Tuning of Power System Stabilizer in Single

Machine Infinite Bus (SMIB) using Genetic

Algorithm and Power Factory Modal Analysis
Ali Kharrazi
School of Engineering and Information Technology
Murdoch University
WA, Australia
Ali.Kharrazi@gmail.com

Abstract--Power system stabilizers have been widely used combination of power and speed have been also
in generation systems to enhance the transient stability proposed in some papers as input signals [3] [4]. They
of the power system. Low frequency oscillation may
occur as a result of excitation system which exerts a add an element to the electrical torque which is in
phase lag. Conventional power system stabilizers using a phase with speed deviation. The conventional power
lead-lag controller have been used in utility for decades. system stabilizer is shown in Fig 1.
Proper tunning of the parameters of stabilizer is
essential for effectiveness of stabilizer. Many
optimization techniques have been proposed for finding
the optimum parameters of stabilizer. In this paper
tuning of parameters of a Conventional Power System
Stabilizer in a Single Machine Infinite Bus (SMIB) using
Genetic Algorithm (GA) is proposed. The power system
is modelled in Power Factory and the optimization is Figure 1 Conventional Power System Stabilizer
done through MATLAB optimization toolbox. The cost
function is to maximize the damping ratio which is It consists of a washout filter to filter the steady state
calculated using Power Factory Modal Analysis. Using voltage and two lead-lag controller blocks followed
DIgSILENT Programming Language (DPL), a real time
data exchange between MATLAB and Power factory is
by limiter block to limit the output. The parameters of
set during optimization process. The result of simulation the lead-lag blocks should be tuned properly to
proves the effectiveness of optimization process. compensate for the phase lag due to exciter system,
generator and power system. Two common
Keywords: Power system stabilizers, Genetic Algorithm,
MATLAB, DIgSILENT, Modal Analysis approaches to provide enough damping to the system
are minimum phase and root locus. In the first
technique, the parameters of stabilizer are adjusted to
compensate the phase lag using frequency response
analysis. In the second one a modal analysis is carried
I. Introduction
out to find out the position of eigenvalues associated
Automatic voltage regulators have been utilized in
with mode of oscillation. Then the parameters of
power systems since 1960’s [1]. Although they
stabilizer are tuned so that the zeros and poles of the
improve steady-state stability of the transmission
stabilizer are placed in the s-plane in order to shift the
system they have a detrimental effect on transient
eigenvalues to the left as much as possible.
stability because they increase the phase lag of the
system [2].Oscillation at low frequencies of 0.2 to 2.5 In this paper the second technique is applied to a
Hz may occur in the system due to large disturbances Single Machine Infinite Bus system. The power
like phase-to-ground faults in a transmission line. system is modelled in DIgSILENT Power Factory
Power system stabilizers are used to add damping of which is a powerful tool for simulation of power
the system through modulation of excitation system systems. The modal analysis tool is used to determine
by adding a component to the electrical torque which the modes of oscillation. The resulting data from
is in phase with speed deviation. They normally use modal analysis is used to optimise a cost function
speed deviation as input signal. Bus frequency and using MATLAB’s Genetic Algorithm optimisation

978-1-4799-8725-2/15/$31.00 ©2015 IEEE

toolkit. Data exchange routine proposed
p in [5] is in this block are defined and maay be changed from the
established using the DIgSILEN NT Programming scripts using DIgSILENT Progrramming Language.
Language in order to run the optimiisation in real time.
In [5] particle swarm optimisation is used to tune the
parameters of two area system. Inn this paper GA is I. Modal Analysis
replaced due to faster responsse in real time Modal Analysis has been used ini many publications to
applications and simplicity of algoorithm. Since it is study the stability of Pow wer Systems [6] [7].
heuristic search technique nonlinnearity and non- Electromechanical modes of oscillation
o are used to
differentiability does not limit its application. determine the characteristics of oscillations. The
Although the solution gained by GA A is not unique but eigenvalues and eigenvector of a linearized state space
it is close enough to global acceptabble solution. model are calculated in most studies. In
Power Factory the modal analysis tool uses
In the next section the power system model in Power Differential Algebraic Equationns (DAEs) as illustrated
Factory is described. In section 3 the
t modal analysis in [8] to calculate the modes of the system. The
are discussed and the result of pow wer system modal simulation results include the imaginary
i and real part
analysis is presented. Section 4 deescribes the tuning of the eigenvalues as well as the damping ratio and
process which consists of GA settinng in MATLAB as frequency of oscillations.
well as automatic data exchangee between Power
Factory and MATLAB. Section 5 depicts
d the result of
tunning as well as the result of simuulation to show the
effectiveness of the proposed tunning approach. The
Final section presents conclusion.

