2018 5th International Conference on Control, Decision Thessaloniki, Greece

and Information Technologies (CoDIT’18) April 10-13, 2018

New Intelligent AVR Controller Based on Particle Swarm

Optimization for Transient Stability Enhancement
A. BENSEDDIK, Student Member, IEEE, M. HASNI, M, MENAA, A. EL FERGOUGUI and M.
BOUDOUR, Senior Member, IEEE

Abstract— In this work, we present a new intelligent control short circuit fault by finding the best parameters at each
scheme to enhance the transient stability of multi-machine power instant.
system based on the generator automatic voltage regulation
system. The algorithm consists on the estimation of the best In this work we have been interested to self-tuning adaptive
parameters of voltage regulators to minimize an objective and robust control strategies. Several works were performed to
function. Particle swarm optimization and genetic algorithm enhance power system transient stability, reference [4] has
methods are used to get the best performance of regulation applied the output feedback linearization (OFL), previously
systems. The control performances damp significantly oscillations this technique has been used by taking the SMIB voltage as
and improve the stability of the system after a given disturbance reference , see also [11]. In [8], the authors design a nonlinear
which enhance the transient stability and obtain a new improved control system in the sense of Lyapunov control technique; a
critical clearing time (CCT). The results are very promising and comparison was made including both AVR and PSS. In [3], the
could be validated in practice. direct feedback linearization (DFL) control has been applied.
The nonlinear excitation system controller designed was
I. INTRODUCTION
composed of two parts; the DFL compensator and the
Power system dynamic stability and control is a major topic nonlinear voltage stabilizer.
that has been carried out during the last two decades, in order
The former works were based on the improvement of the
to solve power system dynamic issues following disturbances
feedback reaction of the voltage regulator by improving the
in short and long terms. The number of disruptive events
Riccati’s equation parameters [30], using back stepping, and
increases respectively with the increase of power demand.
sliding mode techniques. Recently, the development in this
Many solutions were proposed to enhance power system
area consists on the design of a new control law based on PID
stability such as load shedding, area islanding, FACT’s devices
regulation in the AVR system based on evolutionary algorithm
control, power system stabilizer, tripping control, fast valving
[14-16], as well as the design of optimal supplementary
and LFC…etc. Recent researches are performed to design
stabilizing signal based on the power system stabilizer PSS
intelligent control based on artificial intelligence, evolutionary
was performed. The latter has been taking a large
algorithms to deal with power system uncertainty. Several
consideration by electric power system researchers for decades
control schemes are proposed in the literature, robust and
because it can enhance the set of power system stability
adaptive controls are the most adequate for dynamic system as
interval issues. Other achievements are expected by a
well as generator responses [1-12]. Such controls make the
coordination of controller including the renewable energy
system more reliable and resilient to various disturbances.
sources, dispersed generation and the performance of ESS
Reliability is the ability of the power system to deliver
[17].
electricity in the quantity and with the quality demanded by
users, it is generally measured by interruption indices [13]. Others works were performed to overcome the issues of
Whereas, resilience is the new word used for system that can transient stability enhancement by including the evolutionary
deal with disruptive events and provide solutions algorithms and the artificial intelligence. Indeed, in [19-24]
autonomously using the available devices and/or solutions PSO, GA, Fuzzy logic, Expert system, Neural Network are
called self-healing. In [14], resilience is the capacity of system used to enhance the transient stability.
to reduce magnitude and/or duration of events. Resilient
In this work, we have implemented two meta-heuristic
infrastructure efficiency depend on the ability to anticipate
optimization techniques as control actions to enhance the
absorb, adapt and/or fast recover from disruptive events. It is
transient stability of the power system. The control strategy is
well known that power system stabilizer is efficient to damp
the robust and adaptive control based on estimation and self-
small oscillations with adequate parameters; as well as efficient
tuning parameters. The obtained results can be used in training
to enhance transient stability. But Instead of using PSS, we
for making decision in multi-agents system. The approach was
interest to making the regulator parameters more intelligent
applied in multi-machine power system with high order system
and efficient to damp sever disturbances such as three phase
components model for a best CCT values accuracy
determination. We note that the approach represents one task
A. BENSEDDIK, M. HASNI, M. MENAA, A. EL FERGOUGUI and M. of an offline decision database suitable in the agents learning
BOUDOUR are with the Laboratory of Electrical Industrial Systems (LSEI). process [29]. Moreover, several power system reduction and
University of Science and Technologies Houari Boumediene, Bab-Ezzouar equivalence techniques are used to deal with the computation
(USTHB),Algiers,Algeria,(Email:[email protected];elfergougui_ahmed@ time issue.
hotmail.com;[email protected];[email protected];mboudour@iee
e.org).

