IET Generation, Transmission & Distribution

Research Article

Load frequency control of a dynamic ISSN 1751-8687

Received on 8th February 2018

interconnected power system using

Revised 12th July 2018
Accepted on 24th August 2018
E-First on 15th October 2018
generalised Hopfield neural network based doi: 10.1049/iet-gtd.2018.5622
www.ietdl.org

self-adaptive PID controller

Rajeswari Ramachandran1 , Balasubramonian Madasamy2, Veerapandiyan Veerasamy3, Loheswaran
Saravanan1
1Department of Electrical Engineering, Government College of Technology, Coimbatore-641013, India
2Department of Electrical Engineering, Alagappa Chettiar College of Engineering and Technology, Karaikudi-630004, India
3Center for Advanced Power and Energy Research (CAPER), Department of Electrical and Electronics Engineering, Faculty of Engineering,

Universiti Putra Malaysia (UPM), 43400 UPM Serdang, Selangor, Malaysia

E-mail: [email protected]

Abstract: A novel generalised Hopfield neural network (GHNN) based self-adaptive proportional–integral–derivative (PID)
controller for load frequency control (LFC) is designed for a two-area interconnected power system with nonlinearities of
generator rate constraint and governor dead band. The control problem is conceptualised as an optimisation problem with an
objective function as an area control error in terms of the PID controller parameters. The differential equations governing the
behaviour of the GHNN were solved to obtain the controller parameters Kp, Ki and Kd. To test the feasibility and robustness of
the proposed controller, the system is tested in the presence of randomness in load demands, imprecisely modelled system
dynamics, nonlinearities in the system model and uncertainties in the system parameter variations. The proposed method is
simulated using Matlab R2014b/Simulink and the results obtained have shown that the propounded controller performance is
superior to the integral, PID and fuzzy-based proportional–integral controllers. In addition, the Lyapunov stability analysis of the
overall closed-loop system was carried out and the controller is implemented in real-time digital simulator run in hardware-in-the-
loop to validate the effectiveness of the proposed method. Furthermore, the proposed controller is applied to the three-area
power system to test its adaptability.

1 Introduction internal model control scheme with two-degree-of-freedom in [10,

11]. Recently, LFC requires transmitting measurement data from
The recent development of the electrical power system makes the remote terminal units to the control centre and it sends the control
system less complex by sharing the electrical power from the area signals to the plant side. Hence, a delay-dependent robust LFC for
of surplus production to the neighbouring areas through tie-lines. time delays power systems was proposed in [12–14]. Another
In such a network, both the area frequency and tie-line power control strategy through a suboptimal solution was developed using
fluctuate frequently because of randomness in power load demand, an iterative linear matrix inequalities algorithm through the mixed
system parameter uncertainties and also other disturbances due to H2/H∞ control technique proposed in [14, 15] for LFC applications
the varying environmental condition. In this context, the study of
that employ proportional–integral (PI) controller.
load frequency control (LFC) is important for maintaining zero
An artificial intelligent method of the fuzzy and neural based
steady state errors for frequency deviations.
control system is used to design the adaptive PI controller
Generally, an automatic LFC (ALFC) problem can be solved
presented in [16–20] using area-control error and its change. The
through two techniques namely, the classical technique and
evolutionary algorithms such as genetic algorithm and particle-
intelligent technique. Classical control techniques of power
swarm optimisation were also used for effective designing and
systems are based on mathematical models and their modelling is
improving the performance of the controller in [21]. However, it is
not accurate in the presence of uncertainties. On the flipside, the
difficult to guarantee its effectiveness in a theoretical way, because
intelligent techniques do not require the mathematical modelling of
these techniques are essentially stochastic approaches. A novel
the plant to be controlled, strong assumption and the gradient
adaptive feed forward neural network-based PI–derivative (PID)
information. They employ objective functions and simple
controller has been designed and analysed for a multi-area
algorithms.
interconnected power system with a hybrid energy storage system
A systematic method of choosing the frequency bias parameter
in [22].
and the integral gain of the supplementary control using the
From the literature, it is observed all the approaches of tuning
Lyapunov technique with nonlinearities of governor dead band
parameters are random search methods and involve initial
(GDB) is considered in [1]. The design of robust decentralised LFC
assumptions. If the initial assumption of parameters is not assumed
algorithms was proposed for an interconnected power system, for
properly, then all the above approaches will give inaccurate results.
the purpose of regulating area control error (ACE) in the presence
However, the proposed generalised Hopfield neural network
of system uncertainties and external disturbances without
(GHNN) based PID controller violates the initial parameter
considering the nonlinearities of generator rate constant (GRC) and
assumption and it is a single layer feedback neural network with
GDB in [2–7]. All the above-mentioned approaches comprise
weight values as the coefficient of unknown parameters.
complexity in implementation, as a result, a special feature of the
This paper is organised as follows: the model of the multi-area
variable-structure controllers was designed which can greatly
power system and the design of integral, PID and fuzzy-based PI
improve the transient performance of the system by keeping the
controller are presented in Section 2. The generalised architecture
steady-state error at zero [8, 9]. The solution to the LFC problem
of the GHNN and the design of the GHNN based self-adaptive PID
without considering GRC and GDB was also illustrated using the

IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722 5713
© The Institution of Engineering and Technology 2018
 17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 1 Block diagram of the ith pool with GDB, GRC, and controller

controller with the closed-loop stability are discussed in Section 3. dΔPgi

The simulation and experimental results of the proposed controller max = 0.00167 p . u . /s . (3)
dt
with conventional and intelligent controllers for the two-area
power system are explained in Sections 4 and 5, respectively. 2.1 Design of conventional controllers
Section 6 presents the simulation and hardware-in-the loop
simulation of the three-area power system and finally Section 7 An integral controller is generally used for reducing the steady
deals with the conclusion of the proposed work. state error of the system. In this study, the frequency and tie-line
power deviation are represented as ACE and is given below [24]
2 System model
ACE1 = ΔPtie1 + B1Δ f 1, (4)
Consider an interconnected power system with power pools
connected by tie lines. In addition, all generators are assumed to ACE2 = ΔPtie2 + B2Δ f 2 . (5)
form a coherent group in each pool. Fig. 1 shows a block diagram
for the ith pool of an N-pool power system with GDB, GRC, and Thus the speed changer ΔPe1 and ΔPe2 can be represented as
controller [23].
The speed governor dead band in Fig. 1 has a significant effect
on the dynamic performance of the load-frequency control system.
ΔPe1 = − Ki1∫ (ACE1) dt, (6)
In fact, this backlash nonlinearity introduces a time lag associated
with the zero in the governor transfer function, as shown with the ΔPe2 = − Ki2∫ (ACE2) dt . (7)
aid of describing function analysis [1]. The governor and turbine
time constants are TGi = 10 s and TTi = 0.08 s, respectively, and the The constants Ki1 and Ki2 are integral gains. The negative sign in
turbine gain constant (KTi) is 0.2. The system parameters in Fig. 1 the above (6) and (7) indicates the generator in the area should
are defined as follows: increase its generation if either its frequency error Δ f i or its tie-
line power increment ΔPtie1 become negative. As per the design
ΔPtie, i – tie-line power deviation of ith pool procedure in [25], the integral gains are chosen as Ki1 = Ki2 = −0.5
T i j – synchronising coefficient for simulation, since both areas are identical.
To compare the performance of the LFC, the PID controller is
Δ f i – frequency deviation in the ith pool
used whose parameters are obtained using the Ziegler and Nichols
Bi – area frequency response characteristic of the ith pool (Z–N) method of tuning. The frequency deviation in both the areas
ΔPTi – change in turbine output and variation of tie line power are discussed in Section 4. The
Ri – regulation/droop in hertz per megawatt for the ith pool tuned PID controller parameters are obtained as Kp = −0.2284, Ki =
ΔPGi – change in governor output −0.2709 and Kd = 0.4452 using the design procedure in [26].
ΔPDi – change in load demand in the ith pool The fuzzy-PI controller is designed for the two-area
interconnected power system with changes in load condition,
ECAi – ACE of the ith pool system nonlinearities, and uncertainty. During control, a fuzzy
system is used to decide adaptively the proper gain of the PI
For digital simulation, (1) has been used for modelling GDB of controller according to the ACE and its change. In this case,
steam turbine but in practice, there exist constraints on the rates of triangular membership functions are used for the input variables
change of physical variables with five lingual. Fuzzy control rules provide a convenient way of
expressing control policy and domain knowledge of ALFC of the
0.2 power system. Simulations were performed using MATLAB/
F(x, ẋ) = 0.8ẋ − ẋ . (1)
π Simulink in connection with a fuzzy logic tool box. The system
response with a fuzzy logic-based controller is compared with the
Generally, there are two upper limits imposed on the operation of response of other controllers such as integral, PID, and GHNN
steam power plants, they are the rate limits dΔxei/dt of governor based PID controller for the two-area system [17, 26].
valve position and dΔPgi/dt of generating power. The following
rate limits are applied to the governor valve position [9] and the 3 Generalised architecture of GHNN
generating power:
GHNN is a single layer feedback network, which is continuously
dΔxei varying with time called as ‘gradient time network’. The GHNN
max = 0.1 p . u . /s, (2) performs optimisation or error minimisation using a deterministic
dt
search by optimising the fitness/energy function. In this approach,
5714 IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722
© The Institution of Engineering and Technology 2018
 17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
3.1 Design of GHNN based self-adaptive PID controller
The general block diagram of a unity feedback closed loop system
with a PID controller is shown in Fig. 3, where Yref is the set value
of plant output and Yplant is the actual output of the plant, such that
Yplant = T (u(t)), where T denotes the transformation function also
called operation or processing performed by the system on u(t) to
produce the output Yplant. u(t) is the actuating signal (input to the
plant/the output of the controller) to the plant and is given by

