energies

Article
A Neural Network-Based Model Reference Control
Architecture for Oscillation Damping in
Interconnected Power System
Waqar Uddin 1 , Nadia Zeb 2 , Kamran Zeb 1,3 , Muhammad Ishfaq 1 , Imran Khan 1 ,
Saif Ul Islam 1 , Ayesha Tanoli 4 , Aun Haider 4 , Hee-Je Kim 1, * and Gwan-Soo Park 1
1 School of Electrical and Computer Engineering, Pusan National University, 2 Busandaehak-ro 63 beon-gil,
Geumjeong-gu, Busan-City 46241, Korea; waqudn@pusan.ac.kr (W.U.); kami_zeb@yahoo.com (K.Z.);
engrishfaq1994@gmail.com (M.I.); imrankhan@pusan.ac.kr (I.K.); shaheen_575@yahoo.com (S.U.I.);
gspark@pusan.ac.kr (G.-S.P.)
2 Department of Electrical Engineering, COMSATS University Islamabad, Abbottabad Campus, Abbottabad
22010, Pakistan; nadia.zeb@gmail.com
3 Department of Electrical Engineering, National University of Science and Technology, Islamabad 44000,
Pakistan
4 Department of Electrical Engineering, University of Management and Technology, Lahore, Sialkot Campus,
Sialkot 51040, Pakistan; ieesha_tanoli@yahoo.com (A.T.); aun.haider@skt.umt.edu.pk (A.H.)
* Correspondence: heeje@pusan.ac.kr; Tel.: +82-10-3492-9677




Received: 12 July 2019; Accepted: 20 September 2019; Published: 24 September 2019


Abstract: In this paper, a model reference controller (MRC) based on a neural network (NN) is
proposed for damping oscillations in electric power systems. Variation in reactive load, internal
or external perturbation/faults, and asynchronization of the connected machine cause oscillations
in power systems. If the oscillation is not damped properly, it will lead to a complete collapse of
the power system. An MRC base unified power flow controller (UPFC) is proposed to mitigate
the oscillations in 2-area, 4-machine interconnected power systems. The MRC controller is using
the NN for training, as well as for plant identification. The proposed NN-based MRC controller is
capable of damping power oscillations; hence, the system acquires a stable condition. The response
of the proposed MRC is compared with the traditionally used proportional integral (PI) controller
to validate its performance. The key performance indicator integral square error (ISE) and integral
absolute error (IAE) of both controllers is calculated for single phase, two phase, and three phase
faults. MATLAB/Simulink is used to implement and simulate the 2-area, 4-machine power system.

Keywords: power oscillations; UPFC; non-linear control; neural network; model reference control

1. Introduction
The electric power system is composed of various interconnected generating stations that are
connected to load centers through the transmission and distribution systems. The interconnection of
different generating stations increases the complexity of the power network. Due to the stochastic
nature of the consumer profile, a mismatch occurs between the load demand and supply. The load
variation causes a change in system frequency, voltage, active power, and mechanical power. The
variation in these parameters causes fluctuation in the rotor speed. As a result, a low frequency (0.1 Hz
to few hertz) oscillation in active power takes place, known as a power oscillation. These power
oscillation either decays after a specified time or increases in magnitude until the system collapse
(blackout) occurs [1,2].

Energies 2019, 12, 3653; doi:10.3390/en12193653 www.mdpi.com/journal/energies

Energies 2019, 12, 3653 2 of 15

In order to overcome the issues of stability due to power oscillations and control power flow in a
system, a Flexible AC Transmission System (FACTS) has been proposed by scientists and researchers.
FACTS devices serve many functions in a power system, i.e., control of power flow, oscillation damping,
voltage control, and improvement of transient stability. The Unified Power Flow Controller (UPFC) is
an important member of the FACTS family. The UPFC is composed of two voltage source converters
coupled through a DC link. The converters provide both series and shunt compensation to the
transmission line, hence separating control of active and reactive power becomes possible.
The problem of Low-Frequency Oscillation (LFO) is tackled through FACTS devices in the modern
power system. FACT devices and a UPFC based on FACT devices are used mostly for mitigating the
LFOs. These devices are also used for enhancing the stability of the power system. The UPFC ensures
the security and stability of the system and damps out the LFOs.
Different linear and nonlinear control techniques are applied to mitigate the LFOs [2]. A modified
non-dominated sorting genetic algorithm-II (MNSGA-II) is proposed for control of the UPFC to damp
out power oscillation in [3]. The proposed design of the UPFC controller is based on a multi-objective
optimization problem. However, the controller response is slow in the case of a heavy loaded condition.
In [4], the authors presented Variable Neighborhood Search (VNS) optimization algorithm for tackling
the problem of oscillations. Although the design process includes high complexity, the proposed
controller responds well during normal conditions. However, the author did not consider the heavy
loaded or fault condition. An imperialist competition algorithm is used for damping power oscillation
in [5]. The controller is designed for wide range of operating conditions, but the author considered a
Single Machine Infinite Bus (SMIB) system. The artificial bee colony (ABC) algorithm [6] and Support
Vector Regression (SVR) [7] are also proposed for design of a supplementary damping controller.
However, the design is equipped with an additional Power System Stabilizer (PSS), along with
the UPFC, which will increase system cost and complexity. The author in [8] proposed adaptive
input–output feedback linearization control to achieve a non-linear stability estimator and 4th order
linearized model. However, due to the parallel application of these procedures, the system faces
instability during disturbances. The linear observer excludes the system nonlinearities and only
approximate system parameters. The sub-transient state of the synchronous generator is neglected after
the fault clearance, while the FACTS device is turned on and off by using an input–output controller.
In [9], the researcher proposed a co-evolutionary cultural algorithm (CPCA) based on particle swarm
optimization. The author proposed a PSS, along with UPFC, for oscillation damping. But the response
of the controller is slow as it takes too much time in damping the power oscillations. In [10], a PSS
is used to stabilize the LFOs. Also, in case of failure of the PSS, a supplementary control based on a
Linear Quadratic Regulator (LQR) regulates the control input to FACTS devices. In [11], the author
proposed a Fuzzy Neural Non-linear Proportional Integral controller for damping LFOs. The positive
aspects of this controller are its adaptive nature in learning rate, online tuning, simple structure, and
parameter identification. Inter-area oscillations can be efficiently damped through neural predictive
control. It will also enhance the stabilization of the power system [12]. A Takagi-Sugeno controller
produces gain over the large operating region. However, it cannot guarantee the stability of the system,
the robustness of uncertain perturbation in the system, and parameter variations. Transient stability
of a machine connected to an infinite bus system is enhanced by using non-linear coordination of
generator excitation in [13].
In our paper, we present the neural network (NN)-based model reference controller (MRC) control
scheme with a UPFC for mitigation and damping LFOs that is caused by the fault, disturbances,
or due to loss of synchronism in connected machines. The MRC controller uses a NN that has the
capability to learn and store information of non-linearities, using the auto-regression model instead
of tuning of input parameters. The MRC is the integration of the (1) reference model, (2) NN plant,
(3) NN controller, and (4) plant. The incorporation of all these four parts promises the robustness,
intelligence, and adaptation of parameters. The MRC has an inherently adaptive mechanism, which
will help in tracking of variation of process variables. The ability of a NN to learn and model complex
 Energies 2019, 12, 3653 3 of 15

