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

net/publication/323557181

Tuning Controller Parameters for AGC of Multisource Power System using SCA
Algorithm

Conference Paper · February 2017

CITATIONS READS

6 888

2 authors:

Sajjad Asefi Seyed-Jalal Seyed-Shenava

Tallinn University of Technology University of Mohaghegh Ardabili
35 PUBLICATIONS 128 CITATIONS 49 PUBLICATIONS 321 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Sajjad Asefi on 05 March 2018.

The user has requested enhancement of the downloaded file.

 SRPioneers
© Science and Research Pioneers Institute
6th International Conference on Electrical, Computer, Mechanical and Mechatronics Engineering (ICE2017),
9-10 February 2017, Dubai, Emirates

Tuning Controller Parameters for AGC of Multi-

source Power System using SCA Algorithm
SeyedJalal SeyedShenava Sajjad Asefi
Department of Electrical engineering Department of Electrical engineering
University of Mohaghegh Ardabili University of Mohaghegh Ardabili
Ardebil, Iran Ardebil, Iran
[email protected] [email protected]

Abstract— In this paper a new population-based algorithm Automatic Generation Control (AGC). Over the last few
known as Sine Cosine Algorithm (SCA) has applied to decades, different types of generation in each area have
automatic generation control (AGC) of multi-source power been developed considering varieties of models.
system which contains thermal, hydro and gas power plants. And most of the generation in an area is of the same type of
First the algorithm creates multiple initial random solutions
(reheat/non reheat) or hydro generation [2-4, 8]. In real
and by applying a mathematical model based on sine and
cosine functions, makes them to have a wave motion get away
situations, each control area may include different types of
from or closer to the best solution. The proposed algorithm generation such as hydro, thermal, gas, photovoltaic and
has used in a multi-area and multi-source power system which etc. In this investigation, we proposed a model which can
contains DC link between two areas, for tuning the PID be very similar to real-world situation. The results in [1] are
controller that is used for automatic generation control an attempt to study the performance of AGC with thermal,
(AGC). The ability and effectiveness of the proposed algorithm hydro and gas generations in a two area system. So here we
has compared with that of TLBO algorithm and optimal have considered randomly changing load scenario.
output feedback controller tuning performance. The dynamic Previous literature shows that, most of the earlier research
performance of proposed controller has investigated by
in the area of AGC focused only on the optimization of
integral of time multiplied absolute error (ITAE) cost function
and the robustness of the optimized controller has verified by
supplementary control parameters using AI techniques.
its response toward random changing in load and system Only few examine the selection of primary control loop
parameters. It has found that the performance of the proposed parameter R (governor speed regulation parameter) [4]. By
controller is better than that of TLBO optimized controller. considering literature, the importance of the control
parameters such as R and B (frequency bias parameter),
Keywords— Automatic generation control, Multi-source and how they affect the system performance are clearly
power system, HVDC link, Sine Cosine Algorithm (SCA), obvious. Most of these studies are only with hydrothermal
Random load change. systems, they had not discussed the gas generation. This
paper investigates tuning PID controller parameter for each
thermal, hydro and gas generation due to randomly change
I. INTRODUCTION in load and even considers the robustness of controllers for
BULK power systems are often managed by considering deviation in R and B parameters. This paper is an attempt
them as being made up of control areas with to study the AGC of system with thermal-hydro-gas
interconnections between them. A part of automatic generation in each area.
generation control, which is known as automatic load It is to be noted that there has been considerable progress
frequency control has a huge effect on power pool by in intelligent algorithm based controller research work
maintaining scheduled system frequency and scheduled tie- attempting to better control for AGC systems [1]. Sine
line power in normal operation and during slow and small Cosine Algorithm (SCA) [5] is a new optimization
perturbations [1]. Each control area must meet its own technique and it has not been applied to tune the controller
demand and its scheduled interchange power. Monitoring in AGC studies, yet. In this work, we have applied this
deviation in frequency, we can find out any mismatch optimization algorithm for tuning of controller for load
between the generation and load [2]. This balancing frequency control of the present scenario above realistic
between load and generation can be achieved by using power systems having multiple sources of power
 SRPioneers
© Science and Research Pioneers Institute
6th International Conference on Electrical, Computer, Mechanical and Mechatronics Engineering (ICE2017),
9-10 February 2017, Dubai, Emirates