de(t)
u(t) = Kpe(t) + Ki∫ e(t)dt + Kd , (13)
dt
Fig. 2 Generalised architecture of GHNN
where Kp, Ki, and Kd are the proportional, integral and differential
gain of the controller. e(t) is the error signal (input to the
controller), which is the difference between the set point of the
plant out and the actual output of the plant, such that e(t) = Yref –
Yplant. To design the controller based on the proposed neural
network, consider a positive valued objective function to be
minimised. This objective function is the energy function, which is
square of error of the closed loop system. By minimising this
Fig. 3 Closed loop feedback control system energy function the system error is minimised and the system
reaches the stable state. The energy function is given as
the problem is optimised by minimising the energy function
monotonically and it can be applicable to any optimisation problem 1 2
by a proper choice of energy/Lyapunov function [27]. The Minimise E(Kp, Ki, Kd) = y − T Kpe(t)
2 ref
generalised architecture of the GHNN is shown in Fig. 2. It is a (14)
nonlinear interconnected single layer system of ‘n’ neurons. The de(t)
+Ki∫ e(t)dt + Kd .
nonlinear amplifier shown in Fig. 2 has an input–output activation dt
function φ( . ): ℜ → [0 1] . This function is assumed to be
continuously differentiable and strictly increasing monotonically. This objective function is made equivalent to a computational
The most commonly used activation function is the sigmoid energy function for a three-neuron GHNN, whose outputs are Kp,
function and is given by Ki and Kd such that Kp = ϕ(xp), Ki = xϕ(xi), Kd = ϕ(xd), where xp, xi,
and xd are the inputs to the neurons and ϕ is the activation function
x = φ(u) = 1/1 + e− u / θ, (8) of the neuron. In this design of the controller, a linear activation
function is used such that xp = Kp, xi = Ki, xd = Kd. It is required to
where the positive parameter ‘θ’ controls the slope of the activation find the values of Kp, Ki and Kd of the variables to be determined in
function. It is observed from the figure that the product unit Π order to minimise E (Kp, Ki, Kd).
produces the function fij(.),where ‘i’ ∈ [1 … n] and ‘j’ is the In the Hopfield neural network-based optimisation technique,
number of product terms. The output obtained from the product the changes in Kp, Ki, and Kd in time are in the direction of the
units is linearly summed up through the synaptic weights Wij, negative gradient of the computational energy function. Hence, the
where ‘i’ ∈ [1 … n] and ‘j’ is the number of product terms. adaptive rules for changes in Kp, Ki and Kd are shown as
Applying Kirchhoff's current law at the input of each amplifier, the
results are shown as dKp ∂E
= − , (15)
dt ∂Kp
dui ui
Ci
dt
+
Ri
= ∑ Wi j f i j(x , x , … . , xn) + Ii .
1 2 (9)
j dKi ∂E
= − , (16)
dt ∂Ki
The output states of the network are given by x = ϕ(u). The
activation function ϕ(u) is strictly increasing, differentiable, and dKd ∂E
invertible, i.e. = − . (17)
dt ∂Kd
u = φ−1(x) . (10)
The application of (14)–(17), results in the following update rules
for Kp, Ki and Kd as
Rewriting (9) as
dKp
dφ−1(xi) φ−1(xi) = e(t)T′(u(t))e(t), (18)
Ci
dt
+
Ri
= ∑ Wi j f i j(x , … . , xn) + Ii
1 . (11) dt
j
dKi
The energy function of the proposed model is given as = e(t)T′(u(t))∫ e(t)dt, (19)
dt
xi
φ−1(s) dKd
E= ∑∫ Ri
ds ∑ ∑ W i j f i j(x1, x2, … . . , xn)xi +
dt
= e(t)T′(u(t))
de(t)
dt
. (20)
0
i i j
(12)
∑ Iixi . Equations (18)–(20) describe the behaviour of the proposed GHNN
based controller and are solved simultaneously to obtain the values
of Kp, Ki, and Kd using MATLAB/ Simulink as in [28].The
By using conventional methodology we can show that the architecture of the proposed GHNN based PID controller is shown
derivative of the proposed energy function is always less than or in Fig. 4.
equal to zero, i.e. dE/dt ≤ 0 and energy of the system is bounded
[28, 29].

IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722 5715
© The Institution of Engineering and Technology 2018
 17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 4 Architecture of proposed GHNN based controller

Table 1 System parameters

Area capacity: PR1 = PR2 2000 MW
KP1 = KP2 120 Hz/p.u. MW
TP1 = TP2 20 s
TH1 = TH2 0.08 s
TT1 = TT2 0.3 s
R1 = R2 2.4 Hz/p.u. MW
B1 = B2 0.425 p.u. MW/Hz
D1 = D2 8.33 × 10–3 p.u. MW/Hz
12 = T21 0.08674 p.u. MW/rad

In this study, the effectiveness of the proposed controller is

examined using integral absolute error (IAE) and integral time
absolute error (ITAE) as
T
IAEi = ∫0
( | Δ f i | ) dt, i = 1, 2. (21)

T
ITAEi = ∫0
( | Δ f i | ) t dt, i = 1, 2. (22)

3.2 Stability analysis of the proposed controller

The stability of the overall system is proved in this section by
considering a positive definite squared error function of the closed
loop system as a Lyapunov function shown below Fig. 5 Case I-A
(a) Frequency deviation in area-1, (b) Frequency deviation in area-2, (c) Tie-line
1 power deviation
E(Kp, Ki, Kd) = e t 2, (23)
2
monotonically and therefore energy cannot remain constant except
1 at equilibrium point where it will be decreasing to zero with time.
E(Kp, Kd, Ki) = (yref − yp)2, (24) This shows that the stability is in Lyapunov of notion [27].
2

where Yref is the set value for the plant output and Yp is the actual 4 Simulation results and discussion
output of the plant.
A two-area interconnected power system is considered for
Now consider the time derivative of the proposed Lyapunov
simulation of the proposed GHNN based PID controller over the
functions
existing controller. The specifications of the governor, turbine and
re-heater parameters of the area-1 and area-2 are shown in Table 1.
dE ∂E dKp ∂E dKi ∂E dKd To illustrate the robustness of the proposed GHNN based
= + + . (25)
dt ∂Kp dt ∂Ki dt ∂Kd dt controller, the system is analysed with the following different
cases.
By substituting the adaptive rules of KP, KI and Kd in (15)–(17) for
dKp/dt, dKi/dt, and dKd/dt into (25), we obtain Case I-A: In this case, the step increase in the demand of area-2
(ΔPD2) equal to 25 MW is considered with area-1 undisturbed. The
dE ∂E ∂E
2 2
∂E 2
frequency deviation of area-1 and area-2, and inter-area tie line
= − + + . (26) power fluctuations are shown in Fig. 5. From the plot, it is
dt ∂Kp ∂Ki ∂Kd
observed that the frequency and tie-line power variation converges
Equation (26) shows that the derivative of the proposed Lyapunov to zero faster for the proffered controller than the integral, PID and
function is always less than zero, i.e. negative definite (dE/dt≤0) fuzzy PI controller. Tables 2 and 3 depict the transient performance
except for the equilibrium state. The system cannot remain in the indices of peak overshoot, settling time and steady state error for
non-equilibrium state for which the energy profile increases the frequency deviation of area-1 and area-2, respectively. It is
observed that the indices are decreased for the system with the
5716 IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722
© The Institution of Engineering and Technology 2018
 17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 2 Transient performance of frequency deviation in area-1
Case Δf1
Settling time (s) % Steady state error |Peak Overshoot| (Hz)
Integral PID Fuzzy PI GHNN PID Integral × 10–6 PID × 10–6 Fuzzy PI × GHNN PID × Integral PID Fuzzy PI GHNN PID
10–6 10–6
IA 25.34 23.93 21.91 10.12 61.5 5.32 9.48 5.40 0.0439 0.0431 0.0373 0.0255
IB 19.63 19.51 19.8 17.12 25.6 1.79 3.46 1.83 0.0419 0.0421 0.0366 0.0244
IIA 21.91 21.55 19.75 10.39 14.1 3.12 6.67 1.36 0.0421 0.0413 0.0358 0.0248
IIB 16.3 16.59 20.26 11.9 2.11 0.726 0.703 0.177 0.038 0.0388 0.03 0.0214