and nonlinear relationships plays a vital role in control of a complex problem with uncertainties.
After initial weight adjustment, it acquires the ability to predict uncertainties. These features of an
Energies 2019, 12,MRC
NN-based x FORcontroller
PEER REVIEW will tackle the uncertainties and will have a better response than 3 of 15
other
conventional control schemes. The weight of the NN is adjusted with the Back Propagation (BP)
response
algorithm. than other conventional
Eventually, it will dampcontrol schemes.
out the The quickly
oscillation weight of andthe NN isthe
protect adjusted
systemwith
fromthe Back
complete
Propagation (BP) algorithm.
collapse. Moreover, Eventually,
the performance it will damp
of proposed out the
controller oscillation
is compared quickly
with the PIand protectforthe
controller the
system from complete collapse.
same system to validate our claim. Moreover, the performance of proposed controller is compared with
the PI The
controller
paperfor the same system
is organized to validate
as follows: Sectionour claim. the general system model, while Section 3
2 shows
discusses the mathematical modelling of the system. the
The paper is organized as follows: Section 2 shows Thegeneral system model,
supplementary controlwhile Section are
and results 3
discusses the mathematical modelling of the system. The supplementary
discussed in Sections 4 and 5, respectively. Section 6 concludes this paper. control and results are
discussed in Section 4 and Section 5, respectively. Section 6 concludes this paper.
2. System Model
2. System Model
A UPFC installed, 2-area, 4-generator connected power system is depicted in Figure 1. This model
A UPFC
is used installed,
to study LFOs of 2-area,
0.1 to 4-generator
few Hertz. connected power system is depicted in Figure 1. This
model is used to study LFOs of 0.1 to few Hertz.

Area 2 200 Mw 200 Mw Area 1

G3 G2
200 Mw
20 Mvar 1000 Mw
600 Mw
G4 B6 B5 B4 100 Km B3 B1 G1
100 Km B2 200 Km

100 Mw B7 400 Mw Fault 300 Mw

20 Mvar
25 Km

40 Mvar 10 Mvar

150 Km
B8 UPFC B 10
C dc
150 Km

ISe ISh
X Se X Sh
1000 Mw
Vdc 100 Mvar
Mw= Mega-watt
1000 Mw
B9 mSe δ Se mSh δSh
Mvar=Mega-var
100 Mvar

Figure 1. One-line diagram of power system with installed unified power flow controller (UPFC).
Figure 1. One-line diagram of power system with installed unified power flow controller (UPFC).

A DC capacitor voltage regulator is used to damp out the oscillation produces due to the occurrence
A DC capacitor voltage regulator is used to damp out the oscillation produces due to the
of disturbances in power, but it has a negative impact on the generated transients [14] because the
occurrence of disturbances in power, but it has a negative impact on the generated transients [14]
required amount of power is transmitted to the series inverter for injection to the transmission line
because the required amount of power is transmitted to the series inverter for injection to the
through DC link. Similarly, when variations in magnitude of DC link voltage occurs, it affects power
transmission line through DC link. Similarly, when variations in magnitude of DC link voltage
transfer capability of the shunt inverter. Hence, the active power is injected into the system with the
occurs, it affects power transfer capability of the shunt inverter. Hence, the active power is injected
help of series inverter [15].
into the system with the help of series inverter [15].
The synchronous generator connected to the system is used to generate power that is delivered
The synchronous generator connected to the system is used to generate power that is delivered
through the transmission lines. The installation point of the UPFC is between bus 8 and bus 9.
through the transmission lines. The installation point of the UPFC is between bus 8 and bus 9.
The DC link capacitor is used to make the back to back connection of Voltage Sources Converters
The DC link capacitor is used to make the back to back connection of Voltage Sources Converters
(VSCs) possible and transfer the power from the shunt converter to the series converter [16]. The
(VSCs) possible and transfer the power from the shunt converter to the series converter [16]. The
main purpose of Series VSC (VSC-SE) is to control the injection of active power and reactive power to
main purpose of Series VSC (VSC-SE) is to control the injection of active power and reactive power
the system, while Shunt VSC (VSC-SH) is used to adjust active power flow to series converter and
to the system, while Shunt VSC (VSC-SH) is used to adjust active power flow to series converter and
maintain DC link voltage. The connection of the shunt and series VSC is depicted in Figure 1. The
maintain DC link voltage. The connection of the shunt and series VSC is depicted in Figure 1. The
connection of both the converters is such that it diminishes the total power injection to the system.
connection of both the converters is such that it diminishes the total power injection to the system.

𝑃 P+Se𝑃+ P=
Sh 0
=. 0. (1)
(1)

The input control signal to UPFC are 𝑚 ,𝑚 ,𝛿 , and 𝛿 . 𝑚 and 𝑚 is used to represent
the amplitude modulation index of series and shunt compensation, respectively. However, 𝛿 and
𝛿 are used for the representation of series and shunt compensation phase angles [2,17]. In the case
of a series converter, 𝛿 and 𝑚 are the important factors that need to be determined. Because 𝛿
is used to control real power flow in to the high voltage direct current (HVDC) transmission line, and
Energies 2019, 12, 3653 4 of 15

