Ekkawit THAOKEAW1, Krittanon PRATHEPHA1, Jagraphon OBMA2, Worawat SA-NGIAMVIBOOL1

Faculty of Engineering, Mahasarakham University, Mahasarakham 44150, Thailand (1)

Faculty of Engineering, Rajamangala University of Technology ISAN Khonkaen Campus, Khonkaen 40000, Thailand (2)
ORCID: 1. 0000-0001-5796-3260; 2. 0000-0002-4365-9724; 3. 0000-0001-8984-2287; 4. 0000-0001-8580-8962

doi:10.15199/48.2023.02.17

Design of Parallel 2-DOF PID Controller by Bee Algorithm for

Interconnected Thermal Power Systems
Abstract. The article presents the design and analysis of the control of two connected thermal power generator systems by the parallel 2-degree
freedom of proportional-integral-derivative (2-DOF-PID) and modified by the bee algorithm (BA). In general, power and control systems are generally
characterized as non-linear, so the automatic gain controllers (AGCs) will need to be optimized for parameters, particularly the 2-DOF-PID
controllers that use a large number of parameters. The bee algorithms and differential evolution (DE) have been presented to improve
parameterization and compare the performance of such an algorithm while using a dynamic electric load size change approach. As a result of
testing, when the bee algorithm is used to improve the parameters of the generator system, the system will be more robust to changes and have
higher performance.

Streszczenie. W artykule przedstawiono projekt i analizę sterowania dwoma połączonymi układami generatorów energii cieplnej poprzez równoległą
dwustopniową swobodę proporcjonalno-całkująco-różniczkową (2-DOF-PID) i zmodyfikowaną przez algorytm pszczół (BA). Ogólnie rzecz biorąc,
systemy zasilania i sterowania są generalnie charakteryzowane jako nieliniowe, więc automatyczne regulatory wzmocnienia (AGC) będą musiały
zostać zoptymalizowane pod kątem parametrów, w szczególności regulatory 2-DOF-PID, które wykorzystują dużą liczbę parametrów. Algorytmy
pszczół i ewolucja różnicowa (DE) zostały zaprezentowane w celu poprawy parametryzacji i porównania wydajności takiego algorytmu przy użyciu
podejścia do dynamicznej zmiany wielkości ładunku elektrycznego. W wyniku testów, gdy algorytm pszczeli zostanie wykorzystany do poprawy
parametrów systemu generatora, system będzie bardziej odporny na zmiany i będzie miał wyższą wydajność. (Projektowanie równoległego
regulatora PID 2-DOF według algorytmu Bee dla połączonych systemów elektroenergetycznych)

Keywords: PID controller, .Bee algorithm

Słowa kluczowe: sterownik PIDF, algorytm Bee..