generation. Our investigation contains two phases. First we 4. To see the performance of SCA, regarding to randomly
have compared the SCA algorithm dynamic performances changing load and comparing with TLBO.
to that of recently published TLBO tuned controller [1] for 5. To carry out the sensitivity analysis for the proposed
the same AGC system due to a step change in demanded SCA-PID controller with its optimum parameters and its
power of area 1. In the second phase, we have tuned robustness to wide changes in system parameters from their
controller parameters via SCA algorithm while considering nominal values and different cost functions.
the randomly changed load. At the end of phase 2, we have
investigated the robustness of our controller by huge II. SYSTEM UNDER INVESTIGATION
deviations in system parameters such as B and R.
However, the proposed SCA is very simple in concept and A. Multi Source Power System
easy implementation to tune the controller of AGC for the The system under study is like a realistic power system
realistic power system. Here, a brief overview of SCA is which includes reheat thermal, hydro and gas generating
provided. So, for the sake of completeness and better units. The linearized model of single area system is given
reading of the paper we encourage the readers to refer to in references [6,7] and multi area system transfer function
[5] for more details of SCA and its application. The schematic is provided in Fig. 1 and detailed features with
roadmap we have followed in this work is presented below: nominal parameters of the system are provided in [1] for
simulation and AGC study. As it can be seen in Fig. 1 the
1. To propose a new mathematical model based on sine and load in area-1 has been applied to change randomly. While
cosine functions algorithm for the load frequency control of in this investigation for area-2 we have no change in load.
the realistic power system. It is to be noted that the participation rate for each
2. To optimize the PID controller gains and study of its generator is different, and in a control area summation of
dynamic performances for power system. participation factors of all participating generators is equal
to unity. Regulation parameter of each unit is different
3. To compare the dynamic performance of SCA based PID from another and each unit has its own participation factor
controller to that of TLBO tuned controller for multi-source which decides the contribution to the nominal loading.
multi area power system without and with HVDC link.

1/R1
B1 1/R2 Kdc
Delta PD1
1/R3 Tdc × s + 1
- 1
1 Kr × Tr × s + 1
Tsg × s + 1 Trg × s + 1 Ttg × s + 1 KT Kps
+ -
PID - Tps × s + 1
+
-
- - 1 Tr × s + 1 −Tws
×s +1 +
PID Tgh × s + 1 KH +
Trh × s + 1 0.5×Tws×s +1 +
- +

PID - Xc × s + 1 1
1 − Tcr × s + 1 KG
bg × s + cg Yc × s + 1 Tfc × s + 1 Tcd × s + 1
+
2 × π × T 12 +

Tdc × s + 1 -

a12 Kdc
a12 Tdc × s + 1
- 1
1 Kr × Tr × s + 1
Tsg × s + 1 Trg × s + 1 Ttg × s + 1 KT Kps
+ -
PID - Tps × s + 1
+
-
- - 1 Tr × s + 1 −Tws
×s +1 +
PID Tgh × s + 1 KH +
Trh × s + 1 0.5×Tws×s +1 +
- +

B2 PID - Xc × s + 1 1
1 − Tcr × s + 1
KG
bg × s + cg Yc × s + 1 Tfc × s + 1 Tcd × s + 1
+
Delta PD1

B2
B2
B2

Fig. 1. Transfer function model of multi-source multi area with HVDC link
 SRPioneers
© Science and Research Pioneers Institute
6th International Conference on Electrical, Computer, Mechanical and Mechatronics Engineering (ICE2017),
9-10 February 2017, Dubai, Emirates


   
B. HVDC Link  (4)
The two area power system interconnected by parallel 
AC–DC tie lines has simulated using MATLAB Simulink. And the applied objective function in this study is:
The simulated system comprises combination of generating

units such as reheat thermal, hydro and gas units in each     |∆ |  |∆ | (5)
area. Furthermore, the participation rate for each 
participating generator is not same. So, the generators in  |∆  |
each area may or may not participate in the LFC task. As where ts is total simulation time.
mentioned above the summation of participation factors of
all participating generators is equal to unity in a control
area. Transfer function model of multi-source multi area D. Random Change in Load
with HVDC link with integral controllers is presented in In this paper other than commonly used step change in
[1]. load, randomly change in load has used. Fig. 2 shows the
way load has changed during our simulation.