The input control signal to UPFC are mSe , mSh , δSe , and δSh . mSe and mSh is used to represent the
amplitude modulation index of series and shunt compensation, respectively. However, δSe and δSh
are used for the representation of series and shunt compensation phase angles [2,17]. In the case of a
series converter, δSe and mSe are the important factors that need to be determined. Because δSe is used
to control
Energies 2019,real
12, xpower flow
FOR PEER in to the high voltage direct current (HVDC) transmission line, and 4mof
REVIEW is
Se 15
responsible for voltage control. Similarly, in the case of the shunt converter, the reactive power of the
of
ACthe AC terminal
terminal can becan be controlled
controlled throughthrough
mSh , and𝑚 δSh, and 𝛿 to
is used is control
used toDCcontrol DC link(V
link voltage voltage (𝑉 ).
dc ). VSC-SE
VSC-SE injects to
injects voltage voltage
the AC totransmission
the AC transmission line through
line through a series connected
a series connected transformer.
transformer. This
This series series
injected
injected
voltage isvoltage
used toistackle
used different
to tackleproblems,
different problems, i.e., phase
i.e., phase shift, series shift, series compensation,
compensation, voltage control,voltage
and
control,
combinationand combination of these parameters.
of these parameters. VSC-SH is used VSC-SH
to takeis used to takefrom
real power real AC
power from AC
terminal andterminal
transfer
and transfer
it to DC it toThis
terminal. DC power
terminal. Thistransferred
is then power is by then transferred
VSC-SE by VSC-SE
to the system to the system
for compensation. for
These
compensation.
features make the These features
UPFC makeFACTS
a flexible the UPFCdevicea flexible
that has FACTS device that
the properties hasother
of all the properties
FACTS devices.of all
other FACTS
The block devices.
diagram TheUPFC
of the block general
diagramcontrol
of the isUPFC general
depicted control2.isVdepicted
in Figure 1 and V2 inareFigure 2. 𝑉 and
bus voltages to
𝑉 are VSC-SH
which bus voltages to which
and VSC-SE areVSC-SH
connected,and VSC-SE arewhile
respectively, connected, respectively,
“i” is line current. while “𝑖” is line
current.

i V dqse P

Q
V1 V2
i se
i sh

Shunt Converter V dc Series Converter

i sh i
Shunt Converter Series Converter
V1
V1 Control Control
V2

i qref V dcref P dqref

UPFC control.
Figure 2. UPFC control.

3. Mathematical
3. Mathematical Modelling
Modelling of
of Power
Power System
System
The model
The modelused
usedininthis paper
this is composed
paper of a of
is composed UPFC installed,
a UPFC 4-machine,
installed, two-area
4-machine, power system.
two-area power
The UPFC is composed of a VSC − SH, VSC − SE, and DC link.
system. The UPFC is composed of a VSC − SH, VSC − SE, and DC link.
Mathematical Model
Mathematical Model
The non-linear model of the UPFC is presented below from Equation (2) to Equation (3). This
The non-linear model of the UPFC is presented below from Equation (2) to Equation (3). This
model is used for the analysis of stability of the power system. The parameters that affect the dynamics
model is used for the analysis of stability of the power system. The parameters that affect the
of the system are considered in this model, and resistance and transients of transformer is neglected. The
dynamics of the system are considered in this model, and resistance and transients of transformer is
modelling equations given below are transformed to dq − axis by using Park’s transformation [17,18].
neglected. The modelling equations given below are transformed to dq − axis by using Park’s
Series VSC and shunt VSC are modelled as:
transformation [17,18]. Series VSC and shunt VSC are modelled as:
𝑉 Vdse 0 0 −𝑋−Xse 𝐼 Idse mse Vdc𝑐𝑜𝑠𝛿cosδse
" # " #" # " #
(2)
𝑉 Vqse= 𝑋 Xse 0 0 𝐼 Iqse + + ∗ ∗
= , , (2)
2 𝑠𝑖𝑛𝛿 sinδse
𝑉 0 −𝑋 𝐼 𝑐𝑜𝑠𝛿 (3)
𝑉 = 𝐼 + ∗ .
𝑋 0 𝑠𝑖𝑛𝛿
The subscript ′𝑠ℎ’ and ′𝑠𝑒′ is used for shunt and series converter parameters, respectively. 𝑋 ,
𝑋 are equivalent reactance of series and shunt transformer. 𝑉 , 𝑉 are the dq voltage
Energies 2019, 12, 3653 5 of 15

" # " #" # " #

Vdsh 0 −Xsh Idsh msh Vdc cosδsh
= + ∗ . (3)
Vqsh Xsh 0 Iqsh 2 sinδsh
The subscript 0 sh’ and 0 se0 is used for shunt and series converter parameters, respectively. Xse , Xsh
are equivalent reactance of series and shunt transformer. Vdse , Vqse are the dq voltage component of
series branch, while Idse , Iqse are dq-component of series converter current. The dynamic of DC voltage
at DC link is represented as:
" # " #
dvdc 3msh h i I
dsh 3mse h i I
dse
= cosδsh sinδsh + cosδse sinδse . (4)
dt 4Cdc Iqsh 4Cdc Iqse

Vdc shows the DC link voltage, and Cdc is the capacitances of the DC link capacitor. The model of
the UPFC connected power system is presented by:
.
δ = ω0 ∆ω, (5)

. Pmech − Pelec − D∆ω

ω= , (6)
M

. E f −d KA
Efd = − + (Vs0 − Vs ), (7)
TA TA
 
. E f −d − xd − x0d id − E0sq
E0sq = 0 . (8)
Td0
Pmech and Pelec represent mechanical and electrical power, respectively, where M and D represent
inertia and damping coefficient, respectively. ω0 represents synchronous speed. δ and ω are rotor
angle and rotor speed, respectively, while ∆ω = (ω − 1). E f d , E0sq , and V are field, internal, and the
0 represents the time constant of the open circuit. T and K are time
terminal voltage of generator. Td0 A A
constant and exciter gain, respectively, while Vs0 represents reference voltage. xd is the d-axis reactance
of generator, and x0d is transient reactance. Furthermore, the different terms used in Equation (5) to
Equation (8) are given as:

Pelec = Vs−q Is−q + Vs−d Is−d 


q 
2
Vs = Vd + Vq , 2 


. (9)


0 0
Vd = Xd Iq , Vq = Eq − Xd Id 


I = I +I ,I = I +I


d Ed Bd q Eq Bq

Similarly, the line current for VSC-SE and VSC-SH can be obtained as [17–19]:

Xsh 0 Xshd msh sinδsh Vdc X mse sinδse Vdc

 
Ised = − Eq + − td + Vb cosδ , (10)
Xd Σ 2Xd Σ Xd Σ 2

Xshq msh sinδsh Vdc Xshq  mse sinδse Vdc 

Iseq = − + + Vb sinδ , (11)
2Xq Σ Xq Σ 2

Xse 0 Xsed msh sinδsh Vdc X m sinδsh Vdc

 
Ishd = Eq − + shd sh + Vb cosδ , (12)
Xd Σ 2Xd Σ Xd Σ 2

Xshq msh cosδsh Vdc Xtq  mse cosδse Vdc 

Ishq = − + + Vb sinδ , (13)
2Xq Σ Xq Σ 2
where as Xd Σ, Xtd , Xq Σ, and Xtq are defined in [18].
Energies 2019, 12, 3653 6 of 15