II. Power System

A single Machine Infinite bus shown in Fig.2 is
modelled in Power Factory.

.Figure 4 Synchronous Machine Signal

S Interconnections

Figure 2 Single Machine Infinnite Bus

Fig. 5 depicts the result of moodal analysis of SMIB
The control system consists of an IEEE Type AC4A and the real and imaginary part of the Eigenvalues.
Automatic Voltage Regulator (AV VR) and a Steam
Turbine Governor with fast valvinng (TGOV2). The
block diagram shown in The diaagram in Fig.1 is
modelled in Power Factory whicch acts as Power
System Stabilizer with a limiter at the
t end in order to
limit the output to -0.2 to 0.2 pu.

A. Power Factory Models

The control system for the geneerators consists of
Automatic Voltage Regulator (AV VR), Governor and
Power System Stabilizer (PSS). Powwer Factory uses a
frame allowing different controol blocks to be
modelled. This frame indicates the signal flow during
simulation. In Fig.3 the frame settup for controllers
and synchronous machines is show wn. The parameters Figure 5 System eiggenvalues
It is also possible to store the results of modal analysis deploying the algorithm. The search criteria in based
as a spreadsheet in form of a *.csv file. This solely on fitness function. The fitness of each
spreadsheet contains the values such as real and individual is examined and compared. The selection
imaginary part, damping ratio, damping frequencies process is copied from biological evolution including
etc. In Table 1 the values of the dominant eigenvalues mating and mutation. Since the selection of
which relate to modes 11-12 (Electromechanical populations is based on a stochastic process the
Oscillation modes) and mode 8-9 are shown. The solution to the problem is not unique [10]. In many
mode numbers are randomly assigned by Power past papers GA has been applied in the area of power
Factory. system stability. [11] [12] [13] [14].

Table 1 Dominant eigenvalues with no PSS The fitness function is defined using the result of
modal analysis in Power Factory. It is possible to save
Imaginary Damped Damping
Name Real part part Frequency Ratio the result of modal analysis in form of spreadsheet
1/s rad/s Hz (CSV file) that could later be used by MATLAB. The
values of real and imaginary part of the eigenvalues
Mode 11 -0.1416 3.8866 0.61857 0.0364
are saved in form of a CSV file using ASCII RESULT
Mode 12 -0.1416 -3.887 0.61857 0.0364
EXPORT (ComRes) element in Power Factory. This
Mode 8 -22.8 7.6182 1.21247 0.94846 file is loaded in MATLAB environment and the
Mode 9 -22.8 -7.618 1.21247 0.94846 damping ratio of all eigenvalues are calculated using
(1).