C. Control Strategy and Objective Function random load change

0.02
When designing a controller, the main purpose of
controller must be foremost in all considerations. This can 0.015
include system dynamics, robustness to model uncertainty,
ability to follow the set point, sensitivity to measurement 0.01

noise and attenuation of load disturbances. The Proportional

Integral Derivative controller (PID) is the widespread and 0.005
∆PD (pu)

popular feedback controller used in many modern

industries. The popularity of PID controllers stems in part 0

to their wide applicability to a variety of single input single

-0.005
output applications. PID controller is often used when
stability and fast response are required. The PID controller
-0.01
is employed in the present paper for comparative
performance analysis of similar power system. There are -0.015
four kinds of performance criteria, generally considered in 0 5 10 15 20 25 30 35 40
time (sec)
the control design. These are the integral of absolute error
Fig. 2. Random load change
(IAE), integral of squared error (ISE), integral of time
weighted squared error (ITSE) and integral of time
multiplied absolute error (ITAE). However, ITAE and ISE III. PROPOSED ALGORITHM
criterions are mostly used in AGC studies for their better
performance as compared to ISTE and IAE criterion. Both In this paper a novel population-based optimization
to compare SCA with TLBO and further investigations we algorithm called Sine Cosine Algorithm (SCA) for tuning
PID controller parameters has used. After creating multiple
just have used ITAE due to its quick settling feature. Below
initial candidate solutions randomly by SCA, then by using
the equations related to different error criteria and even the
a mathematical model based on sine and cosine functions,
objective function we have used is provided.
algorithm requires them to fluctuate outwards or towards
the best solution. As mentioned above, the proposed
 algorithm utilizes the mathematical functions, i.e. sine and

   | | (1)
cosine, to explore and exploit the space between two


solutions in the search space while trying to find better

    | | (2) solutions. As it is well-known for population-based
 optimization techniques, the optimization process starts
 with a set of random solutions. By using an objective

     (3) function these solutions are evaluated repeatedly and
 improved by a set of equations that are the foundation of an
optimization technique. There is no guarantee of finding a
 SRPioneers
© Science and Research Pioneers Institute
6th International Conference on Electrical, Computer, Mechanical and Mechatronics Engineering (ICE2017),
9-10 February 2017, Dubai, Emirates

solution in a single run, because of stochastic feature of change in load of area one and compare the results with the
population based optimization techniques while looking for obtained results in [1].
the optima of optimization problems. However, we can
increases the probability of finding the global optimum by A. Comparing TLBO and SCA due to Step Load Change
applying enough number of random solutions and In this section simulations are carried out to show the
optimization steps (iterations). Although there might be comparison of TLBO and SCA algorithm regarding 1
differences between algorithms in the field of stochastic percent step change in the load of area one. The parameters
population-based optimization, commonly the optimization obtained by these two algorithms are provided in table 1.
process is divided into two different phases: exploration and Parameters KPi, KIi and KDi indicate the ith PID controller’s
exploitation. By sudden combining random solutions in the proportional, integral and differential coefficients,
set of solutions with a high rate of randomness an respectively. Because the two areas are similar, we have
optimization algorithm finds the promising regions of the provided the parameters just for area 1.
search space. This called exploration phase. In the next
phase, i.e. exploitation, however, there are gradual changes Table 1. Parameters used due to each algorithm
in the random solutions, and random variations are
Parameters TLBO[1] SCA
considerably less than those in the exploration phase [5].
The following position updating equations are applied for KP1 5.0658 9.7744
both phases, in this algorithm:
KI1 3.9658 1.3873
! "  !  #  sin #  |#' ( ! | (6)
K D1 2.4170 4.1195
! "
 !  #  cos #  |#' ( ! | (7) KP2 0.7032 9.8525
Where ! is the position of the current solution in i-th KI2 0.0220 0.0131
dimension at t-th iteration, r1, r2 and r3 are random
numbers,  is position of the destination point in i-th
K D2 0.0264 0.0132
dimension. The above mentioned equations are combined to KP3 8.7211 0.2037
be used as follows: KI3 7.4729 10
! "
!  #  sin #  |#' ( ! |,
K D3 2.4181 0.131
#- . 0.5
+ 
(8)
!  #  cos #  |#'  ( ! |, #- 2 0.5 The simulation results obtained by using parameters of
where ts is a random number in [0,1]. TLBO and SCA provided in table 1 demonstrated in Fig.'s
3-5.
Following steps can be introduced as optimization
procedure for the algorithm:
-3
x 10
Step1: initialize a set of search agents (solutions) (X) 1

0
Step2: evaluate each of the search agents by the objective
function -1

-2
Step3: update the best solution obtained so far (P=X*)
-3
Step4: update r1, r2, r3, and r4
∆ f1 (Hz)

-4

Step5: update the position of search agents using Eq. (8) -5

Step6: if terminations criteria aren’t met go to step2, else -6

return the best solution obtained so far as the global -7

optimum. -8 SCA
TLBO
-9
IV. RESULTS AND SIMULATION 0 5 10 15 20 25 30 35 40
time (sec)
In the following sections, first we are going to compare
TLBO and SCA algorithms in respect to 1 percent step Fig. 3. Frequency deviations of the control area-1 by using PID
controller for 1% change in area 1with AC-DC parallel tie line
 SRPioneers
© Science and Research Pioneers Institute
6th International Conference on Electrical, Computer, Mechanical and Mechatronics Engineering (ICE2017),
9-10 February 2017, Dubai, Emirates
-4