4. Supplementary Control Design

Supplementary control schemes with the MRC are introduced for mitigating LFOs in the power
system. The control design is used in the UPFC internal control structure to enhance the stability of
the power system. As the faults occur on the transmission system, the speed of a different generator
connected to the power system deviates and become out of synchronism. This speed deviation (∆ω) is
treated as an input to the supplementary controller, as depicted in Figure 3. The controllers mitigate
the LFOs by providing suitable input to the UPFC, hence improving the power quality and stabilizing
the power system. The design of the controllers is based on model reference neural control, and it uses
a nonlinear autoregressive moving average model. The MRC is the integration of the (1) reference
model, (2) NN plant, (3) NN controller, and (4) plant. The different parts of controller will be discussed
Energies
in detail2019,
in a12,subsection
x FOR PEERof
REVIEW
this section. 6 of 15

MRC
∆𝝎
Damping
Controller
Pref
P
+ Regulator
V2 Voltage
-
Measurement V2dq
P
V1 PQ
PLL 𝜽 = 𝝎𝒕
Measurement
Q -
Current I dq
I
Measurement + Q
Regulator
𝜽
Q ref
Gating Signals

Vd Vdref
VSC

PWM auto
Modulator Vq Vqref

Figure 3. Control structure for oscillation damping. MRC = model reference controller; VSC = Voltage
Figure 3. Control structure for oscillation damping. MRC = model reference controller; VSC = Voltage
Sources Converters; PQ = Active Power, Reactive Power; PWM = Pulse Width Modulation; PLL =
Sources Converters; PQ = Active Power, Reactive Power; PWM = Pulse Width Modulation; PLL =
Phase Lock Loop.
Phase Lock Loop.
4.1. Neural Network (NN)
4.1. Neural Network (NN)
The NN, inspired by the biological system, is composed of interconnected building blocks known
The NN,
as neurons. Theinspired by the biological
interconnection system, the
weight quantifies is composed
strength ofofinterconnection.
interconnected Because
buildingofblocks
these
known as neurons. The interconnection weight quantifies the strength
interconnected weights, the NN acquires the ability to learn and model complex relationships. of interconnection. Because
Hence,of
these
it interconnected
becomes weights,
more intelligent the NN acquires
in predicting the ability Due
the uncertainties. to learn and
to the model capability
learning complex relationships.
of NN, it has
Hence, it becomes more intelligent in predicting the uncertainties.
a wide range of applications in automation and optimization [20], image processing [21,22], Due to the learning capability
speechof
NN, it has a wide range of applications in automation and
recognition [23,24], control [25,26], modelling, and time series prediction [27,28]. optimization [20], image processing
[21,22],
Thespeech recognition
architecture of NN[23,24], control of
is composed [25,26], modelling,
multiple and and
input layer timehidden
series prediction
layers, as [27,28].
depicted in
The architecture of NN is composed of multiple input layer
Figure 4. This architecture is known as “N-1” feedforward NNs. Whereas ‘N’ represents and hidden layers, as
thedepicted
number ofin
Figure 4. This architecture is known as “N-1” feedforward NNs. Whereas
hidden neurons and it has only one output node. The design of the architecture of NN depends on the ‘N’ represents the number
of hidden
input neurons
weights andoutput
‘zij ’ and it has only one ‘γ
weights output node. The design of the architecture of NN depends on
i ’. The formulation of the NN through the nonlinear function
the input weights ‘𝑧 ’ and output weights
of hidden layer activation function ‘ψi ’ is presented ‘𝛾 ’. The
as:formulation of the NN through the nonlinear
function of hidden layer activation function ‘𝜓 ’ is presented as:
𝑦(𝑘ŷ(+ ) =∑ 𝛾 𝜓γi ψ
1)1= ∑ 𝑧 𝑋 z, X ,
XN Xn+m
(18)

k+ i ij j (14)
where 𝑋 = 𝑢 , 𝑢 , … . , 𝑢 i=1 j=1
is the input vector to NN and ‘𝜓 ’ is the activation function of the ‘𝑖 ’
hidden layer. The neuron in the hidden layer uses the nonlinear logistic function given by Equation
(19):

𝜓 ∑ 𝑧 𝑋 = . (19)
∑

The equation can be rearranged as:

𝑦(𝑘 + 1) = 𝜃 𝜓, (20)
Energies 2019, 12, 3653 7 of 15

where X = [u1 , u2 , . . . ., um+n ] is the input vector to NN and ‘ψi ’ is the activation function of the ‘ith ’
hidden layer. The neuron in the hidden layer uses the nonlinear logistic function given by Equation (15):

1
Xn+m 
ψi zij X j = P . (15)
j=1
1 + exp nj=+1m zij X j

The equation can be rearranged as:

ŷ(k + 1) = θT ψ, (16)

whereas ψ , (ψ1 , ψ2 , . . . ., ψN )T , θT , (γ1 , γ2 , . . . ., γN ). The gradient descent optimization technique

is used to tune and update the output weight only. This will, in turn, reduce the computational
complexity of the NN. The update equation of the NN is given as:

γk+1 = γk α ( yr × ŷ) × ( yr × ŷ)0 .

 
(17)
Energies 2019, 12, x FOR PEER REVIEW 7 of 15

z 11
u1 ψ (.)

𝛾1

u2 ψ (.) 𝛾2
^
y
𝛾𝑚

um ψ (.)

𝛾(𝑚 +𝑛)

u(m+n) ψ (.)

Input Layer Hidden Layer Output Layer

Figure 4.
Figure 4. Neural network (NN)
(NN) architecture.
architecture.

4.2.
4.2. Model
Model Reference
Reference Controller
Controller
The
The NNNN controller
controller is is used
used in in the
the MRC
MRC for
for tracking
tracking the the reference
reference signal,
signal, as
as well
well asas for
for plant
plant
identification.
identification. The MRC uses two NN controllers and minimal online computation. One NN is used
The MRC uses two NN controllers and minimal online computation. One NN is used
to
to train
train the
the controller,
controller,and andthetheother
otherisisused
usedfor
forplant
plantidentification.
identification. The
The data
data set
set for
for training
training the
the NN
NN
controller
controller is
is given
given in in Appendix
Appendix A. A. The
The architecture
architecture of of thethe MRC
MRC is depicted in
is depicted in Figure
Figure 5.5. The
The controller
controller
produces
produces aacontrolled
controlledoutput signalu(𝑢(𝑡)
outputsignal t) that acts
that as input
acts as inputto the plant.
to the ThisThis
plant. control signal
control brings
signal the
brings
output signal
the output signaly ( t ) closer to the reference signal
p 𝑦 (𝑡) closer to the reference signal y r ( t ) . 𝑦 (𝑡).
According to [29], the different inputs and their
According to [29], the different inputs and their respective respective outputs define
outputs the the
define characteristics of the
characteristics of
real-time process. The previous information of input and output is used
the real-time process. The previous information of input and output is used to introduce systemto introduce system dynamics.
Equation
dynamics.(18) is sampled
Equation at sampled
(22) is a regularat interval of time:
a regular interval of time:
𝑦 (𝑡) =yp𝑓(t)𝑦=(𝑡fp−yp1), …1𝑦), .(𝑡. . −
yp (𝑛t −),n𝑢(𝑡
a ), u−
(t 𝑘),
− k)…, .𝑢(𝑡
. . u(−
t −𝑘k−
−𝑛nb )) ,, (22)
h i
(t − (18)
where 𝑘 ≥ 1 shows the time delay in number of samples, 𝑓 (𝑦, 𝑢) is processed by non-linearity,
and 𝑛 , 𝑛 + 1 represents the process output and input, respectively. Equation (22) gets into the
shape of the transfer function below by linearizing the process about the operating point.
( ) ( ) ⋯
=𝑞 =𝑞 . (23)
( ) ( ) ⋯
Energies 2019, 12, 3653 8 of 15