Introduction solution and may avoid a particular solution. This is

The importance of the design and control of the power because the best chromosomes might change, increasing
system is load frequency control and power. Specifically, a the likelihood of obtaining the best chromosome or the
power system where multiple electrical power sources are optimal solution [11]. However, the genetic process is quite
connected and each of which must have the same complicated in order to discover the optimal solution, which
frequency. The core principle of controlling the load requires more time and may not find the optimal solution
frequency in a power system is to maintain the generator within a specific time frame. The annealing simulation
frequency at each source and to manage the power flow algorithm can solve any problem because it is based on the
adequately to satisfy the requirements of the load in both Boltzmann diffusion principle, which shrinks the size or
normal and dynamic states of change. boundary of the source. Nonetheless, a significant issue
In general, a good power system's controller design for with the annealing simulation algorithm is that energy re-
a generator must be able to handle changes in the load, randomization employs only surrounding sets of ideal
and the voltage and frequency must be high quality [1]. In energy values, resulting in an inescapable local solution.
this literature study, electrical control systems for simple Randomization is another significant issue. The initial
electrical systems are proposed [2]. Subsequently, optimal response with a restricted number of randomizations
control was applied for the first time for the design and resulted in a lengthy time to identify the optimal response
control of electrical systems within the connected power when the initial response was far from the optimal
system of a power plant in order to optimize the optimal response. Due to the initial randomization issue when the
control application. This relies on the different control first response is far from the optimal answer and the initial
techniques for designing electrical automation controllers. It unpredictability with a limited number of random replies, it
focuses on the development of automatic electrical takes a long time to discover the optimal answer, and it may
controllers with the application of modern controls to not be possible to get the optimal solution. As there was no
improve electrical automation controllers such as neural perfect solution, a heuristic optimization approach was
networks [5-6], fuzzy system theory [7], reinforcement created to connect the live behavior of an organism to its
learning [8], ANFIS approach [9-10], and particle swarm natural path by generating random beginning responses.
optimization [11], etc. In addition to the development of the There are an infinite number of prototypes, like the
design and application of the optimization of electrical algorithm for the Bee Optimization Method, that can find
controllers, there have been advancements from basic to solutions faster and avoid certain answers.[13-16]
Parallel 2-Degree Freedom of Proportional-Integral- This paper presents the application of the bee algorithm
Derivative (2-DOF-PID) electrical controllers for electrical optimization for the parameterization of the 2-DOF PID
systems with non-linear characteristics. However, very few Controller in the interconnected power system of the
optimization techniques are used for the design of 2-DOF- combined cycle power generator under non-linear features.
PID controllers, especially for power systems It compares the dynamic response based on the dynamic
interconnected between combined cycle power plants. performance, the integral of squared error (ISE), the integral
Heuristic optimization is popular for solving non-linear of time squared error (ITSE), and the sensitivity analysis of
problems by comparing the living behaviors of organisms, the internal controls under abnormal conditions of the power
such as the bee swarm method, ant swarm method, particle system.
swarm method, and genetic method. Genetic algorithms
were the first to be developed; this algorithm could always Interconnected Termal Power System
find a solution to a problem, despite the fact that the initial The purpose of thermal power system connections is to
chromosomal randomization was far from the optimal provide electrical stability and stability. In general, the

104 PRZEGLĄD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 99 NR 2/2023

connection between two combined cycle power plants is combined-cycle power plants, each of which consists of a
simulated as seen in Figure 1, with each source consisting governor with deadband, turbine, and power system
of a generator rated at 2000 MW and supplying 1000 MW to
load. Figure 1 shows a simulation of the connection of two

Fig.1. Transfer function model of interconnected two-area thermal system with governor dead band

THE 2-DOF PID Controller 2. Objective Function

The connection of thermal power plants to the power The structure of the Generally, the primary objectives
system increases the stability of the power system. It of system control in power systems are:
requires control, especially from a 2-DOF-PID controller. In 1.) The frequency difference in each power system must
this topic, controller structure and objective functions are approach zero when the load changes.
presented, respectively. 2.) The integral of the frequency error must be minimal.
3.) The control system characteristics should maintain
the stability of the system.
4.) The power system in each source must be able to
work both when the load is normal and when it changes.
From the above, control objectives can be written into
equations as follow:

(1) Minimize (Ji)

tsim
(2) J1 (ISE) = න ൣ(f1 )2 +(f2 )2 +(PTie )2 ൧ dt
0

J2 (ITSE) = ‫׬‬0 sim ൣ(f1 )2 +(f2 )2 +(PTie )2 ൧ tdt

t
(3)

The objective function must be executed under the

Fig.2. Two degrees of freedom (2-DOF) PIDcontroller maximum (maximum, max) and minimum (minimum, min)
conditions of the parameters. In a parallel 2-DOF PID
1. Controller structure controller presented as shown in Figure 2, it consists of
The structure of the parallel 2-DOF PID controller is proportional setpoint weight (PW), proportional gain (KP),
shown in figure 2. The parallel 2-DOF PID controller integrator gain (KI), derivative setpoint weight (DW),
consists of proportional setpoint weight (PW), proportional derivative gain (KD), and derivative filter coefficient (N) as in
gain (KP), integrator gain (KI), derivative setpoint weight Equations 4 to 9 and Table 1.
(DW), derivative gain (KD), and derivative filter coefficient
(N), respectively. The R (s) is the reference signal and the Y (4) KP minKPKP max
(s) is the feedback signal of the system output when U (s) is (5) KI minKIKI max
the output signal of the parallel 2-DOF PID controller. (6) KD minKDKD max

