applied

sciences
Article
Artificial Intelligence-Based Controller for DC-DC
Flyback Converter
Muhammad Arslan Shahid 1 , Ghulam Abbas 1, * , Mohammad Rashid Hussain 2 ,
Muhammad Usman Asad 3 , Umar Farooq 3,4 , Jason Gu 3 , Valentina E. Balas 5 ,
Muhammad Uzair 6 , Ahmed Bilal Awan 7 and Tanveer Yazdan 1
1 Electrical Engineering Department, The University of Lahore, Lahore 54000, Pakistan;
[Link]@[Link] (M.A.S.); tanveeryazdan23@[Link] (T.Y.)
2 College of Computer Science, King Khalid University, Abha 61421, Saudi Arabia; humohammad@[Link]
3 Electrical and Computer Engineering Department, Dalhousie University, Halifax, NS B3H 4R2, Canada;
usmanasad01@[Link] (M.U.A.); [Link]@[Link] (U.F.); [Link]@[Link] (J.G.)
4 Electrical Engineering Department, University of the Punjab, Lahore 54590, Pakistan
5 Automatics and Applied Software Department, “Aurel Vlaicu” University of Arad, Arad 310130, Romania;
[Link]@[Link]
6 Department of Electrical Engineering, Islamic University of Madinah, Madinah 41411, Saudi Arabia;
uzair91@[Link]
7 Department of Electrical Engineering, College of Engineering, Majmaah University, Al-Majmaah 11952,
Saudi Arabia; [Link]@[Link]
* Correspondence: [Link]@[Link] or engrgabbas@[Link]; Tel.: +92-304-285-4035




Received: 14 October 2019; Accepted: 18 November 2019; Published: 26 November 2019


Abstract: This paper presents an intelligent voltage controller designed on the basis of an adaptive
neuro-fuzzy inference system (ANFIS) for a flyback converter (FC) working in continuous conduction
mode (CCM). The union of fuzzy logic (FL) and adaptive neural networks (ANN) makes ANFIS more
robust against model parameters’ uncertainties and perturbations in input voltage or load current.
ANFIS inherits the advantages of structured knowledge representation from FL and learning capability
from NN. Comparative analysis showed that the ANFIS controller offers not only the superior
transient response characteristics, but also excellent steady-state characteristics compared to those of
the FL controller (FLC) and proportional–integral–derivative (PID) controllers, thus validating its
superiority over these traditional controllers. For this purpose, MATLAB/Simulink environment-based
simulation results are presented for validation of the proposed converter compensated system under
all operating conditions.

Keywords: ANFIS; flyback converter (FC); robust; fuzzy logic control (FLC); PID

1. Introduction
DC-DC converters are employed in a variety of applications such as industrial controls, audio
applications, power adapters and chargers, electric vehicles, electronic appliances, power supplies,
renewable systems, aerospace equipment, and many other modern types of equipment that operate on
DC [1–3]. Flyback converters (FCs) are a frequently used type of DC converter which may operate over
a broad range of unregulated DC voltage. FCs serve as energy storage as well as converter isolation
because of a choke in their topology. This results in reduced noise interference and provided load
protection [4,5]. FCs are able to achieve high efficiency, high stability, small size, lightweight, less cost,
etc., and have more power capacity (usually in the range of 20 to 200 W [6]) than other fundamental
DC-DC converters. They are preferred for their outstanding features including output voltage
waveform shape, power factor control optimization, and device miniaturization [7–10]. FCs exhibit

Appl. Sci. 2019, 9, 5108; doi:10.3390/app9235108 [Link]/journal/applsci

Appl. Sci. 2019, 9, 5108 2 of 22

non-minimum phase characteristics, i.e., the existence of a right half plane zero (RHPZ) in the voltage
transfer function. This complicates the dynamics of FCs, as it results in increased gain and introduces
additional phase lag, thus behaving as a pole. Thus, there is a need to design a feedback controller able
to ensure the static and dynamic performance of the converter [11].
There are several analog and digital methods, such as PID, fuzzy logic (FL), and sliding mode
control (SMC), for regulating the output of DC-DC FCs [12]. The feedback controller in a DC-DC
converter is mainly responsible for delivering regulated supply, ideally with no steady-state error,
less overshoot, and fast dynamic response while maintaining the highest possible efficiency [13,14].
Sliding mode control (SMC) based on the equivalent control method with a constant frequency has been
applied to FCs in References [5,14]. In Reference [15], a peak current control technique for a flyback
converter’s power factor correction was proposed. Reference [16] introduced average current mode
control (ACM) for an FC, which employed an outer voltage feedback loop responsible for maintaining
constant output voltage, and an inner loop engaged in sensing the input current. Quasi-resonant (QR)
controllers for FCs have been presented in References [8,17,18]. PI controllers [19] and PID
controllers [20,21] for FCs have also been reported in the literature. For example, in Reference [22],
an FC with a photovoltaic (PV) panel was controlled by a PI compensator. The problem with these
controllers is that they may not be robust against disturbance and uncertainty. Dynamic modeling
and quantitative controller design of a current-mode controlled FC with optocoupler isolation were
presented in Reference [23]. In Reference [24], the FC operated with a natural switching surface (NSS)
control. Derivation and implementation of NSS with the operation of FCs in boundary conduction
mode (BCM) were presented. A detailed review of the controllers to date, including pulse frequency
modulation (PFM) control, adaptive peak current value control, current estimation control, and duty
cycle control for FCs is presented in Reference [25].
Owing to features such as easier implementation and robustness characteristics, fuzzy logic
controllers (FLCs) have also been employed to control FCs. FLCs are based on natural language
and employ fuzzy “if-then” rules and fuzzy reasoning. They are tolerant of imprecise data, require
fewer tuning parameters, and are flexible in such a way that they add more functionality without
re-starting from the start [26,27]. Such an FLC can be employed to implement a central action for
dealing with the variable structure nature of an FC. In Reference [28], an FLC-based FC is presented
where the FLC responds rapidly to changes in load current or input voltage to ensure better load
regulation. The controller also offers robustness with a good dynamic response. In Reference [29],
a comparison was made between fuzzy and PI-based voltage controllers for an integrated buck flyback
converter. Similarly, in Reference [30], the performance of the FC compensated system was analyzed
via a PID controller and FLC. In these references, FLCs have outshone PIDs in terms of performance.
Although an FL can make decisions based on the knowledge supplied to it, due to the lack of a training
mechanism, it cannot make its knowledge adaptive.
More intelligence can be introduced to fuzzy rules by retuning them with artificial neural networks
(ANN). ANFIS is the fusion of ANN and FL and utilizes the salient features of neural and fuzzy
networks, respectively. The main advantage of ANFIS is that it converges quickly since it minimizes
the search space dimensions of the backpropagation method used in NN [31–33]. An ANFIS-based
maximum power point tracking (MPPT) system for a PV module with a DC-DC converter is presented in
References [34–37]. In Reference [38], the authors proposed an ANFIS-based MPPT system comprising
a PV module connected to a load through a DC-DC Ćuk converter to track the maximum power point
(MPP). The MATLAB/Simulink environment was used to simulate the proposed model. Very limited
research can be found in the literature regarding the control of switching converters using ANFIS.
This paper thus proposes an ANFIS-based controller to regulate and improve the output voltage of FC.
Line and load regulation and response to reference voltage changes are presented and the static and
dynamic performance of the proposed controller is analyzed. Furthermore, its performance was also
compared with FLC and PID controllers to validate its superiority.
Appl. Sci. 2019, 9, 5108 3 of 22