where k ≥ 1 shows the time delay in number of samples, fp [( y, u)] is processed by non-linearity, and na ,
nb + 1 represents the process output and input, respectively. Equation (18) gets into the shape of the
transfer function below by linearizing the process about the operating point.
 
yp ( t ) B q−1 b0 + b1 q−1 + . . . + bnb q−nb
−k
=q = q−k . (19)
u(t) A(q−1 ) 1 + a1 q−1 + . . . + ana q−na
Energies 2019, 12, x FOR PEER REVIEW 8 of 15

Control error
Reference Yr - +
Model
+
Yp
NN Plant
Model

Model error
Command +
Input
u +
NN Controller
-
Plant
Plant Output

Figure
Figure 5. NN-based MRC.
5. NN-based MRC.

In Equation (19), yp (t) and u(t) represents the deviation of process from its operating points. The
In Equation (23), 𝑦 (𝑡) and 𝑢(𝑡) represents the deviation of process from its operating points.
parameters ai and bi are dependent on the operating condition, while q−1 represents the backward shift
The parameters 𝑎 and 𝑏 are dependent on the operating condition, while 𝑞 represents the
operator. The equation for the generalized linear controller is given in Equation (20)
backward shift operator. The equation for the generalized linear controller is given in Equation (24)
1 1   −1 
𝐻(𝑞 H )𝑅(𝑡) )𝑦−1 (𝑡) ,
(24)
  
𝑢(𝑡) u=(t) = q R−( t 𝐺(𝑞
) − G q y p ( t ) , (20)
𝐹(𝑞 F)(q−1 )
where ‘𝑅’ is set point, while 𝐹, 𝐺, and 𝐻 are the polynomial defined below:
where ‘R’ is set point, while F, G, and H are the polynomial defined below:
𝐹(𝑞 ) =1 + 𝑓 𝑞 + ⋯ + 𝑓 𝑞
−1 = 1 + f q−1 + . . . + f q−n f

𝐺(𝑞


 F) q=
 𝑔 + 𝑔 𝑞 1 + ⋯ + 𝑔 𝑞n f . (25)
−1 = g + g q−1 + . . . + g q−n g .

G q
) = ℎ + 0ℎ 𝑞 1 + ⋯ + ℎ 𝑞 ng (21)

𝐻(𝑞


  −1 

 Hq = h0 + h1 q−1 + . . . + h q−nh
The closed-loop transfer function of a linearized system isnhgiven by:
The closed-loop transfer( function
) (
of a linearized ) system
( ) is given by:
=𝑞 ( ) ( )
. (26)
( )
   ) ( )
(
yp (t) structure
Equation (24) and the control q−1 𝑛H qof
𝑛 , 𝑛B and
−1
Equation (25) is used to design a linear
= q−k −1 −1
. (22)
controller model to track theR(reference
t) A (q )FThe
point. ) + q B(q−1B)G
−k
(q polynomial q−1 )
is(neglected to make Equation (26)
equal to the characteristic equation of the system. Hence, B is replaced with another term,
Equation
i.e., 𝐵(𝑞 ) = 𝑏(20) and ).
𝐵 (𝑞 the control (26)
Equation structure n f , n g andinto
will transform nh of
theEquation
following (21) is used to design a linear
equation:
controller model to track the reference point. The polynomial B is neglected to make Equation  (22)
(𝑡)
equal to the characteristic𝑦equation 𝐻(𝑞 B )is replaced with another term, i.e., B q−1 =
of the system. 𝑏Hence,

=𝑞
b0 B+ q−1 . Equation (22)𝑅(𝑡) 𝐴(𝑞into)𝐹(𝑞 ) + 𝑞 equation:
𝑏 𝐺(𝑞 )
 
will transform the following
( ) ⋯ 
y(p ()t)= 𝑞 −k

b0 H q−1 .
⋯
(27)
=q
R(t) A(q−1 )F(q−1 ) + q−1 b0 G(q−1 )
𝑆(𝑞 ) and 𝑇(𝑞 ) are the polynomial that is used to set closed-loop response to a desired
value. 𝑘 is the time delay of the reference model. In order to enable the cancellation of 𝐵 (𝑞 ),
𝐹(𝑞 ) is replaced with 𝐵 (𝑞 )𝐹 (𝑞 ), i.e., 𝐹(𝑞 ) = 𝐵 (𝑞 )𝐹 (𝑞 ). The minimal order of the
unique controller solution is:

𝑛 = 𝑛 − 1, 𝑛 = 𝑛 + 𝑘 − 1, 𝑘 = 𝑘, while 𝐻(𝑞 ) = 𝑆(𝑞 )

Energies 2019, 12, 3653 9 of 15

yp (t) s0 + s1 q−1 + . . . + sns q−ns

= q−km . (23)
R(t) 1 + t1 q−1 + . . . + tnt q−nt
   
S q−1 and T q−1 are the polynomial that is used to set closed-loop response to a desired value.
   
km is the time delay of the reference model. In order to enable the cancellation of B+ q−1 , F q−1 is
         
replaced with B+ q−1 F1 q−1 , i.e., F q−1 = B+ q−1 F1 q−1 . The minimal order of the unique controller
solution is:
  1  
n g = na − 1, n f = nb + k − 1, km = k, while H q−1 = S q−1 (24)
b0
and
nt ≤ na + k − 1. (25)

Hence, the structure of process na and k is used to limit the order of denominator of closed-looped
transfer function nt . The order of nt should fulfill the condition given in Equation (25). The lower order
of denominator is possible by making unused ti coefficient equal to zero. The following Equation (26)
is used to introduce an observer into the controller design.
 