PRZEGLĄD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 99 NR 2/2023 105

(7) Pw minPwPw max Step 10: The best 2-DOF PID controller parameter is
(8) Dw minDwDw max the one that makes each cycle's fitness value the best it can
be.
(9) N minNN max
Results And Discussion
Table 1. Minimum and Maximum Value of The Control Parameters The connection of thermal power plants to the power
Controller Parameters Minimum Maximum system increases the stability of the power system. It
KP 0 1 requires control, especially from a 2-DOF-PID controller. In
KI 0 1 this topic, controller structure and objective functions are
KD 0 1 presented, respectively.
N 10 300 The bee algorithm modifies the settings of the 2-DOF
PW 0 2 PID controller, which controls the connection of the power
DW 0 5
system from two sources. All processes are tested and
assessed using the MATLAB R2013a program. The
The goal of the bee algorithm (BA) is to emulate the program runs on a CPU Corei3 of 2.40 GHz and RAM of
behavior of honey bees as they forage for food in their 4.00 GB. Comparing the results of parameter tuning for 2-
natural environment by communicating with other bees for DOF PID controllers using the bee algorithm and the
instructions and food sources, such as the waggle dance, differential evolution (DE) algorithm with the same settings
which conveys the location of food sources. The bee is shown in Table 3.
algorithm's easy-to-understand approach to communication
has led to the development of artificial intelligence Table 3. Optimization Controller Parameters.
algorithms for solving complex problems with a lot of Objective Differential Evolution Bee Algorithm
variables, like building control systems with optimized function
J1 J2 (ITSE)
parameters. The application of the bee algorithm to adjust /Controller J1 (ISE) J2 (ITSE)
(ISE) [12] [12]
the settings and design the parameters of the 2-DOF PID parameters
controller in order to regulate the power connection KP 0.5409 0.4935 0.6939 0.5285
KI 0.9708 0.7619 1.0793 1.0184
between two sources. It uses the parameters in table 2.
KD 0.5144 0.3007 0.5619 0.4344
N 180.6983 177.4023 152.3835 163.6528
Table 2. Parameters of Bee Algorithm PW 2.0832 0.5997 0.7547 1.1718
n Number of scout bee 10 DW 0.6462 2.5641 0.3238 0.7075
m Number of sites selected for neighborhood 5
search
e Number of best “elite” sites out of m selected 2 Table 3 shows the best settings for the 2-DOF PID
sites controller. These settings were made by the bee algorithm
nep Employed bees for e best sites 4 so that the integral of squared error (ISE) and the integral of
nsp Employed bees for m-e size 2 time squared error would be as low as possible. The
ngh Neighborhood search 0.1 Proportional gain (KP), Integrator gain (KI), and Derivative
NC Number of iterations 30 gain (KD) parameters were found to be close to those
adjusted by the differential evolution (DE) algorithm, with
The bee algorithm includes the following steps: the exception of the Derivative filter coefficient (N),
Step 1: Determine the variables in the bee algorithm as Proportional setpoint weight (PW), and Derivative setpoint
shown in Table 2. weight (DW), which were less than those adjusted by the
Step 2: Pollinating bee populations (number of n) were DE algorithm.
chosen at random to find the parameters of the 2-DOF PID
-3
controls according to Table 1 under rules and conditions. 5
x 10
Step 3: Test the power system connection control from
two sources under the parameters of the 2-DOF PID control
from step 2.
0
Step 4: Use equations (2) and (3) to figure out the
fitness value so that you can get n fitness values.
 f1 (Hz)