Table 3 Transient performance of frequency deviation in area-2

Case Δf2
Settling time (s) % steady state error |Peak overshoot| (Hz)
Integral PID Fuzzy PI GHNN PID Integral × 10–6 PID × 10–6 Fuzzy PI × GHNN PID × Integral PID Fuzzy PI GHNN PID
10–6 10–6
IA 27.74 25.07 23.01 8.82 60.4 13.3 4.00 3.07 0.038 0.0375 0.0348 0.0291
IB 20.76 20.66 20.5 17.48 24.3 7.37 1.42 1.87 0.0376 0.037 0.0334 0.0261
IIA 21.96 22.71 20.9 8.98 13.5 4.16 1.56 1.63 0.0373 0.0367 0.0342 0.0288
IIB 15.08 17.91 20.45 9.45 1.17 0.61 0.373 0.109 0.0419 0.0412 0.0352 0.028

Table 4 Comparison of performance index – IAE

Case IAE
Integral PID Fuzzy PI GHNN PID
Δf1 Δf2 Δf1 Δf2 Δf1 Δf2 Δf1 Δf2
IA 0.1274 0.144 0.1241 0.1337 0.1207 0.1324 0.0723 0.076
IB 0.1166 0.1275 0.1182 0.1225 0.1166 0.1221 0.0881 0.0896
IIA 0.108 0.124 0.1182 0.1187 0.0961 0.1096 0.0721 0.0756
IIB 0.1148 0.1182 0.1129 0.1216 0.1067 0.1271 0.08727 0.09053

proposed controller than the integral, PID and fuzzy-based PI and tie-line power variations for the above uncertain case are
controller [4]. Tables 4–6 depict the steady state performance shown in Fig. 9 and the corresponding changes in the values of KP,
indices such as IAE, ITAE, and control effort are greatly reduced KI and Kd are given in Figs. 10a–c. It is inferred that the values of
for the proposed controller. Moreover, the response of the fuzzy- KP, KI and Kd are tuned automatically even in the presence of
based PI controller is not satisfactory for the system with uncertain load disturbance or uncertainty conditions. In addition, to
uncertainty such as model parameter variation and nonlinearities in show the effectiveness of the overall system under admissible
the system model. uncertainties and load changes, the performance indices such as
Case II-A: In this case, the step increase in the demand of area-2 IAE and ITAE are calculated using (23) and (24), the results are
(ΔPD2) equal to 25 MW and 50% mismatch of inertia constant and tabulated in Tables 4 and 5. Furthermore, to study the potency of
damping co-efficient of area-1 and area-2 are considered with the proposed controller, control effort is also calculated. The
area-1 demand undisturbed. The frequency deviations of area-1 and control effort is the amount of energy or power required for the
area-2, inter-area tie line power fluctuations are shown in Fig. 6. controller to perform its duty. The control effort for the frequency
The oscillations in frequency and tie-line power deviation are deviation of area-1 and area-2 for all the cases is calculated and
damped out quickly with minimum overshoot and steady state error shown in Table 6. To ensure the stability of the overall system the
for the system with a proposed controller than the integral, PID and decrease in energy function profile against time for case I-A with
fuzzy PI controller and the corresponding parameters are given in the proposed controller has been observed and shown in Fig. 10d.
Tables 2–6. This depicts the system behaves in a Lyapunov sense and reaches a
Case I-B: In this study, the perturbation of ΔPD1 = 10 MW and stable equilibrium point [27].
ΔPD2 = 20 MW is applied to area-1 and area-2, respectively. The
frequency deviations of area-1 and area-2 and inter-area tie line
power fluctuations are shown in Fig. 7. From the figures, it is 5 Experimental validation using a real-time digital
observed that the frequency and tie-line power variation are simulator (RTDS)
quickly driven back to zero and the proposed GHNN based
controller shows superior performance than the integral, PID and The proposed GHNN based PID controller is validated using
fuzzy PI controller as shown in Tables 2–6. RTDS. The RTDS is a fully digital electromagnetic transient power
Case II-B: This case is similar to case I-B with 50% of system simulator that operates in real time. HYPERSIM is one of
synchronising power co-efficient mismatch. The frequency the most commercial RTDS available and has been developed by
deviation of area-1 and area-2, and inter-area tie line power Hydro-Québec based on OPAL-RT Technologies Inc. that runs in
fluctuations are shown in Fig. 8. Tables 2–6 show the effectiveness hardware-in-the-loop (HIL). In this work, OP4510 is used to
of the proposed controller against the integral, PID and fuzzy PI validate the proposed GHNN based control strategy for ALFC of
controllers. the multi-area dynamic power system against conventional
Case-III: To show the self-adaptiveness of the proposed approach, controllers. The experimental setup for testing the two-area system
two consecutive load perturbation of 10 and 20 MW are simulated with GHNN self-adaptive PID controllers and conventional
at time of t = 0 and 25 s in area-2. The results obtained show that controllers such as integral, PID and Fuzzy PI is shown in Fig. 11a.
the system settles to the equilibrium state even after the uncertain The results obtained such as frequency deviation and tie-line power
load disturbances. The frequency deviation of area-1 and area-2 deviation in real time for the case I-A of two area are shown in
Figs. 11b–d, respectively.

IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722 5717
© The Institution of Engineering and Technology 2018
 17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 5 Comparison of performance index – ITAE
Case ITAE
Integral PID Fuzzy PI GHNN PID
Δf1 Δf2 Δf1 Δf2 Δf1 Δf2 Δf1 Δf2
IA 0.8681 0.8969 0.6831 0.7106 0.5964 0.6031 0.3337 0.3158
IB 0.5168 0.545 0.4859 0.4966 0.5975 0.6101 0.3912 0.3842
IIA 0.5913 0.6175 0.5323 0.5415 0.5257 0.5402 0.3379 0.3174
IIB 0.4158 0.404 0.4574 0.4847 0.5776 0.5279 0.3927 0.3842

Table 6 Comparison of performance index – control effort

Case Control effort (Joules)
Area 1 Area 2
Integral PID Fuzzy PI GHNN PID Integral PID Fuzzy PI GHNN PID
IA 6.136 6.869 4.19 3.861 6.96 6.64 4.326 3.724
IB 8.5204 7.8663 5.9331 3.872 4.5349 5.8618 7.8559 4.4849
IIA 6.63 7.0548 4.4319 3.477 6.9422 6.803 4.29 3.727
IIB 4.6759 8.7263 4.5174 4.0213 4.6413 8.6128 4.4521 4.0911

Fig. 6 Case II-A

(a) Frequency deviation in area-1, (b) Frequency deviation in area-2, (c) Tie-line power deviation

6 Application of proposed controller to multi-area 7 Conclusion

system This study presents a GHNN based self-adaptive PID controller as
To test the effectiveness of the proposed controller, the controller is applied to the ALFC problem. The simulation results obtained have
applied for a three-area interconnected power system consisting of shown that the proposed controller violates the initial parameter
three identical areas with load disturbances of ΔPD2 = 20 MW, assumption and achieves the robustness and optimality by reducing
ΔPD3 = 10 MW and area-1 load demand undisturbed. The the steady state performance indices such as IAE, ITAE, control
effort and transient performance indices such as peak overshoot,
frequency and tie-line variations in simulation and real-time output
settling time and steady state error of frequency deviation of the
are given in Fig. 12. The results reveal that the proposed GHNN
system than the conventional integral, PID and fuzzy PI
based self-adaptive PID controller is robust for large disturbances
controllers. The effectiveness of the proposed controller is tested
such as variation in load, inertia constant and synchronising power
with different cases namely load change, variation in inertia
coefficient and also applicable to the multi-area power system,
constant and synchronising power coefficient (cases I-A, II-A, I-B
makes the system settle faster with zero or minimum overshoots
II-B and III). Moreover, stability analysis of the proposed
for minimising the system dynamics.
controller was carried out for the sustainable application of GHNN
5718 IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722
© The Institution of Engineering and Technology 2018
17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

5719
(a) Frequency deviation in area-1, (b) Frequency deviation in area-2, (c) Tie-line power deviation

(a) Frequency deviation in area-1, (b) Frequency deviation in area-2, (c) Tie-line power deviation

IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722
© The Institution of Engineering and Technology 2018
Fig. 8 Case II-B
Fig. 7 Case I-B
 17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 9 Consecutive load disturbances
(a) Frequency deviation in area-1, (b) Frequency deviation in area-2, (c) Tie-line power deviation

Fig. 10 Change in GHNN-controller parameters

(a) Change in Kp for consecutive disturbances, (b) Change in Ki for consecutive disturbances, (c) Change in Kd for consecutive disturbances, (d) Energy profile of proposed
controller

5720 IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722
© The Institution of Engineering and Technology 2018
 17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 11 Hardware setup and results
(a) Experimental setup for testing the two area system, (b) Frequency deviation in area-1 (Case-IA), (c) Frequency deviation in area-2 (Case-IA), (d) Tie-line power deviation (Case-
IA)

Fig. 12 Results of the three-area system with GHNN PID controller

(a) Simulation results of frequency deviation, (b) Simulation results of tie-line power deviation, (c) RTDS results of frequency deviation, (d) RTDS results of tie-line power deviation

to ALFC problem. Thus the energy profile of the proposed in a deterministic way by moving in the path of negative gradient
controller reaches zero and converges the system steady state error direction by ensuring the stability of the system in a Lyapunov

IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722 5721
© The Institution of Engineering and Technology 2018
 17518695, 2018, 21, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-gtd.2018.5622 by Bangladesh Hinari NPL, Wiley Online Library on [14/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
sense. Furthermore, to test the adaptability, the proposed controller [10] Tan, W.: ‘Unified tuning of PID load frequency controller for power systems
via IMC’, IEEE Trans. Power Syst., 2010, 25, (1), pp. 341–350
is tested for the three-area interconnected power system and the [11] Saxena, S., Hote, Y.V.: ‘Load frequency control in power systems via internal
results reveal that the proposed controller is suitable for any linear model control scheme and model-order reduction’, IEEE Trans. Power Syst.,
or nonlinear plant, which is imprecisely modelled and suffering 2013, 28, (3), pp. 2749–2757
from plant parameter variations. In addition, the simulation results [12] Bevrani, H., Hiyama, T.: ‘On load–frequency regulation with time delays:
design and real-time implementation’, IEEE Trans. Energy Convers., 2009,
of the two- and three-area dynamic power system are validated 24, (1), pp. 292–300
with the hardware results using HIL simulation. The proposed [13] Jiang, L., Yao, W., Wu, Q.H., et al.: ‘Delay-dependent stability for load
controller can also be applied and validated for the multi-area frequency control with constant and time-varying delays’, IEEE Trans. Power
interconnected power system constituted by the combination of the Syst., 2012, 27, (2), pp. 932–941
[14] Zhang, C.-K., Jiang, L., Wu, Q.H., et al.: ‘Delay-dependent robust load
thermal, hydel and renewable generation system. Moreover, the frequency control for time delay power systems’, IEEE Trans. Power Syst.,
concept of GHNN can be used for tuning the parameters of the 2013, 28, (3), pp. 2192–2201
fractional-order PID controller and the same can also be applied to [15] Yousef, H.A., AL-Kharusi, K., Albadi, M.H., et al.: ‘Load frequency control
the ALFC problem. of a multi-area power system: an adaptive fuzzy logic approach’, IEEE Trans.
Power Syst., 2014, 29, (4), pp. 1822–1830
[16] Chang, C.S., Fu, W.: ‘Area load frequency control using fuzzy gain
8 Acknowledgements scheduling of PI controllers’, Electr. Power Syst. Res., 1997, 42, (2), pp. 145–
152
The authors are thankful to the principal and authorities of [17] Abdennour, A.: ‘Adaptive optimal gain scheduling for the load frequency
Government College of Technology, Coimbatore-641013, India for control problem’, Electr. Power Comput. Syst., 2002, 30, (1), pp. 45–56
[18] Kocaarslan, I., Cam, E.: ‘Fuzzy logic controller in interconnected electrical