4
x 10 As the results obtained in table 2 shows SCA is faster
2
and this is one of the reasons we have chosen the SCA
algorithm for rest of our study.
0
Table 2. Comparing best run of both algorithms
-2

-4 Time
No. of iter. No. of pop. Cost
(sec)
∆ f2 (Hz)

-6

-8 SCA 100 30 ≈ 440 1.12

-10

-12 TLBO 100 30 ≈ 1400 1.12

-14 SCA
TLBO
-16
0 5 10 15 20 25 30 35 40
time (sec)
B. System Response for Random Load Change
Fig. 4. Frequency deviations of the control area-2 by using PID In this section we have provided the change of frequency of
controller for 1% change in area 1with AC-DC parallel tie line both areas and tie line power while load changes randomly.
x 10
-4 Note that the profile of random load change is provided in
2
figure 2. Applied PID coefficients which obtained using
SCA, are presented in table 3.
0
0.015

0.01
-2

0.005
∆P tie (pu)

-4
0
∆ f1 (Hz)

-0.005
-6

-0.01

-8
-0.015
SCA
TLBO
-0.02
-10
0 5 10 15 20 25 30 35 40
time (sec) -0.025
0 5 10 15 20 25 30 35 40
time (sec)
Fig. 5. Tie line power deviation by using PID controller for 1%
change in area-1 with AC-DC parallel tie line Fig. 6. Frequency deviations of the control area-1 by using PID
controller for 1% change in area 1with AC-DC parallel tie line
As it is obvious from the above results, the proposed and random load change
algorithm gets slightly better results than TLBO. It is worth x 10
-3

3
noting that by increasing the number of iteration and
population of algorithms they may lead to same answers, 2

but the main feature is the speed in which the algorithms 1

reach to the best solution. We have examined both TLBO
0
and SCA algorithm 10 times and at each run we have set
∆ f2 (Hz)

the iteration 100 by 30 number of population. The results of -1

comparison for the best cost obtained amongst 10 run of -2

both algorithms are provided in table 2. Table 2 shows that
-3
both SCA and TLBO converge to same amount for
objective function but there would be a difference of about -4

factor 3 in the time elapsed for the best answer. It is to be -5

0 5 10 15 20 25 30 35 40
noted that we have used same computer system for this time (sec)

comparison.
Fig. 7. Frequency deviations of the control area-2 by using PID
controller for 1% change in area 1with AC-DC parallel tie line
and random load change
 SRPioneers
© Science and Research Pioneers Institute
6th International Conference on Electrical, Computer, Mechanical and Mechatronics Engineering (ICE2017),
9-10 February 2017, Dubai, Emirates
-3 -3
x 10 x 10 Change in frequency bias parameter (B)
1.5 4

1
2
0.5

0
0
∆ Ptie (pu)

∆f2 (Hz)
-0.5
-2

-1

-4
-1.5
zero
+%25
-2 -6 -%25
+%50
-2.5 -%50
0 5 10 15 20 25 30 35 40
-8
time (sec) 0 5 10 15 20 25 30 35 40
time (sec)
Fig. 8. Tie line power deviation by using PID controller for 1%
change in area-1 with AC-DC parallel tie line and random load Fig. 10. Frequency deviation of area-2 using optimized PID
change controller and change in B
-3 Change in frequency bias parameter (B)
x 10
From Fig.'s 6 – 8 it can be understood that the system can 1.5

easily damp each change occurs in load and never let 1

system to be unstable. 0.5

C. Analysis of Robustness 0

To study the robustness of power system to wide changes

∆Ptie (pu)

-0.5

in the system parameters such as frequency bias parameter -1

(B) of both areas and Governor speed regulation parameter -1.5

(R) of both areas simultaneously are varied in the range of zero
-2
+50% to –50% from their nominal values in steps of 25% +%25
-%25
taking one at a time. The dynamic behavior of the system -2.5 +%50
-%50
with SCA optimized PID controller for 1% step load -3
0 5 10 15 20 25 30 35 40
changes in area-1 at t = 0 s for the above parameters time (sec)

variation are shown in figures 9–11. It can be observed that Fig. 11. Tie line power deviation of area-1 using optimized PID
the proposed strategy provides a robust and stable control controller and change in B
satisfactorily at nominal parameters and need not be reset Change in governor speed regulation parameter (R)
for wide changes in the system parameters. 0.015