Step 5: Choose the number of m bee populations that

-5
yields the best fitness value. The m bee populations were
split into e groups made up of the e bee populations with
the highest fitness values. The bee population surveyed the
-10
number of n = m - e. J1 : DE 2-DOF PID [11]
Step 6: Assign nep bees worker to find the parameter of
J1 : BA 2-DOF PID
the 2-DOF PID control, the surrounding bee survey e with a
-15
fixed distance of ngh, and set the nsp bee worker to find the 0 5 10 15
parameter of the 2-DOF PID control. Surrounding of n = m - Time (Sec)
e survey bees with fixed spacing ngh The parameters of the Fig.3. Frequency deviation of area-1 for 1% change in area-1 with
2-DOF PID controls are given in Table 1 under various ISE objective function.
nep+nsp set conditions and regulations.
Step 7: Test the power system connection control from Figures 3 and 4 compare the difference in the frequency
two sources under the parameters of the 2-DOF PID control change of the generator systems 1 and 2 in a two-source
from step 6. power connection control system. The parameters of the 2-
Step 8: Evaluate the fitness value according to DOF PID controller were designed with the integral of
equations (2) and (3) to get nep + nsp fitness values. squared error (ISE) at its lowest value by the bee algorithm
Choose the best fitness value in this round. and a differential evolution (DE) algorithm. The conditional
Step 9: Return to step 2 until the specified number of is a dynamic change by a 1% change in the load connected
repetitions has been completed. to the generator systems 1 and 2. Figures 3 and 4 show

106 PRZEGLĄD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 99 NR 2/2023

that the 2-DOF PID control with parametric tuning of the the bee algorithm gives greatest control over both the
bee algorithm gives greatest control over both the overshoot and the transition to the steady state. Figure 8
overshoot and the transition to the steady state. Figure 5 shows the difference in electrical power (〖 "P" 〗_"Tie" )
shows the difference in electrical power in the connection of in the connection of the power system from two sources. It
the power system from two sources. It was found that the 2- was found that the 2-DOF PID controller parameterized by
DOF PID controller parameterized by the bee algorithm had the bee algorithm had a more efficient power differential
a more efficient power differential control effect than the control effect than the differential evolution (DE) method.
differential evolution (DE) method -3
-3 x 10
x 10
1 1

0
0

-1

 f2 (Hz)
-1
 f2 (Hz)

-2

-2 -3

-4 J2 : DE 2-DOF PID [11]

-3 J1 : DE 2-DOF PID [11]
-5 J2 : BA 2-DOF PID
J1 : BA 2-DOF PID
-4 0 5 10 15
0 5 10 15
Time (Sec)
Time (Sec)
Fig.4. Frequency deviation of area-2 for 1% change in area-1 with Fig.7. Frequency deviation of area-2 for 1% change in area-1 with
ISE objective function. ITSE objective function.
-3 -3
x 10 x 10
0 0

-0.5
-0.5
 Ptie (P.U.)
 Ptie (P.U.)

-1
-1

-1.5
-1.5

J1 : DE 2-DOF PID [11] -2 J2 : DE 2-DOF PID [11]

-2 J1 : BA 2-DOF PID J2 : BA 2-DOF PID
-2.5
0 5 10 15 0 5 10 15
Time (Sec) Time (Sec)
Fig.5. Tie line power deviation for 1% change in area-1 with ISE Fig.8. Tie line power deviation for 1% change in area-1 with ITSE
objective function objective function.
-3
x 10
Also, when comparing the convergent sequence to the
2
least differential of the integral of squared error (ISE, J1)
0 and integral of time squared error (ITSE, J2), the
-2 parameterization of 2-DOF controls PID for controlling
power connection from 2 sources. It was found that the
-4
design of the 2-DOF PID controller with the objective
 f1 (Hz)

