H. 1968, Computer
parabolically with either tap setting of transformer or Methods in Power System Analysis, McGraw-Hill
capacitor reactive power. Hence, there is a point Book Company, New York.
where minimum real power losses occur. This is an 4. Weedy, B.M. and Cory, B.J., 1998, Electric Power
optimum point for the operation of an electrical Systems, 4th ed., John Wiley & Sons, West Sussex.
power system as long as the voltage is within 5. Lukman, D., Blackburn, T.R and Walshe, K, Loss
allowable range. Minimization in Industrial Power System Operation,
Proceedings of the Australasian Universities Power
Engineering Conference (AUPEC’94), Brisbane,
Australia, 24-27 September 2000.

6/7
 6. Penny, J. and Lindfield, G., 1995, Numerical Similar equations hold in terms of ∆P and ∆V, and ∆Q in
methods using MATLAB, Ellis Horwood Limited, terms of ∆δ and ∆V.
Hertfordshire
7. Wood, A.J. and Wollenberg, B.F., 1996, Power Hence, the Newton-Raphson method requires that a set of
Generation, Operation and Control, 2nd ed., John linear equations be formed expressing the relationship
Wiley and Sons, Inc., New York. between the changes in real and reactive powers and the
8. Del Toro, V., 1992, Electric Power Systems, components of bus voltages and phase angles. The
Prentice Hall, Inc., New Jersey. Jacobian matrix can be partitioned into four blocks J1, J2,
9. ANSI/IEEE Standard 399-1980, IEEE J3 and J4.
recommended practice for industrial and ∆P   J1 J 2 ∆δ  Eq. [A.4]
commercial power systems analysis, 1980, Power ∆ Q  =  J  
System Technologies Committee of the IEEE    3 J 4  ∆V 
Industry Applications Society. The partial derivatives in each block can be derived from
10. Parker, A.M., The Modeling of Power System Equations [A.1] and [A.2]. The unknown quantities in
Components, 1997 Residential School in Equation [A.4] are the elements of the column matrix of
Electrical Power Engineering, UNSW, Australia, the changes in the phase angle and voltage of each bus.
26 Jan – 14 Feb 1997. Convergence criteria are often based on ∆y(i) or power
11. Jabr, R., Coonick, A.H and Cory, B.J., A Study of mismatches rather than ∆x(i) or phase angle and voltage
the Homogeneous Algorithm for Dynamic magnitude mismatches.
Economic Dispatch with Network Constraints
and Transmission Losses, IEEE Transactions on Once the voltage at each bus is computed, line flows can be
Power Systems, Vol. 15, No. 2, May 2000, pp calculated. The current at bus k in the line connecting k
605-611 and n is given by:
y ' kn
i kn = (V k − V n )y kn + V k
12. Gonen, T. 1988, Modern Power System Analysis,
John Wiley and Sons, Inc., New York Eq. [A.5]
2
where ykn = line admittance, y’kn = total line charging
APPENDIX admittance and (Vk×y’kn)/2 = current contribution at bus k
due to line charging. The real and reactive power flow
Power flow solutions by Newton-Raphson are based
from k to n is found to be:
on the non-linear power flow solutions. It can be
y'kn
shown that the power flow equations of y = f(x) can Pkn − jQkn =Vk *ikn =Vk *(Vk −Vn)ykn +Vk *Vk Eq. [A.6]
be written as: 2
N
yk = Pk = Pk (x) = Vk ∑YknVn cos(δ k − δ n −θkn ) Eq.[A.1]
The power loss in line k-n is the algebraic sum of the
power flows in either direction.
n=1
N
yk+N = Qk = Qk (x) = Vk ∑YknVn sin(δ k −δ n −θkn ) Eq.[A.2]
n=1
where k = 2, 3, …, N
Ykn is the element of the bus admittance
matrix between buses k and n
Hence, there are two non-linear simultaneous
equations for each node. The real and reactive
powers depend on the product of the sum of the
voltages connected between two buses and the
admittance between the buses. The bus admittance
matrix can be first formed from the impedances
connected to a bus or between two buses.

Changes in P and Q are related to changes in V and δ

by Equations [A.1] and [A.2], e.g.
∂P2 ∂P ∂P
∆P2 = ∆δ2 + 2 ∆δ3 +...+ 2 ∆δ N Eq.[A.3]
∂δ2 ∂δ3 ∂δ N

7/7

View publication stats