978-1-5386-5065-3/18/$31.00 ©2018 IEEE -587-

 The organization of the paper is as follows. Section II A. Differential algebraic equations
describes the dynamical model of the system under The In aim to predict the transient behavior of the system a
consideration. Section III includes the presentation of the time domain simulation was performed using a numerical
proposed self-tuning control strategies. In Section IV, the integration method (e.g. Rung-Kutta 4, Newton-Raphson) see
simulation and interpretation of the obtained results have been [28-31]. In Matlab, the considered system model consists in
presented. The conclusions are considered in Section V. the nonlinear differential algebraic equation (DAE) of the
turbo alternator (1)-(10), together with the set of algebraic
II. POWER SYSTEM MODEL equations representing the network (11)-(12). Loads are
Control efficiency techniques were mostly performed on a modeled as a constant admittance in (13)-(14).
reduced system. In this study, for the transient stability purpose
we have performed simulation on the single machine infinite di
 i s (1)
bus (SMIB) presented in Fig. 1 and Fig. 2, as well as the dt
western system coordinate council (WSCC) 9-buses 3-
machines test cases presented in Fig. 3. The synchronous 2Hi di ' ' ' '
machines were modeled by the fourth order model equipped  TMi  Edi Idi  Eqi Iqi  (Xqi  Xdi )Iqi Idi
with the standard third order model of IEEE Type 1 excitation
s dt (2)
system [33], as presented in (1)-(4) and (5)-(7), respectively. Di (i s )
'
' dEqi ' '
Td0i  Eqi  (Xdi  Xdi )Idi  Efdi (3)
dt
'
' dE ' '
Tq0i di  Edi  (Xqi  Xqi )Iqi (4)
dt
Figure 1. Single line diagram of the OMIB test case.
dEfdi
TEi  (KEi  SEi (Efdi ))Efdi VRi (5)
dt
dRfi K
TFi  Rfi  Fi Efdi (6)
dt TFi

dVRi K K
TAi  VRi  KAi Rfi  Ai Fi Efdi  KAi (Vrefi Vi ) (7)
dt TFi

VRimin VRi VRimax (8)

1 Rsi
' '  '
Edi'' Vi sini i 
Figure 2. OMIB test case response w(δ). Idi  2
Rsi  Xqi Xdi Xqi
 Eqi Vi cosi i   (9)

1 Rsi Eqi' Vi cosi i 

' '  ' 
 Edi Vi sini i  
Iqi  2 ' (10)
Rsi  Xqi Xdi Xdi

 j i 
 
Ve
i
ji
 di qi  2  PLi Vi  jQLi Vi  
I  I e
 

(11)
N

j i k ik 
VVi kYike
k1


N

PLi Vi  jQLi Vi  VV

j i k ik 
i kYike (12)
k1
P Q
GLi  Li
2 (13) Li   Li
2 (14)
Figure 3. Single line diagram of multi machines power system, case of 3- VLi VLi
machines, 9-buses WSCC.