The paper is organized as follows. Section 2 describes the modeling of an FC in continuous

Appl. Sci. 2019, 9, x FOR PEER REVIEW 3 of 22
conduction mode (CCM). The design parameter selection of the FC used throughout the paper for the
design of the controllers is described in Section 3. Section 4 presents the detailed controller design
the design of the controllers is described in Section 3. Section 4 presents the detailed controller design
procedure for the FLC, ANFIS-based controller, and PID controller. Section 5 shows the results obtained
procedure for the FLC, ANFIS-based controller, and PID controller. Section 5 shows the results
after performing the simulations of FC with feedback controllers. Finally, the results are concluded in
obtained after performing the simulations of FC with feedback controllers. Finally, the results are
Section 6.
concluded in Section 6.
2. Modeling of Flyback Converter
2. Modeling of Flyback Converter
The block diagram of the complete system, along with the circuit diagram of the FC, is presented
The block
in Figure 1. FC diagram
behaves of the
like complete
a plant to besystem, alongTowith
controlled. the circuit
develop diagram
a controller forofit,the
itsFC, is presented
dynamics must
in Figure
be known.
1. FC behaves like a plant to be controlled. To develop a controller for it, its dynamics must
be known.

Figure 1. Block diagram of the complete system.

Figure 1 shows the FC circuit consisting of a DC voltage source “E”, a power MOSFET “S”,
transformer
transformer“T” “T”forfor
isolation purposes,
isolation diodediode
purposes, “D”, capacitor “C”, and“C”,
“D”, capacitor loadand
and magnetizing inductance
load and magnetizing
of transformer “L ”.
inductance of transformer
M The DC“LM”. The DC voltage source can be the output of an uncontrolledfiltering)
voltage source can be the output of an uncontrolled rectifier (after rectifier
that
(afterconverts ACthat
filtering) to DC. The switching
converts AC to DC. MOSFET S is fed with
The switching a PWM Sat is
MOSFET high
fedfrequency.
with a PWM The purpose
at high
of the transformer
frequency. is also of
The purpose to the
provide better matching
transformer is also tobetween
provideinput
betterand output voltage
matching betweenand current
input and
requirements. The diode D rectifies the secondary voltage of the transformer
output voltage and current requirements. The diode D rectifies the secondary voltage of the T. The capacitor C is for
filtration
transformer purposes.
T. The capacitor C is for filtration purposes.
For determining
determiningFCFC dynamics
dynamics in CCM,
in CCM,a state-space averaging
a state-space technique
averaging was employed
technique [5,24,28].
was employed
Inductor
[5,24,28].current ILM and
Inductor voltage
current across
I LM andthevoltage VC were
capacitoracross theconsidered
capacitoras independent state variables
V C were considered as
of the state vector X. Thus,
independent state variables of the state vector X ". Thus,#
ILM
X= (1)
 IVLM 
X= C (1)
 VC 
The independent input vector u is expressed in Equation (2) with input voltage E and voltage
across the diode V D as constraints.

E
u=  (2)
 VD 
Appl. Sci. 2019, 9, 5108 4 of 22

The independent input vector u is expressed in Equation (2) with input voltage E and voltage
across the diode VD as constraints. " #
E
u= (2)
VD
The output vector consists of input current IS and output voltage VO , since in the modeling
procedure, input and output are to be modeled, which are dependent quantities.
" #
IS
Y= (3)
VO

CCM operation occurs in two modes:

Mode 1: When the switching MOSFET S is in ON state, current flows through the primary side of
the transformer with magnitude IS , and the diode is reversed biased at the secondary side. Therefore,
no current flows in the secondary of the transformer. The capacitor’s stored energy supplies the output
voltage. The state conditions for Mode 1 are:

dILM E I R
= − S S (4)
dt LM LM

dVC VC
=− (5)
dt (R + RC )C
VC R
VO = (6)
R + RC
IS = ILM (7)
dI
where LM is the magnetizing inductance of transformer, dtLM is the rate of change of current in the
primary of the transformer, E is the supply voltage, IS is the supply current, RS is the resistance of the
dV
switch S, C is the capacitance, dtC is the rate of change of capacitor voltage, VC is the capacitor voltage,
R is the load resistance, RC is the internal resistance of capacitor, and VO is the output voltage. Thus,
the state-space representation for Mode 1 is as follows:
 dILM
  RS " # " 1 #" #
.   − L 0  ILM 0 E
+ LM
 dt
x = Ax + Bu ⇒  dVC
 =  M
1
 (8)
  0 − C(R+ VC 0 0 VD

dt R C)

" # " #" # " #" #

IS 1 0 ILM 0 0 E
y = Cx + Du ⇒ = + (9)
VO 0 − R+RR VC 0 0 VD
C
.
where x is a state vector, x is the state vector derivative with regard to time, A is the system matrix, B is
the input matrix, u is the input or control vector, y is the output vector, C is the output matrix, and D is
the disturbance matrix.
Mode 2: When the switching MOSFET S is in OFF state, current flows through the secondary of
the transformer and the diode is forward biased. The energy stored in the magnetizing inductance LM
is transferred to the load and output capacitor.
The state conditions for Mode 2 are:

dILM n2 ILM RRC nRVC nVD

=− − − (10)
dt LM (R + RC ) LM (R + RC ) LM

dVC nILM R VC
= − (11)
dt C(R + RC ) C(R + RC )
nRRC R
VO = ILM + VC (12)
R + RC R + RC
Appl. Sci. 2019, 9, 5108 5 of 22

IS = 0 (13)

where n is the turn ratio of the transformer. Thus, the state-space representation for Mode 2 is as follows:
2
  − L n(RRR
 dILM
  nR
"
C
− LnM
# " #" #
−L  ILM 0 E
M +RC ) M (R+RC )
 dt  = 
 VC + 0 (14)

dVC

nR 1 0 VD
− C(R+
  
dt C ( R + RC ) RC )