C q−1 = 1 + c1 q−1 + c2 q−2 + cnc q−nc , (26)

while
nc ≤ na + k − 1 − nt . (27)

Then, Equation (25) is updated as:

  1    
H q−1 = V q−1 C q−1 . (28)
b0

Although due to nonlinearity, the parameter of controller will change, but the controller designed
structure in Equation (20) will not change  for
 any operating point of the plant in Equation (18)
despite setting the observer polynomial C q−1 = 1 and ti coefficient to zero. As in Equation (20), a
nonlinear controller design with the same input and output can smoothly interpolate between different
parameters of linear controllers as the operating points of the plant changes. The MRC in Equation (29)
is proposed by considering Equation (20) in difference equation form and linearizing a nonlinear
system into the form of Equation (20)
h    i
u(t) = fr u(t − 1), . . . u t − n f , r f (t), yp (t), . . . yp t − n g . (29)

In Equation (29), the NN will be used to realize fr [n], and Equation (24) will define the order of n g
   
and n f , while Equation (27) defines nc and r f (t) = S q−1 C q−1 r(t).

Controller Adaption with BP

The gradient descent algorithm with time delay is implemented to update the parameter of NN.
The output of NN with a multilayer of a neuron is given as:
X  X  
u ( t ) = f0 z2j fh z1ij xi (t) + b1j + b2 . (30)
j j

xi is the input to the NN, and b and z are biases and weights of the NN, respectively. f0 represents the
nonlinear function of the input layer, and fh is function of the hidden layer. The cost function that
needs to be minimized is given as:

1 1 2
J (t) = e(t j)2 = yr (t) − yp (t) . (31)
2 2
Energies 2019, 12, 3653 10 of 15

yr and yp are the reference and measured output of the system, respectively. The weight parameter ‘z’
is used in opposite direction in cost function gradient to minimize the error, given as:

∂J (t)
z(t) = z(t − 1) − η . (32)
∂z
The unknown ‘η’ represents the learning rate. The BP algorithm is used to calculate the gradient
∂J (t)
algorithm of Equation (32) Hence, Equation (18) and ∂yp (t)
= −e(t) gives:

∂J (t) ∂yp (t) ∂u(t − k)

= −e(t) . (33)
∂z ∂u(t − k) ∂z

In Equation (33), the first derivative is called process Jacobian. The second NN is used to estimate
it as a process model, i.e., the NN model given by:
X  X  
ym (t) = fm [n] = f0 z2m
j fh z 1m m
x
ij i ( t ) + b1m
j + b 2m
. (34)
j i

The delayed model inputs/outputs are represented by xm i

(t), as shown in Equation (18) in the
form of u(t − k). Equation (35) shows one of the estimations of Jacobian.

∂yp (t) ∂ym (t) X

≈ = f00 z2m fh0 z1m . (35)
∂u(t − k) ∂u(t − k) j j j 1j

The process sign is used to approximate Jacobian if the model is not available. Equation (36)
shows the update rule for the different parameter of the controller.

∂yp (t) ∂u(t − k)

!
z(t) = z(t − 1) + ηe(t)sgn . (36)
∂u(t − k) ∂z

This equation is used to update the rule to get better dynamics of the system.

5. Results and Discussion

The performance of the designed controller is evaluated and compared with the PI controller
under symmetrical and asymmetrical faults. The faults are applied for 50 ms, i.e., the fault is initiated
is at t = 200 ms and cleared at t = 250 ms. Moreover, the fault is applied on the mid-point of the
lower line of the double circuit line between buses B1 and B2 , as depicted in Figure 1. The proposed
controller has the capability to mitigate the LFOs and enhance the stability of the power system. The
parameters used for comparison of the proposed control structure are: (1) load bus, voltage, (2) speed
deviation (∆w), and (3) performance indices of controllers.

5.1. Asymmetrical Faults

Both asymmetrical and symmetrical faults are applied to the system for 50 ms (at t = 200 ms) to
create oscillations in the system. The fault creates a voltage dip of 10% during the 1-Φ fault, while a
25% dip is created during the 2-Φ to ground fault. After the clearance of the 1-Φ fault, the voltage
profile is restored at t = 2 s due to the adaptive and predictive nature of MRC, while the response
with conventional PI is oscillatory and oscillation is damped out after t = 12 s, as shown in Figures 6
and 7, respectively. Similarly, the speed deviation shows high amplitude oscillation in the case of the
PI controller, while the MRC has the capability to damp out the oscillation in comparatively less time.
Figure 8 shows that the MRC has damped out the oscillation at t = 2 s, and speed deviation (∆w ≈ 0)
is approximately zero, while in comparison with MRC, the PI controller deviates from zero till t = 6 s,
as depicted in Figure 8.
deviates from zero till 𝑡 = 6 s, as depicted in Figure 8.
Similarly, during the 2-Ф to ground fault, the MRC has a better response in mitigation of
oscillation as compared to the PI. As depicted in Figure 9, the MRC has smoothed the voltage
response abruptly after fault clearance at 𝑡 = 2.5 s , while the PI controller failed to show the
robustness. The oscillation is completely damped out at 𝑡 = 12 s, as shown in Figure 7. Similarly,
Energies 2019, 12, 3653 11 of 15
speed deviation becomes zero at 𝑡 = 2 s in the case of the MRC, while the PI vanishes the oscillations
at 𝑡 = 8 s.

Voltage (v) p.u

Energies 2019, 12, x FOR PEER REVIEW 11 of 15

Energies 2019, 12, x FOR PEER REVIEW 11 of 15
Figure 6. Load bus voltage 1-Φ
Figure 6. 1-Ф to ground fault.
SpeedDeviation
Speed p.u
Deviation( ( w)w)p.u

Figure 7. Speed
Speed deviation
deviation during 2-Ф to
to ground
ground fault.
Figure
Figure 7.
7. Speed deviation during
during 2-Φ
2-Ф to ground fault.
fault.
SpeedDeviation
Speed p.u
Deviation( ( w)w)p.u

Figure 8.
Figure Speed deviation
8. Speed deviation during
during 1-Ф
1-Φ to
to ground
ground fault.
fault.
Figure 8. Speed deviation during 1-Ф to ground fault.
Similarly, during the 2-Φ to ground fault, the MRC has a better response in mitigation of oscillation
as compared to the PI. As depicted in Figure 9, the MRC has smoothed the voltage response abruptly
p.u
)v)p.u
 Speed Deviati
Energies 2019, 12, 3653 12 of 15

after fault clearance at t = 2.5 s, while the PI controller failed to show the robustness. The oscillation is
completely damped out at t = 12 s, as shown in Figure 7. Similarly, speed deviation becomes zero at
Figure
t = 2 s in the case of the MRC, 8. Speed
while the deviation during
PI vanishes the 1-Ф to groundatfault.
oscillations t = 8 s.
Voltage (v) p.u