providing facilities procured under TEQIP fund to carry out the power systems for load-frequency control’, Int. J. Electr. Power Energy Syst.,
research. 2005, 27, (8), pp. 542–549
[19] Juang, C.F., Lu, C.F.: ‘Load-frequency control by hybrid evolutionary fuzzy
PI controller’, IEE Proc. Gener. Transm. Distrib., 2006, 153, (2), pp. 196–204
9 References [20] Arivoli, R., Chidambaram, I.A.: ‘CPSO based LFC for a two-area power
[1] Tripathy, S.C., Hope, G.S., Malik, O.P.: ‘Optimization of load-frequency system with GDB and GRC nonlinearities interconnected through TCPS in
control parameters for power systems with reheat steam turbines and governor series with the tie-line’, Int. J. Comput. Appl., 2012, 38, (7), pp. 1–10
dead band nonlinearity’, IEE Proc. Gener. Transm. Distrib., 1982, 129, (1), [21] Talaq, J., Al-Basri, F.: ‘Adaptive fuzzy gain scheduling for load frequency
pp. 10–16 control’, IEEE Trans. Power Syst., 1999, 14, (1), pp. 145–150
[2] Lim, K.Y., Wang, Y., Zhou, R.: ‘Robust decentralised load-frequency control [22] Xu, D., Liu, J., Yan, X.-G., et al.: ‘A novel adaptive neural network
of multi-area power systems’, IEE Proc. Gener. Transm. Distrib., 1996, 143, constrained control for multi-area interconnected power system with hybrid
(5), pp. 377–386 energy storage’, IEEE Trans. Ind. Electron., 2017, 65, (8), pp. 6625–6634
[3] Bevrani, H.Y., Mitani, Y., Tsuji, K.: ‘Robust decentralised load-frequency [23] Rahman, M.M., Chowdhury, A.H., Hossain, M.A.: ‘Improved load frequency
control using an iterative linear matrix inequalities algorithm’, IEE Proc. control using a fast acting active disturbance rejection controller’, Energies,
Gener. Transm. Distrib., 2004, 151, (3), pp. 347–354 2017, 10, pp. 1–18
[4] Sudha, K.R., Vijayasanthi, R.: ‘Robust decentralized load frequency control [24] Gupta, E., Saxena, A.: ‘Performance evaluation of antlion optimizer based
of interconnected power system with generation rate constraint using type-2 regulator in automatic generation control of interconnected power system’, J.
fuzzy approach’, Int. J. Electr. Power Energy Syst., 2011, 33, (3), pp. 699–707 Eng., 2016, 2016, pp. 1–14, http://dx.doi.org/10.1155/2016/4570617
[5] Dong, L., Zhang, Y., Gao, Z.: ‘A robust decentralized load frequency [25] Elgerd, O.I.: ‘Electric energy systems theory – an introduction’ (Tata
controller for interconnected power systems’, ISA Trans.., 2012, 51, (3), pp. McGraw-Hill, New York, 2007, 2nd edn.)
410–419 [26] Vijaya Chandrakala, K.R.M., Balamurugan, S., Sankaranarayanan, K.:
[6] Bevrani, H., Daneshmand, P.R.: ‘Fuzzy logic-based load-frequency control ‘Variable structure fuzzy gain scheduling based load frequency controller for
concerning high penetration of wind turbines’, IEEE Syst. J., 2012, 6, (1), pp. multi area hydro thermal system’, Int. J. Electr. Power Energy Syst., 2013, 53,
173–180 pp. 375–381
[7] Trinh, H., Fernando, T., Lu, H.H.C., et al.: ‘Quasi-decentralized functional [27] Veerapandiyan, V., Rajeswari, R., Mariammal, T., et al.: ‘Load flow analysis
observers for the LFC of interconnected power systems’, IEEE Trans. Power using generalized Hopfield neural network’, IET Gener. Transm. Distrib.,
Syst., 2013, 28, (3), pp. 3513–3514 2018, 12, (8), pp. 1765–1773
[8] Sivaramakrishnan, A.Y., Hariharan, M.V., Srisailam, M.C.: ‘Design of [28] Balasubramonian, M., Rajamani, V.: ‘Design and real-time implementation of
variable-structure load-frequency controller using pole assignment technique’, SHEPWM in single-phase inverter using generalized Hopfield neural
Int. J. Control, 1984, 40, (3), pp. 487–498 network’, IEEE Trans. Ind. Electron., 2014, 61, (11), pp. 1–11
[9] Chan, W.-C., Hsu, Y.-Y: ‘Automatic generation control of interconnected [29] Wen, U.-P., Lan, K.-M., Shih, H.-S.: ‘A review of Hopfield neural networks
power systems using variable-structure controllers’, IEE Proc. Gener. Transm. for solving mathematical programming problems’, Eur. J. Oper. Res., 2009,
Distrib., 1981, 128, (5), pp. 269–279 198, (3), pp. 675–787

5722 IET Gener. Transm. Distrib., 2018, Vol. 12 Iss. 21, pp. 5713-5722
© The Institution of Engineering and Technology 2018