" #  " # " #" #

IS  0 0  ILM 0 0 E
=  nRRC R
 + (15)
VO R+RC R+RC
 V
C 0 0 VD
Now, the state-space averaged model for CCM can be written as [5]:
  RS 2
  − L S1 − L n(RRR
"
dILM
S2 − L (RnR+R ) S2  ILM
 C
#
M +RC )
 dt
  =  M M C
dVC

nR 1  VC
S − C(R+
  
dt C(R+RC ) 2 RC ) (16)
− L1M S1 − LnM S2
" #" #
E
+
0 0 VD
" #  " #
IS  0 0  ILM
=  nRRC R
 (17)
VO R+RC S2
 V
R+RC C

Leading towards the derivation of the transfer function, the small-signal model was employed to
approximate the behavior of the FC model. In the state-space model (SSM), state variables and control
input are comprised of DC and AC quantities. Capital subscripts represent DC steady-state values,
and small subscripts represent AC perturbations.

iLM = ILM + ilm (18)

vO = VO + vo (19)

D = ∆+d (20)

Inserting the above quantities into the state-space averaged model (SSAM) and hence taking
Laplace transforms results in the voltage transfer function expressed in Equation (21).
 
ELM ∆
− s + ER
vo n(1−∆)2
Hvo (s) = =   (21)
d (LM RC)s2 + (LM )s + Rn2 (1 − ∆)2

The negative sign in the numerator with the term ‘s’ indicates that it is a non-minimum phase
system, i.e., it has a right half plane zero (RHPZ).

3. Design of FC with Specifications

Let us consider the efficiency of FC to be 90%.

1  2
ELM = LM ILMmax − ILMmin (22)
2
1  2
Pout = LM ILMmax − ILMmin × fS × η (23)
2
For a 90% efficiency, η = 0.9.

Pout Pout
η= and Pin = = 1.11Pout (24)
Pin 0.9
Appl. Sci. 2019, 9, 5108 6 of 22

VO 2
Pout = (25)
R
Substituting Equation (25) into Equation (24) gives Equation (26).
 2
1
V 2 2 LM ILMmax − ILMmin
Pin = 1.11 × O = (26)
R TS

where η is the practical efficiency of the design, ILMmax and ILMmin are the maximum and minimum
magnitudes of current ILM , respectively, and TS is the reciprocal of fS . For calculating the inductance
and capacitance, the following formulae can be used:

VS × D
LM = (27)
2π × fS × ∆I

VS × D2
C= (28)
n(1 − D)R × fS × ∆V
The typical parameter values used throughout the paper are tabulated in Table 1.

Table 1. Designed values of flyback controller (FC) components.

Parameter/Component Value
Input voltage E 12 V
Output voltage VO 24 V
Transformer power 120 W
Magnetizing inductance of transformer LM 250 µH
Output filter capacitor C 200 µF
Load resistance R 10 Ω
Duty cycle D 0.5
Transformer turn ratio n1 /n2 = N 0.5
Switching frequency fS 100 kHz

With the above-mentioned design specifications, the FC with unity feedback showed output
voltage across the resistive load with a peak value of 30.76 V and a steady-state voltage of 22.19 V,
Appl. Sci. 2019, 9, x FOR PEER REVIEW 7 of 22
as shown in Figure 2. An intelligent controller needed to be introduced into the loop.

Figure 2. The output of FC with unity feedback.

Figure 2. The output of FC with unity feedback.

4. Controller Design
After describing the flyback converter modeling, a detailed design procedure of feedback
controllers, i.e., FLC, ANFIS-based controller, and PID controller, is presented in this section.
Appl. Sci. 2019, 9, 5108 7 of 22
Figure 2. The output of FC with unity feedback.

4. Controller Design
4.
After describing
After describing the flyback
flyback converter
converter modeling,
modeling, a detailed design procedure
procedure of of feedback
feedback
controllers, i.e.,
controllers, i.e., FLC,
FLC, ANFIS-based
ANFIS-based controller,
controller,and
andPID
PIDcontroller,
controller,isispresented
presentedininthis
thissection.
section.

4.1.
4.1. FLC
FLC
Fuzzy
Fuzzy logic
logic(FL)
(FL)isisanan
extension
extensionof multivalued
of multivalued logic. It is It
logic. related to thetotheory
is related of fuzzy
the theory of sets,
fuzzy which
sets,
refers to classes of objects with blunt boundaries in which membership is a matter
which refers to classes of objects with blunt boundaries in which membership is a matter of degree. of degree. It emulates,
unlike crisp unlike
It emulates, data, the ability
crisp data,tothe
reason
abilityand uses approximate
to reason data for finding
and uses approximate solutions.
data for FLCs are
finding solutions.
knowledge-based
FLCs are knowledge-based controllers that consist of linguistic if-then rules, whichbycan
controllers that consist of linguistic if-then rules, which can be constructed using
be
the knowledge of experts in a particular field of interest. FLCs have shown
constructed by using the knowledge of experts in a particular field of interest. FLCs have shown their their ability in many
applications,
ability in many especially in complex
applications, nonlinear
especially systems,
in complex for which
nonlinear analytical
systems, formodeling is difficult
which analytical [39,40].
modeling
Figure 3 illustrates the block diagram of a traditional FLC system, which consists
is difficult [39,40]. Figure 3 illustrates the block diagram of a traditional FLC system, which consists of four components:
fuzzification interface fuzzification
of four components: (FI), rule-baseinterface
(RB), inference mechanism
(FI), rule-base (RB),(IM), and defuzzification
inference mechanism (IM), interface
and
(DI). The FI classifies
defuzzification the inputs
interface intoFIsuitable
(DI). The classifieslinguistic values/sets.
the inputs into suitableThe linguistic
RB is comprised of fuzzy
values/sets. Thesets
RB
(database) and fuzzy control rules. The IM is the kernel of the FLC, which is
is comprised of fuzzy sets (database) and fuzzy control rules. The IM is the kernel of the FLC, which capable of simulating
human
is capabledecision makinghuman
of simulating based decision
on fuzzymaking
concepts andon
based inferring fuzzy control
fuzzy concepts actionsfuzzy
and inferring usingcontrol
fuzzy
relations
actions using fuzzy relations and rules of inference. The DI makes conclusions from the IM and input
and rules of inference. The DI makes conclusions from the IM and gives an output, i.e., gives
for the planti.e.,
an output, [41,42].
input for the plant [41,42].

Figure 3.
Figure 3. Fuzzy logic controller (FLC) block diagram.
diagram.