0.01
Change in frequency bias parameter (B)
0.015
0.005
0.01
0
0.005
∆f1 (Hz)

-0.005
0

-0.005 -0.01
∆f1 (Hz)

-0.01 -0.015 zero

+%25
-0.015 -%25
-0.02
+%50
zero -%50
-0.02
+%25 -0.025
-%25 0 5 10 15 20 25 30 35 40
-0.025 time (sec)
+%50
-%50
-0.03
0 5 10 15 20 25 30 35 40
Fig. 12. Frequency deviation of area-1 using optimized PID
time (sec) controller and change in R
Fig. 9. Frequency deviation of area-1 using optimized PID
controller while changing B
 SRPioneers
© Science and Research Pioneers Institute
6th International Conference on Electrical, Computer, Mechanical and Mechatronics Engineering (ICE2017),
9-10 February 2017, Dubai, Emirates
-3 Change in governor speed regulation parameter (R)
3
x 10
V. CONCLUSION
2 In this paper a real-world power system containing two
areas which include three different types of generating
1
units, i.e. gas, reheat thermal and hydro is formulated as an
0
optimization problem to automatic generation control
∆f2 (Hz)

-1 (AGC). A new algorithm is introduced and applied for the

optimization problem. The tow area power system is
-2
simulated by MATLAB/Simulink while subjected to
zero
-3
+%25
randomly changing in load. From the results, it is clearly
-4
-%25
+%50
obvious that the algorithm works well in solving the
-%50 optimization problem by providing reasonable coefficients,
-5
0 5 10 15 20
time (sec)
25 30 35 40
sooner than the time needed for the algorithms applied in
the literature. And the robustness of controlling system, due
Fig. 13. Frequency deviation of area-2 using optimized PID to change in system parameters such as B and R is
controller and change in R
confirmed while using the obtained controller parameters
-3 Change in governor speed regulation parameter (R)
x 10
1.5 by SCA.
1
REFERENCES
0.5
[1] A. Barisal, "Comparative performance analysis of teaching learning
0 based optimization for automatic load frequency control of multi-
source power systems", International Journal of Electrical Power &
∆Ptie (pu)

-0.5 Energy Systems, vol. 66, pp. 67-77, 2015.

-1
[2] H. A. Shayanfar, H. Shayeghi, and A. Younesi, "Optimal PID
-1.5 zero controller design using Krill Herd algorithm for frequency
+%25
-%25
stabilizing in an isolated wind-diesel system," In proceedings on the
-2 international conference on artificial intelligence (ICAI). The
+%50
-%50 steering committee of the world congress in computer science,
-2.5
0 5 10 15 20 25 30 35 40 computer engineering and applied computing (WorldComp), 2015.
time (sec)

Fig. 14. Tie line power deviation of area-1 using optimized PID [3] M. Carmeli, F. C. Dezza, M. Mauri, G. Marchegini and D. Rosati,
controller and change in R "Control strategies and configurations of hybrid distributed
generation systems," Rene Energy, vol. 41, pp. 294-305, 2012.
It can be seen from the above figures that the system has
high robustness due to change in both parameters B and R. [4] K. K. Challa, and N. P. Rao, "Analysis and design of controller for
two area thermal-hydro-gas AGC system." Power Electronics,
Table 3. Parameters used by SCA algorithm due to random Drives and Energy Systems (PEDES) & 2010 Power India, 2010
change in load Joint International Conference on. IEEE, 2010.

Parameters SCA [5] S. Mirjalili, "SCA: A sine cosine algorithm for solving optimization
problems",Knowledge-Based Systems, vol. 96, pp. 120-133, 2016.
KP1 10
KI1 10 [6] K. Parmar, S. Majhi and D. Kothari, "Load frequency control of a
realistic power system with multi-source power
K D1 3.8739 generation", International Journal of Electrical Power & Energy
Systems, vol. 42, no. 1, pp. 426-433, 2012.
KP2 10
[7] B. Mohanty, S. Panda and P. Hota, "Controller parameters tuning of
KI2 0.01 differential evolution algorithm and its application to load frequency
control of multi-source power system",International Journal of
K D2 0.0167 Electrical Power & Energy Systems, vol. 54, pp. 77-85, 2014.
KP3 10
[8] O. Abedinia, N. Amjady, A. Ghasemi, and H. Shayeghi, "Multi-stage
KI3 9.9897 fuzzy load frequency control based on multiobjective harmony
search algorithm in deregulated environment", Journal of operation
K D3 0.8432 and automation in power engineering, vol. 1, no. 1, pp. 63-73, 2013.

View publication stats