CoDIT'18 / Thessaloniki, Greece - April 10-13, 2018 -588-

 All power systems test case parameters and initial - Variable constraints
conditions calculations can be found in [29-32].
imin  i  imax (17)
III. PARTICLE SWARM OPTIMIZATION
min max
VRi  VRi  VRi
PSO was used in various controllers for optimal parameters
selection, due to their advantage as the less code line program
and simulation control parameters. In Fig. 3, we present the
flowchart of this technique which was proposed by [25-27]. The proposed control strategy will perform the estimation
of the best parameters at each instant in a real time simulation.
The Automatic voltage regulators parameters are modified
consecutively, as shown in the Fig. 4 below:
Start

KAoptth
Initialize PSO Machine δ
parameters
PSO / GA m

Generate first swarm

KAopt VRmax SE  f Efd 

VRref
Evaluate the fitness
of all particles δi
Vt  KA KE Kg

Record personal best Update the position 1sTA 1sTE 1sTg
fitness of all particles of particles
VRmin
Find global best Update the velocity
particle of particles sKF
1sTF
Swarm
Convergence Figure 5. The adaptive control strategy based on evolutionary algorithm
applied to the IEEE Type 1 Excitation system.

The optimal gain values of parameters KAopt are

End determined simultaneously by applying the PSO process in the
aim to satisfy the overall transient stability constraint δmax.
We note that the used approach has a decentralized goal which
is the stability enhancement of each generator relatively to the
Figure 4. Standard flowchart of PSO. reference machine. In contrast, KA parameters can be
optimized individually for each machine locally to satisfy a
local stability goal but we expect a poor efficiency behavior by
A. Adaptive control strategy this control action in case of sever perturbation.
The control action is based on the minimization of an
Firstly, the propose strategy was applied on the SMIB to
objective function, which defines the gap between the power
check the implementation of the optimization as control
angle of the ith machine and the power angle of the reference
algorithm. The multi-machine power system is used to
machine as described in (15):
implement the coordination strategy between AVRs. All
t1 generators are considered having the same voltage regulators,
min f   i dt (15) which is not the real case. The stability is improved
t simultaneously by considering the variation of all voltage
regulators in the network at the same time, which allows us to
Subject to: avoid mistuning of one parameter of an individual machine
- Control constraint that can worsen the system transient stability.
min max
KAi  KAi  KAi (16)

CoDIT'18 / Thessaloniki, Greece - April 10-13, 2018 -589-

 using a high gain values determined automatically by the PSO
algorithm. The depicted figures are obtained at the new CCT
- Read power system Data values given by the new proposed strategy and they show that
- Matrices initialization the conventional controller with a fixed gain values will lead to
- Initialize DAE solution instability states at earlier stages compared to the proposed
robust controller based PSO.

t=0

Apply fault

t = t + Δt Time domain
simulation

PSO
t < tsimu Move to the new
best positions

Evaluate Figure 7. Power Angle of generator 2 next bus 1 three phase short circuit
eliminated after 313ms.
objective
function

δ optimal

Stop

Figure 6. Optimale power angle deviation flowchart algorithm using RPSO

controller.

In figures 6, we present the RPSO controllers flowchart

algorithms. The simulations and obtained results are discussed
in the next section.
Figure 8. Power Angle of generator 3 next bus 1 three phase short circuit
eliminated after 313ms.
IV. SIMULATIONS AND RESULTS
In simulations, we have applied a sever disturbance at 1 sec In Fig. 9 and 10, the generators power angles are depicted at
near terminal of generator 1 including the following cases: the CCT limit using a conventional controller. We observe that
the oscillations are reduced and a new steady state operating
1. Simulation with conventional controller and fixed point is obtained with an increase of 20 degrees using the
parameters; robust tuning based PSO controller.
2. Simulation with conventional controller and Robust
variable parameters (Robust self-tuning control);
3. Simulation with conventional controller and adaptive
variable parameters (Adaptive self-tuning control).
A comparison study is performed using different control
strategies in the aim to get the best critical clearing time CCT
to prevent the system instability, as presented in table 1.