Figure presents the

Figure 4 presents themodel
modelofofananFCFC with
with a fuzzy
a fuzzy logic
logic controller
controller in MATLAB/Simulink.
in MATLAB/Simulink. The
The output
output voltage
voltage is measured
is measured and is and is compared
compared with thewith the reference
reference value,
value, and thus,and thus,isan
an error error is
generated.
generated. Two namely
Two quantities, quantities, namely
error error e andofderivative
e and derivative of the
the error Δe, areerror ∆e, are
generated asgenerated
the inputsasforthe
theinputs
fuzzy
for thecontroller.
logic fuzzy logic controller.
The The controller
fuzzy logic fuzzy logicprocesses
controllerthe
processes the based
quantities quantities based on membership
on membership functions
functions (MFs) and fuzzy rules to generate an output, which is the duty cycle d. This signal is
compared with a 100 kHz triangular waveform to generate duty cycle pulses for the gate of MOSFET S.
Input MFs are presented in Figure 5a,b. Five MFs were assigned to each input variable for the
universe of discourse of −24 to 24. The output MFs of a Sugeno type FIS are constants or linear
functions. In this case, constants were selected. Finite values were assigned to each MF, as shown in
Figure 5c. FLC works based on fuzzy rules. Table 2 illustrates the fuzzy rules.

Table 2. Rules for the FLC.

e
Neg. Big Neg. Small Zero Pos. Small Pos. Big
∆e
Neg. Big 0 0 0 0.25 0.5
Neg. Small 0 0 0.25 0.5 0.75
Zero 0 0.25 0.5 0.75 1
Pos. Small 0.75 0.5 0.75 1 1
Pos. Big 0.5 0.75 1 1 1
(MFs) and fuzzy rules to generate an output, which is the duty cycle d. This signal is compared with
a 100 kHz triangular waveform to generate duty cycle pulses for the gate of MOSFET S.

Appl. Sci. 2019, 9, x FOR PEER REVIEW 8 of 22

Appl. Sci. 2019, 9, 5108 8 of 22

(MFs) and fuzzy rules to generate an output, which is the duty cycle d. This signal is compared with
a 100 kHz triangular waveform to generate duty cycle pulses for the gate of MOSFET S.

Figure 4. FC model with FLC in MATLAB/Simulink.

Input MFs are presented in Figure 5a,b. Five MFs were assigned to each input variable for the
universe of discourse of −24 to 24. The output MFs of a Sugeno type FIS are constants or linear
functions. In this case, constants were selected. Finite values were assigned to each MF, as shown in
Figure 4. FC model with FLC in MATLAB/Simulink.
Figure 5c. FLC works based on fuzzy
Figure 4. FC rules.
model Table 2 illustrates
with FLC the fuzzy rules.
in MATLAB/Simulink.

Input MFs are presented in Figure 5a,b. Five MFs were assigned to each input variable for the
universe of discourse of −24 to 24. The output MFs of a Sugeno type FIS are constants or linear
functions. In this case, constants were selected. Finite values were assigned to each MF, as shown in
Figure 5c. FLC works based on fuzzy rules. Table 2 illustrates the fuzzy rules.

(a) (b)

(a) (b)

(c)

Figure 5. MFs
MFs of
of (a)
(a) input ∆e, and (c) output variable duty cycle d.
input variable error e, (b) input variable Δe,

The 3D graphical representation of the rule

Table surface
2. Rules relating
for the FLC. output value for any combination of
the two inputs is presented in Figure 6.
e Neg. Big Neg. Small (c) Zero Pos. Small Pos. Big
4.2. ANFIS-Based Controller
Figure 5. MFs of (a) input variable error e, (b) input variable Δe, and (c) output variable duty cycle d.
The adaptive neuro-fuzzy inference system (ANFIS) is the combination of fuzzy logic (FL) and
Table
neural networks (NN) developed by Jang 2. Rules
in 1993 for the
[43,44]. [Link] and NN, it inherits their advantages,
From
structured knowledge representation,
e Neg. Big and learning
Neg. Small capabilities,
Zero [Link].
Small Pos. In BigFL, membership
Appl. Sci. 2019, 9, x FOR PEER REVIEW 9 of 22

Δe
Neg. Big 0 0 0 0.25 0.5
Appl. Sci. 2019, 9, 5108Neg. Small 0 0 0.25 0.5 0.75 9 of 22
Zero 0 0.25 0.5 0.75 1
Pos. Small 0.75 0.5 0.75 1 1
functions and their distribution and fuzzy rules are obtained based on expert knowledge. In ANFIS,
Pos. Big 0.5 0.75 1 1 1
MFs and rules obtained from the FLC are further tuned using neural networks to generate optimized
The 3D graphical representation of the rule surface relating output value for any combination of
MFs, and thus rules [36,44,45].
the two inputs is presented in Figure 6.

Figure 6.
Figure 6. Rules
Rules representation
representation by
by the
the 3D
3D surface
surface view.
view.

ANFIS uses NN
4.2. ANFIS-Based to adjust the parameters of MFs while tuning fuzzy rules. The ANN part of the
Controller
ANFIS contributes to the reduction of error, thus optimizing the parameters. On the other hand, the FI
The adaptive neuro-fuzzy inference system (ANFIS) is the combination of fuzzy logic (FL) and
part of the ANFIS deals with uncertainty efficiently. The purpose of ANFIS is to apply a hybrid learning
neural networks (NN) developed by Jang in 1993 [43,44]. From FL and NN, it inherits their
algorithm, adopting input-output data sets and then accomplishment with a desired input-output
advantages, structured knowledge representation, and learning capabilities, respectively. In FL,
mapping, to identify the optimal MFs and other parameters of the equivalent FIS [44–46].
membership functions and their distribution and fuzzy rules are obtained based on expert
ANFIS possesses a layered architecture similar to NN. Nodes in the layers can be adaptive or
knowledge. In ANFIS, MFs and rules obtained from the FLC are further tuned using neural networks
fixed. Adaptive and fixed nodes are illustrated by squares and circles in the architecture diagram,
to generate optimized MFs, and thus rules [36,44,45].
respectively. The general architecture of ANFIS (with five layers, as we considered in our work) is
ANFIS uses NN to adjust the parameters of MFs while tuning fuzzy rules. The ANN part of the
illustrated in9,Figure
Appl. Sci. 2019, 7 [43,44,47,48].
x FOR PEER REVIEW 10 of 22
ANFIS contributes to the reduction of error, thus optimizing the parameters. On the other hand, the
FI part of the ANFIS deals with uncertainty efficiently. The purpose of ANFIS is to apply a hybrid
learning algorithm, adopting input-output data sets and then accomplishment with a desired input-
output mapping, to identify the optimal MFs and other parameters of the equivalent FIS [44–46].
ANFIS possesses a layered architecture similar to NN. Nodes in the layers can be adaptive or
fixed. Adaptive and fixed nodes are illustrated by squares and circles in the architecture diagram,
respectively. The general architecture of ANFIS (with five layers, as we considered in our work) is
illustrated in Figure 7 [43,44,47,48].