-6 function integral of time squared error (ITSE, J2) has the

-8 least value by parameter tuning by the swarm algorithm
showed good convergence to the answer as shown in Fig.
-10
9.
J2 : DE 2-DOF PID [11] -3
-12 x 10
J2 : BA 2-DOF PID 1
-14 J1 : BA 2-DOF PID
0 5 10 15 J2 : BA 2-DOF PID
Time (Sec) 0.8
Fig.6. Frequency deviation of area-1 for 1% change in area-1 with
ITSE objective function.
0.6
Error

Figures 6 and 7 compare the difference in the frequency

change of the generator systems 1 and 2 in a two-source 0.4
power connection control system. The parameters of the 2-
DOF PID controller were designed with the integral of time 0.2
squared error (ITSE) at its lowest value by the bee
algorithm and a differential evolution (DE) algorithm. The
conditional is a dynamic change by a 1% change in the load 0
0 5 10 15 20 25 30
connected to the generator systems 1 and 2. Figures 6 and Iteration
7 show that the 2-DOF PID control with parametric tuning of Fig.9. Convergence curves

PRZEGLĄD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 99 NR 2/2023 107

 Table 4 displays the controller performance the controller. This results in complicated power system
parameterized by the bee method and the differential management using a 2-DOF PID controller capable of
evolution (DE) algorithm with the objective of minimizing the controlling frequency and power effectively. For the goal,
integral of squared error (ISE, J1) and integral of time the integral of squared error and the integral of time
squared error (ITSE, J2). Table 4 displays the following squared error are the smallest, and the dynamic change
performance indicators: peak undershoot, settling time and conditions of the thermal power generator system are linked
errors of f1, f2 and Ptie. From Table 4, for all control to two power sources.
purposes, controllers parameterized by the bee algorithm In addition, the bee approach is used to construct and
provided better control performance in all aspects than parameterize the 2-DOF PID controller. Compared to
controllers parameterized by the differential evolution (DE) controllers configured using the differential evolution (DE)
algorithm algorithm, it provides superior control over all aspects of
In complicated power system management, controller complicated power systems.
design issues are a major problem. Specifically, the 2-DOF
PID controller with high performance. Utilizing bee
algorithms allows design by optimizing the parameters of

Table 4. Values of Settling Times, Peak Overshoot,and Errors

Optimization/ Peak overshoot (OS) Settling times (2% band) Ts (s)
Controllers f1 f2 Ptie f1 f2 Ptie Errors
-5
J1 DE : 2-DOF PID [11] 0.0032 0.0010 0.0000 7.6676 7.6676 7.6318 8.9321 x 10
-5
J1 BA : 2-DOF PID 0.0030 0.0006 0.0000 5.2110 6.0338 6.1814 8.4822 x 10
-4
J2 DE : 2-DOF PID [11] 0.0025 0.0011 0.0000 7.6854 7.7748 7.6318 1.0116 x 10
-5
J2 BA : 2-DOF PID 0.0028 0.0006 0.0000 4.7012 5.6296 6.6568 5.7807 x 10