A. Robust tuning based on PSO controller:

In Fig. 7 and 8, a comparison of power angle of generator 2
and 3 respectively is presented, regarding the performance of
the adopted strategy based on PSO.
The figures show that the transient instabilities of
generators were eliminated by damping the high oscillations Figure 9. Power Angle of generator 2 next bus 1 three phase short circuit
eliminated after 285ms

CoDIT'18 / Thessaloniki, Greece - April 10-13, 2018 -590-

 TABLE I. CCT COMPARISON TABLE BETWEEN CONVENTIONNEL
VOLTAGE REGULATOR AND CONVENTIONNEL VOLTAGE REGULATOR
INCLUDING INTELLIGENT TUNNING CONTROLLER USING PSO

Critical clearing time (millisecond)

Fault
buses Conventional Enhancement AVR +
AVR+RPSO
AVR ΔCCT APSO
Bus 1 285 313 28 316

Bus 2 197 218 21 220

Bus 3 237 258 21 260

Bus 4 266 292 26 295

Bus 5 322 374 52 379

Bus 6 351 415 64 420

Figure 10. Power Angle of generator 3 next bus 1 three phase short circuit
Bus 7 203 224 21 226
eliminated after 285ms
Bus 8 274 314 40 320

B. Adaptive tuning based PSO controller: Bus 9 232 253 21 255

In Fig. 11 is depicted the optimal gain values KA at each
step time for the considered multi machine test case and Discussion: from table 1 we observe that the best
consequently in Fig. 12, we can see the performance of the improvement is seen at the load buses: 5, 6 and 8, where the
proposed strategy to enhance the transient instability of the fault occurrence probability is higher. In other side, we obtain
system.
a low enhancement of the transient stability margin at the
generation power plant due to the severity of the fault. Load
shedding remains the ideal but not the best way to improve the
transient stability. We can conclude that the proposed
controller is very efficient in damping the oscillations after fast
sever contingencies. In contrast, the power system stabilizer is
more dedicated to damp low oscillations in long term stability
assessment.

V. CONCLUSION
This paper introduces a new nonlinear controller to enhance
the power system transient stability namely robust tuning and
adaptive tuning based particle swarm optimization, (APSO)
and (RPSO) respectively and alternatively based on genetic
Figure 11. AVR_1 Sample values of optimal parameters control for the multi- algorithm namely (RGA) and (AGA). These controllers can
machine power system test case next bus 6 three phase short circuits deal with any unpredictable behavior as the case of the
eliminated after 420ms.
renewable energy generation behavior connected to the grid
which makes the system more resilient to disturbances. The
evolutionary algorithms are very useful to deal with nonlinear
system as the case of transient behavior of power system. The
proposed controllers give the expected results without
exceeding the variables limits. The biggest enhancement is
obtained at the load levels. The use of this strategy can be
considered for training a neural network system to take the
right action at the right moment and for each given situation.
The controller is applied to both SMIB and multi-machine
power systems.

FURTHER WORK
In future work, we will analyze the effect and efficiency of
Figure 12. Power Angle of generator 2 using adaptive tuning based PSO the proposed controllers including their coordination on more
controller next bus 6 three phase short circuit eliminated after 420ms. realistic power systems. Dynamic load models as the case of
induction motors (IM) and the doubly fed induction generator

CoDIT'18 / Thessaloniki, Greece - April 10-13, 2018 -591-

(DFIG) with variable wind profile will be considered.
[12] J. Machowski, J. W. Bialek, S. Robak and J. R. Bumby, “Excitation
ABBREVIATIONS control system for use with synchronous generators,” lEE Proc. Tran.
Dist., Vol.145, No.5, pp. 537-546, Sep. 1998.