ିఈ
As seen in Table 1 the damping ratio of ݀ܽ݉‫ ݎ݌‬ൌ (1)
ඥఈ మ ାఉమ
electromechanical modes is around 3%. Under
different loading conditions this ratio could In (1), ߙ and ߚ are real and imaginary parts of
deteriorate and the system could become unstable i.e.
the Eigenvalues in the S-plane. This ratio is
the oscillations could persist which may cause the
calculated for all eigenvalues with positive
separation of the network.
imaginary part and the minimum damping ratio
is calculated. The fitness value for GA as
II. Tuning Process described in (2) is finding the possible maximum
By adding PSS units to the generator’s control loop value of the minimum damping ratio.
the poles and zeros of the PSS change the position of
the eigenvalues in the S-plane. The ultimate goal of ‫ ܬ‬ൌ ƒšሼ‹ሺ݀ܽ݉‫ݎ݌‬௜ ሻሽ ݅ ൌ ͳǡʹ ǥ ǥ (2)
the tunning procedure is to shift all oscillatory modes
to the left of the S-plane to an optimum place. The As the aim is to tune the parameters for
parameters to be changed are gain and time constant maximum damping ratio and GA is a
of lead lag blocks. Many optimization techniques minimization algorithm the reverse of damping
have been used by past researchers to find the ratio is considered in encoding. The genetic
optimum parameters including particle swarm encoding is made of 5 parameters including
optimization [5], Tabu search, Fuzzy logic, Genetic ܶଵ ǡ ܶଶ ܶଷ ܶସ and K. The upper and lower limits of
Algorithm and other Artificial Intelligent techniques
the parameters are defined according to Table 2.
[9]. In this work Genetic Algorithm solver in the
MATLAB Optimization Toolbox is used to find the B. Automatic Data Exchange between
optimal values for PSS parameters. MATLAB and POWERFACTORY
Power factory uses a set of script as command codes
A. Genetic Algorithm
to facilitate the process of simulation. These
GA is a search technique based on biological
commands are written using DIgSILENT
evolution to find the optimal or a near optimal
Programming Language (DPL) and they are used to
solution to an optimization problem. A fitness
add some sort of automation to the simulations.
function should be defined which is the cost function
Changing parameters, running different simulation
of the optimization problem. A random population of
and commands could be arranged through DPL
individual is encoded in form of binary strings. The
scripts.
definition of this code is done by the engineers
 Table 2 Parameters upper and lower limit After running a modal analysis in Power Factory the
new positions of the eigenvalues are found and the
Parameters Min Max
K 10 80 oscillation modes which are 7-8 and 14-15are shown
T1 0.1 3 in Table 4.
T2 0.05 1
As seen in Table 4deployment of PSS has deteriorated
T3 0.1 2
the stability of the system and has made the system
T4 0.05 1
unstable with negative damping ratio in modes 14 and
The data exchange proposed in [5] is used to run the 15. The DPL script in power factory and GA script
simulation process. The flowchart shown in Fig. 6 MATLAB are executed simultaneously to perform the
depicts the sequence of the data exchange. process shown in Fig.6. The number of generations is
limit to 10 as the stop criteria. The result after running
Start the algorithm is shown in Fig. 8. The best fitness
value as seen in Fig. 8 is 0.538 which is an acceptable
damping ratio. The solutions which are the parameters
Power Factory runs modal
analysis and writes data as of PSS are set as new parameters of PSS block and the
spread sheet
MATLAB waits for Power
modal analysis is run again which are shown in Table
factory to finish checking 5 and the solution in Table 6.
switch

MATLAB Power factory

Table 3 Randomly chosen Parameters for PSS
changes the changes the
switch switch Parameters Min
K 50
MATLAB runs GA T1 3
optimization and
writes PSS T2 0.03
parameters as spread
sheet T3 2
T4 0.01

Best Fit Found

Table 4 Dominant Eigenvalues with un-tuned PSS
No
Yes
Real Imaginary Damped Damping
End Name part part Frequency Ratio
1/s rad/s Hz

Mode 14 0.2 2.3 0.36 -0.08

Figure 6 Data exchange between MATLAB Power Factory
Mode 15 0.2 -2.3 0.36 -0.08
The simulation is run until the best fit is found and
Mode 7 -20.4 78.9 12.55 0.25
GA stops, returning the best parameters.
Mode 8 -20.4 -78.9 12.55 0..25

III. Simulation and results

In this section the results of simulation after running Table 5 Dominant Eigenvalues with tuned PSS
the tuning process are discussed. The PSS block
Real Imaginary Damped Damping
shown in Fig. 7 is modelled in power factory with Name part part Frequency Ratio
initial parameters shown in Table 3. The output limit 1/s rad/s Hz
is set to ±0.2 p.u. and the washout filter time constant Mode 14 -1.72 2.69 0.42 0.538
to 10s.
Mode 15 -1.72 -2.69 0.42 0.538

Mode 8 -22.2 6.14 0.9 0.96

Mode 9 -222.2 -6.14 0.9 0.96

.Figure 7 PSS block as designed in Power Factory

 Table 6 Tunned Parameters of PSS MATLAB and modal analysis in Power Factory. The
problem of tuning the parameters is converted to an
Parameters value
74.6769 optimization problem which is solved to maximize the
K
1.9514 damping ratio of dominant oscillatory modes. A real
T1
0.2066 time data exchange between MATLAB and Power
T2
1.8068 Factory is established using DIgSILENT
T3
0.6946 Programming Language in form of executable script
T4
in Power Factory. The result of simulation shows the
effectiveness of tuning process. The proposed
To show the effectiveness of the tunned stabilizer a 3 technique could be easily adapted to be used for multi
Phase short circuit in transmission line with 500ms machine power systems with many mode of
duration is modelled and the result of the simulation oscillation by changing the codes in MATLAB and
are shown in Fig 9. PowerFactory scripts.