Figure 7. Adaptive neuro-fuzzy inference system (ANFIS) general architecture with two inputs (with
five membership functions
functions for
for each
each input),
input), one
one output,
output, and
and 25
25 rules.
rules.

The first order Sugeno type with two inputs “x”, i.e., e and “y”, i.e., Δe, and the rules for a Sugeno
based ANFIS are expressed by Equations (29) and (30). The construction of two rules is presented
here for the sake of saving space.
If X is A 1 and Y is B1 , then
Appl. Sci. 2019, 9, 5108 10 of 22

The first order Sugeno type with two inputs “x”, i.e., e and “y”, i.e., ∆e, and the rules for a Sugeno
based ANFIS are expressed by Equations (29) and (30). The construction of two rules is presented here
for the sake of saving space.
If X is A1 and Y is B1 , then
f1 = p1 x + q1 y + r1 (29)

If X is A2 and Y is B2 , then
f2 = p2 x + q2 y + r2 (30)

4.2.1. Layer 1
Layer 1 has adaptive nodes (squares). Input MF values are calculated in this layer. The output of
nodes in this layer is a membership grade of inputs.
Output O1,i for node i = 1, 2, · · · , 5

O1,i = µAi (x) (31)

Output O2,i for node i = 1, 2, · · · , 5

O2,i = µBi ( y) (32)

where x and y are the inputs to the node i; Ai and Bi are linguistic labels such as negative, zero,
and positive; and µAi (x) µBi ( y) are the MFs. The number of nodes is the same as the fuzzy sets into
which each input is quantified. Ai and Bi are determined and adjusted adaptively during the training
process by ANFIS.

4.2.2. Layer 2
In Layer 2 (circle), nodes are fixed. T-norm operators such as “min”, “product”, “fuzzy AND”,
and so on are employed. In this layer, the firing strength wi of each rule is computed and is given by
Equation (33).
wi = µAi (x).µBi ( y) i = 1, 2, · · · , 5 (33)

4.2.3. Layer 3
In Layer 3 (circle), nodes are fixed. Rules are normalized. The output from this layer that becomes
the input of Layer 4 is said to be normalized firing strength wi which is calculated by Equation (34).
w
wi = P i i = 1, 2, · · · , 5 (34)
wi

4.2.4. Layer 4
Layer 4 consists of adaptive nodes (squares). Consequent parameters or the output is determined,
and the function of each node becomes the product of Layer 3’s output and Sugeno fuzzy if-then rules
fi , i.e., a simple linear equation. The output from each rule is calculated using Equation (35).

wi · fi = wi (pi x + qi y + ri ) (35)

where the quantities pi , qi , and ri are the consequent parameters that are adjusted during the training
procedure using a least-squares error algorithm (LSA).
Appl. Sci. 2019, 9, 5108 11 of 22

4.2.5. Layer 5
Layer 5 is the last layer for the output and is comprised of a single node (fixed (circle)). All the
outputs from Layer 4 from each rule are added up here, and a single output, i.e., the duty cycle is
generated and is calculated by Equation (36).
P
X wi · fi
wi · f i = P (36)
wi

In this paper, a hybrid algorithm (HA) was used for the training purpose. The HA was the
hybrid combination of the back-propagation algorithm (BPA) and least-squares error algorithm (LSA).
It determined the antecedent (input) and consequent (output) parameters. First, the consequent
(output) parameters were revised using the LSA and, antecedent (input) parameters were then updated
using BPA. Furthermore, the gradient descent algorithm (GDA) back-propagated the errors that still
existed [45].
The squared error k AX − B k2 was actually minimized by LSA, where A represents the outputs
produced according to Equation (34), B represents a target output, and X is the unknown consequent
value. X can be obtained using the pseudo-inverse of X, expressed in Equation (37).

−1
X ∗ = ( AT A ) AT B (37)

−1
where AT represents the transposition of A and (AT A) AT is the pseudo-inverse of A if (AT A) is
non-singular. However, as ANFIS is based on data training, in certain iterations, the AT A produced
a singular matrix. To overcome this problem, the recursive LSA was implemented as presented in
Equation (38).
Si a(i+1) aT(i+1) Si


Si+1 = Si −


T
for i = 1, . . . , P

1+a Si a(i+1) 
(38)
(i+1)
T T

Xi + 1 = Xi + S ( i + 1 ) a ( a + 1 ) ( b ( i + 1 ) − a ( i + 1 ) Xi ) 



where X0 = 0, S0 = γI (where γ represents a positive large number and I is the identity matrix), aTi is
the ith line of matrix A, and bTi is the ith element of matrix B and X∗ = XP (where P represents the
training data points number).
After the identification of consequent parameters, the network output was computed and the
expression in Equation (39) was then employed to measure the error for the Pth entry of the training
data.
EP = (TP − OP )2 (39)

where TP and OP represent the desired and the ANFIS output, respectively. Finally, the root mean
square error (RMSE) of the training data was calculated using Equation (40).
rP
EP
E= (40)
P

The RMSE in Equation (40) can be expressed generally as

s
(Epredicted,i − Eactual,i )2
PP
i=1
RMSE = (41)
P

This RMSE is actually minimized during the training of data in ANFIS.

The flowchart in Figure 8 shows the ANFIS implementation. First of all, the training data were
loaded for the generation of FIS. Initial antecedent (input) parameters and MFs were set. Next, the
FIS model optimization method (BPA or HA) was chosen. After this, training and testing (optional)
parameters were defined. The number of training epochs was set to a specific value, e.g., 50 or 100.
Appl. Sci. 2019, 9, 5108 12 of 22

The training data were input to the ANFIS system. The training of the data was done N times, and
when finished, the results were obtained. The testing data were loaded (optional), and testing was
done. Eventually, the FIS structure and output surface could be viewed along with the modified rules
[Link].
Appl. Sci.2019,
2019,9,9,xxFOR
FORPEER
PEERREVIEW
REVIEW 13of
13 of22
22
and MFs adjusted by ANFIS.

Figure 8.
Figure8.
Figure Flowchartof
[Link]
Flowchart of implementation
ofimplementation of
implementationof ANFIS.
ofANFIS.
ANFIS.