Figure 9.
Figure Load bus
9. Load bus voltage 2-Ф
2-Φ to
to ground
ground fault.
fault.
Energies 2019, 12, x FOR PEER REVIEW 12 of 15
5.2. Symmetrical Faults
5.2. Symmetrical Faults
In the
In thecase
case of asymmetrical
of an an asymmetrical fault, fault,
the 3-Φthe 3-Ф to fault
to ground ground fault is
is applied applied the
to evaluate to evaluate
performance the
performance
of the proposed of the proposed
controller, as controller,
depicted inasFigure
depicted10. in Figure
The MRC10. The
has MRC has
a better a better
approach approach
toward the
toward the mitigation of oscillations occurring after the voltage dip, as shown
mitigation of oscillations occurring after the voltage dip, as shown in Figure 11. The oscillation in Figure 11. Theis
oscillation damped
completely is completely
out at damped outthe
t = 2.5 s in 𝑡 =of2.5
atcase thes MRC,
in thewhile
case the
of PI
thehas
MRC, while the oscillations
low amplitude PI has low
amplitude
till t = 8 s. oscillations till 𝑡deviation
Similarly, speed = 8 s. Similarly, speed
is subjected deviation
to larger is subjected
amplitude to larger amplitude
high frequency oscillation high
and
frequency oscillation and higher settling time using the PI, while the proposed
higher settling time using the PI, while the proposed controller settles the response to zero at controller settles
t = 3thes.
response to zero at 𝑡 = 3 s.
Voltage (v) p.u

Figure 10. Load

Figure 10. Load bus
bus voltage
voltage 3-Φ
3-Ф to ground
ground fault.
fault.
w) p.u
Speed Deviation (
Energies 2019, 12, 3653 13 of 15
Figure 10. Load bus voltage 3-Ф to ground fault.

w) p.u
Speed Deviation (

Figure 11.
Figure Speeddeviation
11. Speed deviation during
during 3-Ф
3-Φto
toground
groundfault.
fault.

Besides other performance evaluation parameter, Table 1 shows the performance indices of both
Besides other performance evaluation parameter, Table 1 shows the performance indices of both
controllers during all three types of faults. The lower value of integral square error (ISE) and integral
controllers during all three types of faults. The lower value of integral square error (ISE) and integral
absolute error (IAE) shows better response of controller. The MRC controller has both a lower value of
absolute error (IAE) shows better response of controller. The MRC controller has both a lower value
ISE error and IAE.
of ISE error and IAE.
Table 1. Performance indices. ISE = integral square error; IAE = integral absolute error.
Table 1. Performance indices. ISE = integral square error; IAE = integral absolute error.
1-Φ Fault 2-Φ Faults 3-Φ Fault
Controllers
1-Ф fault 2-Ф faults 3-Ф fault
Controllers ISE IAE ISE IAE ISE IAE
PI 0.0009028ISE 2.432 × 10
IAE−7 ISE
0.001574 6.569IAE
× 10−7 ISE
0.00253 IAE× 10−6
1.07
MRC PI 0.0005561 −7 0.0005802 × 10−7 0.0005248 1.111 × 10−7
1.07e-6
0.00090281.301 × 10
2.432e-7 0.001574 1.41
6.569e-7 0.00253
MRC 0.0005561 1.301e-7 0.0005802 1.41e-7 0.0005248 1.111e-7
6. Conclusions
The mitigation of power oscillations is a demanding task to enhance the stability of the power
system. An intelligent, optimized, and robust design of the UPFC plays a vital role in damping of these
oscillations. With these abilities, the MRC controller has shown tremendous performance in mitigating
power oscillations, enhancing system stability, and improving the capability of the control system.
Moreover, our control scheme shows robustness and intelligence during system perturbation. In order
to show the effectiveness of proposed control scheme, it was compared with a traditional PI controller.
The results clearly show the superiority of the NN-based MRC over the traditional PI controller, as it is
better in oscillation damping. The NN-based MRC diminishes the amplitude of power oscillation,
produced due to fault.
However, the MRC-based on a NN is complex algorithm. Due to its complexity, it requires
considerable amount of time for calculation and implementation. Due to the integration of the
NN controller in the MRC frame, it is associated with several issues and complications. The
complications include selecting the input and output of the controller, as well as training of the NN for
its weight adjustment.
Moreover, soon the proposed control will be extended to more than a 2-area system with a greater
number of generation stations. Additionally, a new control scheme will be designed, and its results
will be compared with both the PI and the MRC.

Author Contributions: W.U., N.Z., and K.Z. propose the main idea of the paper. W.U., M.I., S.U.I., A.T. and A.H
have implemented the mathematical derivations, simulation verifications, and analyses. The paper is written by
W.U. and is revised by N.Z., I.K., A.H., and H.-J.K. All the authors were involved in preparing the final version of
this manuscript. Besides, this whole work is supervised and funded by H.-J.K. and G.-S.P.
Energies 2019, 12, 3653 14 of 15

Funding: This research was supported by Basic Research Laboratory through the National Research Foundations
of Korea funded by the Ministry of Science, ICT and Future Planning (NRF-2015R1A4A1041584).
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A

Table A1. Parametric Value of NN.

Parameters Value
Controller training Epochs 50
Controller Training sample 100
Sampling Interval (sec) 0.05
Maximum interval value (sec) 20
Minimum interval value (sec) 5
Controller training segment 50
Number of hidden layers 2
No. of delayed Plant Inputs 1
No. of delayed Plant outputs 1