CCT: Critical clearing time; [13] https://ics-cert.us-cert.gov/

DAE: Differential algebraic equation; [14] D. Devaraj, and B. Selvabala, “Real-coded genetic algorithm and fuzzy
SMIB: Single machine infinite bus; logic approach for real-time tuning of proportional-integral - derivative
controller in automatic voltage regulator system,” IET Generation,
RPSO: Robust particle swarm optimization controller; Transmission and Distribution, 3(7): 641-649, 2009.
APSO: Adaptive particle swarm optimization controller; [15] R. Chaudhary and al., “A Novel Approach to PID Controller Design for
RGA: Robust genetic algorithm controller; Improvement of Transient Stability and Voltage Regulation of Nonlinear
Power System,” Int. Journal of Electrical and Computer Engineering
AGA: Adaptive genetic algorithm controller; (IJECE), vol/issue: 6(5), pp. 2225-2238, 2016.
IM: Induction machine; [16] N. Farouk, and T. Bingqi, “Application of self-tuning fuzzy PID
DFIG: Doubly Fed Induction Generator; controller on the AVR system,” In Proce. of Int. IEEE Conf. of
Mechatronics and Automation (ICMA), pp. 2510-2514, 2012.
[17] A. Vural, and K. Bayindir, “Transient Stability Enhancement of the
APPENDIX Power System Interconnected with Wind Farm Using Generalized
Unified Power Flow Controller with Simplex Optimized Self-Tuning
TABLE II. PARTICLE SWARM OPTIMIZATION AND GENETIC ALGORITHM Fuzzy Damping Scheme,” Int. Rev. of Elec. Eng. 7(4), 2012.
PARAMETERS [18] S. Wei, Y. Zhou and Y. Huang, "Synchronous Motor-Generator Pair to
Enhance Small Signal and Transient Stability of Power System With
PSO Parameters Values GA Parameters Values High Penetration of Renewable Energy," in IEEE Access, vol. 5, pp.
Maximum number Maximum number 11505-11512, 2017.
of generation
10 of generation
10 [19] H. Mokhlis, J. Laghari, A. Bakar and M. Karimi, “A fuzzy based under-
frequency load shedding scheme for islanded distribution network
Population size 20 Population size 20 connected with DG,” Int. Rev. of Elec. Engineering.7: 49925000, 2012.
Coefficent c1 2 Crossover rate 0.25 [20] D. Flynn, B. W. Hogg, E. Swidenbank and K. J. Zachariah, "Expert self-
tuning control applied to automatic voltage regulation," IEE Colloq. on
Coefficent c2 2 Mutation rate 0.01 Adaptive Controllers in Practice - Part One, London, , pp. 1/1-1/4,
1995.
[21] A. G. Abro and J. Mohamad-Saleh, “A Model Free Estimation Based
REFERENCES Neurocontroller for Synchronous Generator Excitation to Enhance
[1] Y. Wang, G. Guo and D. J. Hill, “Robust decentralized nonlinear Transient Stability,” 2011 Fifth Asia Modelling Symposium, Kuala
controller design for multimachine power systems,” In Automatica, Vol. Lumpur, pp. 151-156, 2011.
33, Issue 9, p1725-1733, 1997. [22] R. Ramya, K. Selvi, and S. Nivethitha, “Optimization of synchronous
[2] Y. Wang, L. Xie, D. J. Hill and R. H. Middleton, “Robust nonlinear generator excitation controller parameters,” IEEE Proc. Int. Conf. of the
controller design for transient stability enhancement of power systems,” Power, Energy and Control (ICPEC), 585-590, 2013.
in Proc. of the 31st IEEE Conf. on Decision and Control, Tucson, AZ, [23] P. P. Kumar, I. Kar and L. Behera, “Variable-Gain Controllers for
vol.1, pp. 1117-1122, 1992. Nonlinear Systems Using the T–S Fuzzy Model,” in IEEE Trans. on
[3] L. Gao, L. Chen, Y. Fan and H. Ma, “A nonlinear control design for Systems, Man, and Cybernetics, Part B (Cybernetics), vol. 36, no. 6, pp.