The data
The data from
data from FLC
from FLC were
FLC were extracted
were extracted using
extracted using the
using the MATLAB
the MATLAB environment
MATLAB environment
environment andand are
and are presented
are presented
presented inin
in
The
Appendix
Appendix A. A. Two
A. Two hundred
Two hundred data
hundred data points
data points were
points were extracted
were extracted from
extracted from the
from the FLC,
the FLC, which
FLC, which included
which included all
included all possible
all possible
possible
Appendix
combinations
combinations ofof
ofNeg.
[Link],
Big,Neg.
[Link],
Small,Zero,
Zero,Pos.
[Link],
Small,and
andPos.
[Link]
Bigvalues
valuesofof
error and
error andderivative of
derivative
combinations Neg. Big, Neg. Small, Zero, Pos. Small, and Pos. Big values of error and derivative
the error.
ofthe
the The
error. data
The from
data fromAppendix
AppendixAA Awere
werethen
thenloaded
loadedinto
intothe
theneuro-fuzzy designer,
neuro-fuzzydesigner, as
designer,as shown
asshown
showninin
in
of error. The data from Appendix were then loaded into the neuro-fuzzy
Figure
Figure 9.
9.
Figure 9.

Figure 9.
Figure9.
Figure Loadingof
[Link]
Loading oftraining
of trainingdata
training datainto
data intothe
into theneuro-fuzzy
the neuro-fuzzydesigner.
neuro-fuzzy designer.
designer.

The FIS
The FIS from
from the
the loaded
loaded data
data was
was generated
generated using
using the
the grid
grid partition
partition clustering
clustering type.
type. In
In grid
grid
partition, each input variable has an evenly distributed and fixed number of input
partition, each input variable has an evenly distributed and fixed number of input membership membership
functions and
functions and one
one rule
rule for
for each
each MF
MF combination.
combination. The
The consequent
consequent (output)
(output) of
of each
each rule
rule corresponds
corresponds
toone
to oneoutput
outputMF MFagainst
againsteach
eachfuzzy
fuzzyrule.
rule.
The hybrid optimization method integrating backpropagation
The hybrid optimization method integrating backpropagation and and least-squares
least-squares estimation
estimation for
for
Appl. Sci. 2019, 9, 5108 13 of 22

The FIS from the loaded data was generated using the grid partition clustering type. In grid
partition, each input variable has an evenly distributed and fixed number of input membership
functions and one rule for each MF combination. The consequent (output) of each rule corresponds to
one output MF against each fuzzy rule.
The hybrid optimization method integrating backpropagation and least-squares estimation for
the parameters associated with the inputs and the output membership functions, respectively, was
Appl.
[Link].
Sci.2019,
employed for
2019, 9, xxFOR
FORPEER
9,training PEERof REVIEW
the data with 50 epochs and targeting zero error tolerance. After the training
REVIEW 14
14of
of2222
of data, a new FIS was generated by the ANFIS, and this FIS could then be compared with the training
of
ofdata,
datadata,
fromaanew
the FIS
new oldwas
FIS FIS generated
was [Link]
generated
from by the
This ANFIS,
ANFIS,and
thecomparisonandthis
thisFIS
FIScould
between could then
thenbe
the trainingbecompared
compared
data and the with
with the
FIS training
theoutput
training is
data
data from
fromin
presented the
the old
old FIS
Figure FIS from
from FLC.
9, where FLC.
the This
This comparison
training data pointsbetween
comparison between
are markedthe training
theby circles,data
training data and the
anddata
and the FIS
FIS output
the pointsoutput
of new is
is
presented
presented in
in Figure
Figure 9,
9, where
where the
the training
training data
data points
points are
are marked
marked by
by circles,
circles,
FIS output are marked by a star symbol. On the graph in Figure 10, the data points of circles and stars and
and the
the data
data points
points of
of
new
are FIS
FIS output
newalmost output
in the are
are marked
marked
same by aa star
bybut
place, star
very symbol.
small On
symbol. On the
thegraph in
in Figure
graphwere
adjustments Figure
made 10,
10,
bythe
the
thedata
data points
points
ANFIS. of
of circles
This cannotand
circles andbe
stars
seen are
stars in almost
arethe
almost
graph in the
the same
invery same place,
butbut
place,
clearly, but very
very
it was small
small adjustments
evident in the MFs ofwere
adjustments were made
made
the new by the
the ANFIS.
by where
FIS, ANFIS. This
This cannot
the membershipcannot
be
beseen
seenin
functions the
thegraph
inand graph very
veryclearly,
rules were clearly,but
slightly butititwas
adjusted evident
wasduring
evident in
inthe
theMFs
training MFs of
ofthe
by [Link]
newFIS,
FIS,where
wherethe themembership
membership
functions and rules were slightly adjusted during
functions and rules were slightly adjusted during training by ANFIS. training by ANFIS.

Figure 10. Comparison of training data and output FIS after training.
Figure
Figure 10.
10. Comparison
Comparison of
of training
training data
data and
and output
output FIS
FIS after
after training.
training.
Figure 11 shows the neural network diagram of ANFIS, where there are two inputs and one
Figure
Figure 11
11 shows
shows the the neural
neural network
network diagram
diagram of of ANFIS,
ANFIS, where
where there
there are
are two
two inputs
inputs and
and one
one
output with five MFs for each input. Twenty-five MFs of output corresponded to an output value, and
output
output with
with five
five MFs
MFs forfor each
each input.
input. Twenty-five
Twenty-five MFs MFs of of output
output corresponded
corresponded to to an
an output
output value,
value,
the AND operation was performed in fuzzy rules. In the FLC there were originally five MFs for the
and
and the
the AND
AND operation
operation was was performed
performed in in fuzzy
fuzzy rules.
rules. In
In the
the FLC
FLC there
there were
were originally
originally five
five MFs
MFs for
for
output variable, and the ANFIS resulted in 25 MFs for the output variable. The output values were
the output variable, and the ANFIS resulted in 25 MFs for the output variable.
the output variable, and the ANFIS resulted in 25 MFs for the output variable. The output values The output values
then added together to obtain a single output. In the new FIS generated by ANFIS, the rules were
were
were then
then added
added together
together to to obtain
obtain aa single
single output.
output. InIn the
the new
new FIS
FIS generated
generated by by ANFIS,
ANFIS, thethe rules
rules
adjusted a little bit, which can be seen in the 3D surface view of the rules in Figure 12.
were
were adjusted
adjusted aa little
little bit,
bit, which
which can
can be
be seen
seen inin the
the 3D
3D surface
surface view
view of
of the
the rules
rules in
in Figure
Figure 12.
12.

Figure 11. Neural network diagram.

Figure
Figure 11.
11. Neural
Neural network
network diagram.
diagram.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 15 of 22

[Link].
Appl. Sci.2019,
2019,9,9,5108
x FOR PEER REVIEW 15of
14 of22
22

Figure 12. New rules surface, after training in ANFIS.