References

[1] D. Swann and E. Larsen, “Applying Power

System Stabilizer, Part1; General Concept,Part2;
Performance Objective and Tuning Concepts,
Part3; Practical Consideration',” IEEE Trans. on
Power Apparatus and Systems, vol. PAS100, pp.
3017-3046, 1981.
Figure 8 the result of GA optimization
[2] F. d. Mello and C. Concordia, “Concepts of
Synchronous Machine Stability as Affected by
Excitation Control,” IEEE Trans Power
Apparatus and Systems, Vols. PAS-87, pp. 316-
329, 1969.

[3] A. Sharma and M.L. Kothari, “Intelligent dual

input power system stabilizers for multi-machine
system,” in Transmission and Distribution
Conference and Exhibition 2002: Asia Pacific.
IEEE/PES, 2002.

[4] G. Rogers, “The application of power system

stabilizers to a multigenerator plant,” Power
Figure 9 the result of RMS simulation of 3 phase fault in transmission line Systems, IEEE Transactions on, vol. 15, no. 1,
pp. 350 - 355, 200.
As seen in Fig.9 the tuned PSS has better performance in
response to disturbance. [5] A. Stativa, M. Gavrilas and V. Stahie, “Optimal
tuning and placement of power system stabilizer
using particle swarm optimization algorithm,” in
IV. Conclusion
International Conference and Exposition on
A new tunning approach for power system stabilizer
Electrical and Power Engineering (EPE),,
in single machine infinite bus was proposed in this
Romania, 2012.
paper by deploying genetic algorithm toolkit in
[6] J. Ning, X. Pan and V. Venkatasubramanian, [15] A. F. P.M. Anderson, Power System Control and
“Oscillation modal analysis from ambient Stability, First ed., Iowa: The Iowa State
synchrophasor data using distributed frequency University Press, 1997, pp. 37-39.
domainoptimization,” IEEE Trans. Power Syst,
vol. 28, pp. 1960-1968, May 2013.

[7] C. L, M. Y, C. Y and H. W, “Evaluation of

generator damping using oscillation energy
dissipation and the connection with modal
analysis,” IEEE Trans Power Syst, vol. 29, no. 3,
pp. 1393-1402, 06 December 2013.

[8] S. Pizarro-Gálvez, H. Pulgar-Painemal and V.

Hinojosa-Mateus, “Parameterized Modal
Analysis Using DIgSILENT Programming
Language,” in PowerFactory applications for
power system analysis, Cham, Springer, 2014,
pp. 221-248.

[9] L. Hassan and H. M. M. Moghavvemi, “Power

system stabilization based on artificial intelligent
techniques ; A review,” in Technical
Postgraduates (TECHPOS), 2009 International
Conference for, Kuala Lumpur, 2009.

[10] R.H. Dinger, “Engineering design optimization

with genetic algorithms,” in Northcon/98
Conference Proceedings, Seattle, 1998.

[11] A. D. Bomfim, G. Taranto and D.M.Falcao,

“Simultaneous tuning of power system damping
controllers using genetic algorithms,” Power
Systems, IEEE Transactions on, pp. 163 - 169, 06
August 2002.

[12] A. Hasanovic and A. Feliachi, “Genetic

algorithm based inter-area oscillation damping
controller design using MATLAB,” in Power
Engineering Society Summer Meeting, 2002
IEEE, Chicago, 2002.

[13] K. Sebaa and M. Boudour, “Optimal Locations

and tuning of Robust Power System Stabilizers,”
in IEEE Industrial Electronics, IECON 2006 -
32nd Annual Conference on, Paris, 2006.

[14] M. Abido and Y. Abdel-Magid, “Coordinated

design of a PSS and an SVC-based controller to
enhance power system stability,” Electrical
Power and Energy Systems 25, p. 695–704,
2003.