References
1. Wu, W.; Chen, Y.; Fei, Y.; Zhen, H.; Zhou, B.; Wang, Z.; Chen, W. A novel damping strategy for
low-frequency oscillation suppression with MMC-type unified power flow controller. In Proceedings
of the IEEE International Conference on Industrial Technology, Lyon, France, 20–22 February 2018.
2. Gandoman, F.H.; Ahmadi, A.; Sharaf, A.M.; Siano, P.; Pou, J.; Hredzak, B.; Agelidis, V.G. Review of FACTS
technologies and applications for power quality in smart grids with renewable energy systems. Renew.
Sustain. Energy Rev. 2018, 82, 502–514. [CrossRef]
3. Kannayeram, G.; Manoharan, P.S.; Iruthayarajan, M.W.; Sivakumar, T. UPFC damping controller design
using multi-objective evolutionary algorithms. Int. J. Bus. Intell. Data Min. 2018, 13, 52–74. [CrossRef]
4. Fortes, E.; Macedo, L.; Araujo, P.B.; Romero, R. A VNS algorithm for the design of supplementary damping
controllers for small-signal stability analysis. Int. J. Electr. Power Energy Syst. 2018, 94, 41–56. [CrossRef]
5. Banaei, M.R.; Toloue, H.; Kazemi, F.M.; Oskuee, M.R.J. Damping of power system oscillations using
imperialist competition algorithm in power system equipped by HVDC. Int. J. Ain Shams Eng. J. 2015, 6,
75–85. [CrossRef]
6. Martins, L.F.B.; Araujo, P.B.; De Vargas Fortes, E.; Macedo, L.H. Design of the PI–UPFC–POD and PSS
Damping Controllers Using an Artificial Bee Colony Algorithm. J. Control Autom. Electr. Syst. 2017, 28,
762–773. [CrossRef]
7. Shahriar, M.S.; Shafiullah, M.; Rana, M.J. Stability enhancement of PSS-UPFC installed power system by
support vector regression. Electr. Eng. 2018, 100, 1601–1612. [CrossRef]
8. Shojaeian, S.; Soltani, J.; Arab Markadeh, G. Damping of low-frequency oscillations of multi-machine
multi-UPFC power systems, based on adaptive input-output feedback linearization control. IEEE Trans.
Power Syst. 2012, 27, 1831–1840. [CrossRef]
9. Esmaili, M.R.; Khodabakhshian, A.; Bornapour, M. A new coordinated design of UPFC controller and PSS
for improvement of power system stability using CPCE algorithm. In Proceedings of the IEEE Conference
on Electrical Power and Energy EPEC, Ottawa, ON, Canada, 12–14 October 2016.
10. Pandey, R.K.; Gupta, D.K. Knowledge domain states mapping concept for controller tuning in an
interconnected power network. Int. J. Electr. Power Energy Syst. 2016, 80, 160–170. [CrossRef]
11. Tavakoli, A.R.; Seifi, A.R.; Arefi, M.M. Fuzzy-PSS and fuzzy neural network non-linear PI controller-based
SSSC for damping inter-area oscillations. Trans. Inst. Meas. Control 2016, 40, 733–745. [CrossRef]
12. Moravej, Z.; Pazoki, M.; Khederzadeh, M. New Pattern-Recognition Method for Fault Analysis in Transmission
Line With UPFC. IEEE Trans. Power Deliv. 2015, 30, 1231–1242. [CrossRef]
13. Mahmud, M.A.; Pota, H.R.; Hossain, M.J. Full-order nonlinear observer-based excitation controller design
for interconnected power systems via exact linearization approach. Int. J. Electr. Power Energy Syst. 2012, 41,
54–62. [CrossRef]
Energies 2019, 12, 3653 15 of 15

14. Parimi, A.M.; Elamvazuthi, I.; Kumar, A.V.P.; Cherian, V. Fuzzy logic based control for IPFC for damping
low-frequency oscillations in the multimachine power system. In Proceedings of the 2015 IEEE IAS Joint
Industrial and Commercial Power Systems/Petroleum and Chemical Industry Conference (ICPSPCIC),
Hyderabad, India, 19–21 November 2015.
15. Singh, B.; Mukherjee, V.; Tiwari, P. A survey on impact assessment of DG and FACTS controllers in power
systems. Renew. Sustain. Energy Rev. 2015, 42, 846–882. [CrossRef]
16. El-Zonkoly, A. Optimal sizing of SSSC controllers to minimize transmission loss and a novel model of SSSC
to study transient response. Electr. Power Syst. Res. 2008, 78, 1856–1864. [CrossRef]
17. Rodríguez, O.; Medina, A.; Andersson, G. Closed-form analytical characterization of non-linear oscillations in
power systems incorporating a unified power flow controller. IET Gener. Transm. Distrib. 2015, 9, 1019–1032.
[CrossRef]
18. Wang, H. A unified model for analysis of FACTS Devices in Damping Power System Oscillations. Part III:
Unified Power Flow Controller. IEEE Trans. Power Deliv. 2000, 15, 978–983. [CrossRef]
19. Torkzadeh, R.; Nasrazadani, H.; Aliabad, A.D. A genetic algorithm optimized fuzzy logic controller for
UPFC in order to damp of low-frequency oscillations in power systems. In Proceedings of the 2014 22nd
Iranian Conference on Electrical Engineering (ICEE), Tehran, Iran, 20–22 May 2014.
20. Li, Z.; Xia, Y.; Su, C.Y.; Deng, Y.; Fu, J.; He, W. Missile guidance Law Based on robust model predictive control
using Neural Network Optimization. IEEE Trans. Neural Netw. Learn. Syst. 2015, 26, 1803–1809. [CrossRef]
[PubMed]
21. Xu, R.; Tao, Y.; Lu, Z.; Zhong, Y. Attention-Mechanism-Containing Neural networks for high resolution
remote sensing image classification. Electronics 2018, 10, 1602. [CrossRef]
22. Barone, E.R.; Salerno, V.; Siniscalchi, S.M. An introductory study on deep neural networks for high resolution
areal images. AIP Conf. Proc. 2013, 1558, 1232. [CrossRef]
23. Sinniscalchi, S.M.; Salerno, V.M. Adaptation to new microphones using artificial neural networks with
trainable activation functions. IEEE Trans. Neural Netw. Learn. Syst. 2017, 28, 1959–1965. [CrossRef]
24. Hautamaki, V.; Sinniscalchi, S.M.; Behravan, H.; Salerno, V.M.; Kukanov, I. Boosting universal speech
attributes classification with deep Neural Network for foreign accent characterization. In Proceedings of the
16th Annual Conference of the International Speech Communication Association, Dresden, Germany, 6–10
September 2015.
25. Fu, Y.; Chai, T. Neural Network Based nonlinear adaptive dynamical decoupling control. IEEE Trans. Neural
Netw. 2007, 18, 921–925. [CrossRef]
26. Zeb, K.; Mehmood, C.A.; Khan, B.; Ali, S.M.; Jadoon, A.M.; Uddin, W. Fault tolerant speed regulation
of induction motor using artificial neural network. In Proceedings of the IEEE Conference on Emerging
Technologies, Peshawar, Pakistan, 19–20 December 2015.
27. Chae, S.; Kwon, S.; Lee, D. Predicting infectious Disease Using Deep learning and Big Data. Int. J. Environ.
Res. Public Health 2018, 15, 1596. [CrossRef] [PubMed]
28. Salerno, V.M.; Rabbeni, G. An Extreme learning machine approach to effective energy disaggregation.
Electronics 2018, 7, 235. [CrossRef]
29. Douratsos, I.; Gomm, J.B. Neural Network based model reference adaptive control for process with time
delay. Int. J. Inf. Syst. Sci. 2006, 3, 161–179.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).