4.3. PID Controller

The transfer function (duty-cycle-to-output) of the flyback converter is well known. The
Figure 12. New rules surface, after training in ANFIS.
Figure 12. New
heuristic Ziegler–Nichols closed-loop tuningrules surface,
methodafter wastraining
employed in ANFIS.
to tune the parameters of the
4.3.
PID PID Controller
controller for the flyback converter. MATLAB’s sisotool was used to further tune the parameters.
4.3. PID Controller
As a The
result, the PID
transfer compensator
function can be expressed
(duty-cycle-to-output) by [49]
of the flyback converter is well known. The heuristic
The transfer function (duty-cycle-to-output) of the to flyback converter isof well known. The
Ziegler–Nichols closed-loop tuning method  was1 employed tune the parameters the PID controller
heuristic
for Ziegler–Nichols
the flyback GPID ( s) = sisotool
closed-loop
converter. MATLAB’s  1 + was
K ptuning + Td s  to
method was employed to parameters.
tune the parameters of the
the
PID controller for the flyback converter. Ti s used
MATLAB’s
further tune the As a result,
sisotool was used to further tune the parameters.
PID compensator can be expressed by [49]
As a result, the PID compensator can be expressed K by [49] K p (42)
= K p + i + Kd s ∴ Ki = ;K  d = K pTd ;
1
GPID(ss) = Kp 1 + +TT
 Ti s i d s
1
GPID ( s)==K  1 + + T
Ki99,999.88 s  Ki = Kp ; Kd = Kp Td ;
K p  +s T s d  +
= 18.45
p + + K sd ∴ (42)
 i
s  0.00071
99,999.88
Ti s
= 18.45 + s +
K p0.00071s
Ki (42)
= Kp + + Kd s ∴ Ki = ; Kd = K pTd ;
5. Results and Discussion s Ti
5. Results and Discussion
99,999.88
To analyze the
To analyze theperformance
performanceofof =the
18.45
theANFIS + controller
ANFIS + 0.0
controller 0071
and toscompare
and to compare its performance
its performance with
with other
s
other controllers
controllers such such
as FLCas FLC
andandPID, PID, simulations
simulations werewereperformed
performedusing using thethe MATLAB/Simulink
MATLAB/Simulink
environment. The results are presented and discussed
environment. The results are presented and discussed in this section. in this section.
5. Results and Discussion
5.1. Nominal Performance
5.1. Nominal
To analyzePerformance
the performance of the ANFIS controller and to compare its performance with other
FC withsuch
controllers unityas feedback,
FLC and FLC,
PID, ANFIS, and PIDwere
simulations controllers
performedwere simulated
using thefor a fixed load. The
MATLAB/Simulink
FC with unity feedback, FLC, ANFIS, and PID controllers were simulated for a fixed load. The
output voltageThe
environment. offered byare
results these controllers
presented and is shown ininFigure
discussed this 13. The PID parameters used were
section.
output voltage offered by these controllers is shown in Figure 13. The PID parameters used were
summarized in Table 3.
summarized in Table 3.
5.1. Nominal Performance
FC with unity feedback, FLC, ANFIS, and PID controllers were simulated for a fixed load. The
output voltage offered by these controllers is shown in Figure 13. The PID parameters used were
summarized in Table 3.

Figure 13. The output voltage of FC with unity feedback, FLC, ANFIS, and proportional–integral–
derivative (PID) controllers.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 16 of 22
Appl. Sci. 2019, 9, 5108 15 of 22
Figure 13. The output voltage of FC with unity feedback, FLC, ANFIS, and proportional–integral–
derivative (PID) controllers.
Table 3. PID parameters.

PIDTable 3. PID parameters.

Parameters Values
Proportional gain KP
PID Parameters 18.45
Values
Integral gain KI KP
Proportional gain 99,999.88
18.45
Derivative gain KD 0.00071
Integral gain KI 99,999.88
Derivative gain KD 0.00071
The response
The response parameters
parametersextracted
extractedfrom
fromFigure
Figure1313are
aretabulated
tabulated inin
Table 4. 4.
Table From
From Table 4, it4,can
Table be
it can
deduced
be deducedthat thethe
that ANFIS-based
ANFIS-based feedback
feedback controller
controllerhad
hadthetheleast
leastrise
risetime,
time,settling
settling time,
time, percentage
percentage
overshoot, peak voltage, peak time, and steady-state error at 0.8827 ms, 6.3 ms, 0.5603%,
overshoot, peak voltage, peak time, and steady-state error at 0.8827 ms, 6.3 ms, 0.5603%, 24.1345 24.1345V, V,
1.8
1.8 ms,
ms, and
and 1.04%,
1.04%, respectively,
respectively, as
as compared
compared to to FLC
FLC and
and PID.
PID. Only
Only the
the percentage
percentage undershoot
undershoot was was
less, 0.0149%, in the case of FLC. The ANFIS controller completely outshone the other
less, 0.0149%, in the case of FLC. The ANFIS controller completely outshone the other controllers. controllers.

Table 4.
Table Comparison of
4. Comparison of response
response parameters
parameters of
of unity
unity feedback,
feedback, FLC,
FLC, ANFIS,
ANFIS, and
and PID.
PID.
Parameters
Parameters Unity Feedback
Unity Feedback FLCFLC ANFIS
ANFIS PID PID
RiseRise
time time
(ms) (ms) 1.0 1.0 0.8925
0.8925 0.8827
0.8827 0.980.98
Settling time (ms) (ms)
Settling time NaN NaN 6.9 6.9 6.3
6.3 22.122.1
Overshoot (%) 28.1803 11.0875 0.5603 19.5644
Overshoot (%) 28.1803 11.0875 0.5603 19.5644
Undershoot (%) 0.0022 0.0149 0.0205 0.3464
Undershoot
Peak (V) (%) 0.0022
30.7633 0.0149
26.6610 0.0205
24.1345 0.3464
28.6955
Peak timePeak
(ms)(V) 2.430.7633 26.6610
1.9 24.1345
1.8 28.69552.7
Steady-state
Peakerror
time(%)(ms) 7.54 2.4 1.83 1.9 1.04
1.8 2.71.79
Steady-state error (%) 7.54 1.83 1.04 1.79
5.2. Load Regulation
5.2. Load Regulation
Load regulation was performed by changing the value of the load resistance and checking the
capability regulation
Load was performed
of the controller to maintainbyachanging the value
constant output of the For
voltage. loadtheresistance and checking
load resistance changethe
of
10 Ω to 14 of
capability Ω and from 10 Ωtotomaintain
the controller 6 Ω, loadacurrent
constant output voltage.
variations are shownForinthe load 14.
Figure resistance
Reduced change
peak of
to
10 Ω to 14 Ω and from 10 Ω to 6 Ω, load current variations are shown in Figure 14. Reduced
peak voltage spike with reduced recovery time was observed at the time of load transience. The ANFIS peak to
peak voltage spike with reduced recovery
controller offered excellent load regulation. time was observed at the time of load transience. The
ANFIS controller offered excellent load regulation.