Authors: Ekkawit Thaokeaw, Ph.D.(Candidate), Faculty of .(2018), Load frequency control of a dynamic interconnected
Engineering Mahasarakham University, Tambon Khamriang, power system using generalised Hopfield neural network based
Kantharawichai District, Maha Sarakham 44150 Thailand. self-adaptive PID controller, IET Generation, Transmission &
Krittanon Prathepha Ph.D., Faculty of Engineering, Mahasarakham Distribution, 12(21), 5713 – 5722.
University, Tambon Khamriang,Kantharawichai District, Maha [8] Mohamed Mokhtar 1, Mostafa I. Marei 1, Mariam A. Sameh 2,*
Sarakham 44150 Thailand. and Mahmoud A. Attia 1(2022). An Adaptive Load Frequency
Jagraphon OBMA, Ph.D., Faculty of Engineering, Rajamangala Control for Power Systems with Renewable Energy Sources,
University of Technology ISAN Khonkaen Campus, Khonkaen Energies 2022, 15, 573.
40000, Thailand. [9] J. Li, N. Yan and Z. Fu (2012). ANFIS Approach for Noise
Assoc.Prof.Worawat Sa-ngiamvibool, Ph.D., Faculty of Reduction of Lightning Current Online Monitoring System.
Engineering, Mahasarakham University, Tambon Khamriang, Przegląd Elektrotechniczny,88(7) : 108–112.
Kantharawichai District, Maha Sarakham 44150 Thailand. [10] Khuntia SR, Panda S. (2012).Simulation study for automatic
e-mail : [email protected] (Corresponding Author) generation control of a multi-area power system by ANFIS
approach. Appl Soft Comput 2012;12: 333–41.
REFERENCES [11] Sangram Keshori Mohapatra, Saumya Narayan Behera ,
[1] Hussein, Tawfiq, and Awad Shamekh. (2019). "Design of PI Susanta Kumar Biswal, and Kabita Bhuyan.(2021), Application
Fuzzy Logic Gain Scheduling Load Frequency Control in Two- DE Optimization Algorithm based Automatic Generation
Area Power Systems" Designs 3, no. 2: 26. Control of Interconnected Multi Area Power System., Journal of
[2] Chen, Chunyu, Kaifeng Zhang, Kun Yuan, and Xianliang Teng. Xi'an University of Architecture & Technology, Volume XIII,
(2017). "Tie-Line Bias Control Applicability to Load Frequency Issue 5, 2021, 86-96.
Control for Multi-Area Interconnected Power Systems of [12] Kumar SahuRabindra, Sidhartha Panda, Kumar Rout Umesh.
Complex Topology" Energies 10, no. 1: 78. “DE optimized parallel 2-DOF PID controller for load frequency
[3] Yu-Qing BAO, Yang LI, Beibei WANG, Minqiang HU & Peipei control of power system with governor dead-band non-
CHEN (2017). Demand response for frequency control of multi- linearity”. Electric Power Energy Syst 2013; 49: 19–33
area power system. Journal of Modern Power Systems and [13] Srisuwan, W. and Sa-ngiamvibool, W.(2019), Optimal
Clean Energy volume 5 : 20–29. Controllers Design for Isolated Hybrid Wind-Diesel Power
[4] Kumar, T. B., & Vani, M. U. (2014). Load frequency control in System using Bee Algorithm. Revue Roumaine Des Sciences
two area power system using ANFIS. Computer Engineering Techniques , vol. 64. no.4, pp.341-348.
and Intelligent Systems, 5: 27-35. [14] Durongdumrongchai, P. Sa-ngiamvibool, W., Aurasopon, A and
[5] Mosaad, M. I., & Salem, F. (2014). LFC based adaptive PID Pothiya, P.(2014), Robust and Optimal Fuzzy Logic
controller using ANN and ANFIS techniques. Journal of Proportional Integral Derivative Controllers Design by Bee
Electrical Systems and Information Technology, 1(3), 212-222. Algorithm for Hydro-Thermal System. Revue Roumaine Des
[6] Rostyslav YATSIUK and Dmytro MAMCHUR .(2020), The Sciences Techniques (Série Électrotechnique et Énergetique),
design of field-oriented control system with artificial neural vol. 59, no.2. pp.193-203.
network to control faulty induction motor, Przeglad [15] Rongrong Yu, Ye-Hwa Chen, Baokun Han, Han Zhao, "A
Elektrotechniczny , 96(9) : 30–33. Hierarchical Control Design Framework for Fuzzy Mechanical
[7] Rajeswari Ramachandran, Balasubramonian Madasamy, Systems With High-Order Uncertainty Bound", IEEE
Veerapandiyan Veerasamy andLoheswaran Saravanan Transactions on Fuzzy Systems, vol.29, no.4, pp.820-832, 202

108 PRZEGLĄD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 99 NR 2/2023