power systems,” In Automatica, Vol. 28, Issue 5, p975-979, 1992. 1442-1449, Dec. 2006.
[4] G. Gurrala and I. Sen, “A Nonlinear Voltage Regulator With One [24] N. I. Abdul Wahab, and A. Mohamed, “Transient stability emergency
Tunable Parameter for Multimachine Power Systems,” In IEEE Trans. control using generator tripping based on tracking area-based rotor angle
on Power Systems, vol. 26, no. 3, pp. 1186-1195, Aug. 2011. combined with UFLS,” Int. Rev. of Elec. Eng.. 5(5): pp. 2317-2326,
[5] Z. Lan, D. Gan and Y. Ni, “Robust adaptive control design for generator 2010.
excitation,” IEEE Power Engineering Society General Meeting, [25] M. Srinivas and L. M. Patnaik, "Genetic algorithms: a survey," in
Montreal, Que., pp. 5 pp.- , 2006. Computer, vol. 27, no. 6, pp. 17-26, June 1994.
[6] J. W. Park, R. G. Harley and G. K. Venayagamoorthy, “Adaptive-critic- [26] K. F. Man, K. S. Tang and S. Kwong, "Genetic algorithms: concepts and
based optimal neurocontrol for synchronous generators in a power applications [in engineering design]," in IEEE Trans. on Industrial
system using MLP/RBF neural networks,” In IEEE Trans. on Industry Electronics, vol. 43, no. 5, pp. 519-534, Oct 1996.
Applications, vol. 39, no. 5, pp. 1529-1540, Sept.-Oct. 2003. [27] J. Kennedy, R.C. Eberhart, “Particle swarm optimization,” Proc.of the
[7] Y. Guo, D. J. Hill, and Y. Wang, “Global Transient Stability and IEEE Int. Conf. on Neural Networks, USA, 1995, Vol. 4, pp. 1942-
Voltage Regulation for Power Systems,” In IEEE Trans on. Power 1948.
Systems, Vol. 16, N°4, pp. 678-688, 2001. [28] A. Benseddik, M. Hasni, and Al., “~TRANS: A Novel Transient
[8] M. Ouassaid, A. Nejemi, M. Cherkaoui, “A New Nonlinear Excitation Stability Simulation Tool for Large Scale Power Systems,” Inderscience
Controller for Synchronous Power Generator, ” Proc. of the 5th publishers, Int. J. of Power and Energy Conversion, submitted for
WSEAS/IASME Int. Conf. on Electric Power Systems, High Voltages, publication.
Electric Machines, Tenerife, Spain, December 16-18 (pp98-103), 2005. [29] A. Benseddik, and al, "Application of multi-agents system in multi-
[9] G. Fusco and M. Russo, “Nonlinear control design for excitation machine power system transient stability and control,” 3rd Int. Conf. on
controller and power system stabilizer,” In Control Engineering Systems and Control, Algiers, Algeria, Oct. 27-29 2013, pp. 1043-1048.
Practice, Vol. 19, Issue 3, pp. 243-251, 2011. [30] Yao Nan Y., Electric Power System Dynamics, United Kingdom Edition
[10] T. Lahdhiri, A.T. Alouani, “Nonlinear excitation control of a published by .Academic Press, Inc. (London) Ltd, 1983.
synchronous generator with implicit terminal voltage regulation,” In [31] Paul M. Anderson; A. A. Fouad, Power System Control and Stability,
Electric Power Systems Research, Vol. 36, Issue 2, pp.101-112, 1996. 2nd Ed., NJ: USA, Wiley-IEEE Press, 2003.
[11] W. A. Mittelstadt and J. L. Saugen, “A Method of Improving Power [32] P. W. Sauer and M. A. Pai, Power System Dynamics and Stability, NJ:
System Transient Stability Using Controllable Parameters,” in IEEE USA, Prentice Hall, 2005.ch.7.
Trans. on Power Apparatus and Systems, vol. PAS-89, no. 1, pp. 23-27, [33] IEEE Standard, IEEE Recommended Practice for Excitation Systems
Jan. 1970. for Power system Stability Studies, 2005.

CoDIT'18 / Thessaloniki, Greece - April 10-13, 2018 -592-