(a) (b)

Figure 14. Output voltage and load current when the load resistance was (a) increased and (b) decreased.
Figure 14. Output voltage and load current when the load resistance was (a) increased and (b)
decreased.
Appl. Sci. 2019, 9, 5108 16 of 22

Appl. Sci. 2019, 9, x FOR PEER REVIEW 17 of 22

Appl. Sci. 2019, 9, x FOR PEER REVIEW 17 of 22
5.3. Line Regulation
5.3. Line Regulation
5.3. Line
LineRegulation
regulation was performed by changing the value of the supply voltage and checking the
Line regulation
capability was performed
of the controller byachanging
to maintain constant the value
output of the The
voltage. supply voltage
supply and was
voltage checking the
changed
Line regulation was performed by changing the value of the supply voltage and checking the
capability
from of
12 V to the controller to maintain a constant output voltage. The supply voltage was changed
capability of15 V controller
the and 12 V toto9 V. The output
maintain voltageoutput
a constant across the load resistance
voltage. The supplyis presented in Figure
voltage was 15.
changed
fromcontroller
The 12 V to 15was
V and 12to
able V compensate
to 9 V. The output
for voltage
the changesacross
in the
the load resistance
supply voltage, is presented
thus offering in Figure
superior
from 12 V to 15 V and 12 V to 9 V. The output voltage across the load resistance is presented in Figure
15. The
line controller was able to compensate for the changes in the supply voltage, thus offering superior
regulation.
15. The controller was able to compensate for the changes in the supply voltage, thus offering superior
line regulation.
line regulation.

Figure 15.
Figure Output voltage
15. Output voltage and
and load current when the input voltage was increased and decreased.
Figure 15. Output voltage and load current when the input voltage was increased and decreased.
5.4. Change in Reference Voltage
5.4. Change in Reference Voltage
5.4. Change in Reference Voltage
The reference voltage was changed from 24 to 28 V. The output voltage waveform is presented in
The reference voltage was changed from 24 to 28 V. The output voltage waveform is presented
FigureThe reference
16a. voltage
Similarly, was changed
the reference from
voltage 24changed
was to 28 V. The
fromoutput voltage
24 to 20 waveform
V at time 0.01 s. isThe
presented
output
in Figure 16a. Similarly, the reference voltage was changed from 24 to 20 V at time 0.01 s. The output
in Figurewaveform
voltage 16a. Similarly, the reference
is presented voltage
in Figure [Link]
Thechanged from 24 to 20followed
output successfully V at timethe
0.01changes
s. The output
in the
voltage waveform is presented in Figure 16b. The output successfully followed the changes in the
voltage waveform
reference signal. is presented in Figure 16b. The output successfully followed the changes in the
reference signal.
reference signal.

(a) (b)
(a) (b)
Figure 16. Output voltage when the reference voltage was (a) increased and (b) decreased.
Figure 16. Output voltage when the reference voltage was (a) increased and (b) decreased.
Figure 16. Output voltage when the reference voltage was (a) increased and (b) decreased.
In a nutshell, Table 5 presents the complete summary of results related to line and load regulation
In a nutshell, Table 5 presents the complete summary of results related to line and load
In a nutshell,
and set-point Table
tracking. The5 percentage
presents the complete
load and line summary
regulationsofwere
results
belowrelated to linefor
1%, except and
theload
line
regulation and set-point tracking. The percentage load and line regulations were below 1%, except
and set-point
regulation (when tracking.
the input voltageThewaspercentage
increased load
fromand
12 Vline regulations
to 15 were below
V), i.e., 1.67%. 1%, except
The settling time
for the line regulation (when the input voltage was increased from 12 V to 15 V), i.e., 1.67%. The
the line
for load regulation
regulation was (when thethe
less than input
line voltage wasThe
regulation. increased from error,
steady-state 12 V when
to 15 V),
the i.e., 1.67%.value
reference The
settling time for load regulation was less than the line regulation. The steady-state error, when the
settling time
changed, wasfor
alsoload regulation
around 1% withwas less than
settling timesthe
of line regulation. 4.0–4.5
approximately The steady-state
ms. error, when the
reference value changed, was also around 1% with settling times of approximately 4.0–4.5 ms.
reference value changed, was also around 1% with settling times of approximately 4.0–4.5 ms.
Table 5. Summarized results of (a) load regulation, (b) line regulation, and (c) change in the reference
Table 5. Summarized results of (a) load regulation, (b) line regulation, and (c) change in the reference
voltage.
voltage.
Appl. Sci. 2019, 9, 5108 17 of 22

Table 5. Summarized results of (a) load regulation, (b) line regulation, and (c) change in the
reference voltage.

(a)
Step Load
Parameters
10 Ω to 14 Ω 10 Ω to 6 Ω
Load regulation (%) 0.31 0.65
Settling time (ms) 2.5 3.0
(b)
Step Line
Parameters
12 V to 15 V 12 V to 9 V
Line regulation (%) 1.67 1.0
Settling time (ms) 5.5–6.0 5.5–6.0
(c)
Reference Value Change
Parameters
24 V to 28 V 24 V to 20 V
Steady-state error (%) 0.9 1.05
Settling time (ms) 4.0–4.5 4.0–4.5

6. Conclusions
To keep the output voltage of an FC system regulated as necessary for modern electronic devices
irrespective of the changes in input voltage or load current, an ANFIS-based controller was proposed
in this research. A FIS structure containing 25 fuzzy rules (in our case) relating the inputs and the
output originally generated by the fuzzy logic system was tuned by ANFIS. It was observed that
the FIS tuned by ANFIS exhibited better output voltage response under static and transient loads.
Better set-point tracking was also achieved. ANFIS may utilize past results or experiences to make
immediate decisions about the response. Training of fuzzy rules by ANFIS made the ANFIS-based
controller more adaptive. Simulation results depicted that both the static and dynamic performance
of the ANFIS-based controller was better in all aspects compared to the FLC and PID controllers.
Hardware implementation of the proposed compensated system can be carried out in future.

Author Contributions: conceptualization, M.A.S. and G.A.; data curation, M.R.H. and A.B.A.; formal analysis,
G.A. and V.E.B.; investigation, U.F.; methodology, M.A.S., G.A., and M.U.A.; resources, M.U.A., M.U. and M.R.H.;
supervision, G.A.; validation, J.G. and V.E.B.; visualization, U.F. and T.Y.; writing—original draft, M.A.S. and G.A.;
writing—review & editing, A.B.A., M.U., M.U.A., T.Y., J.G. and U.F.
Funding: The authors extend their appreciation to the Deanship of Scientific Research at King Khalid University for
funding this work through Research Group Project under grant number R.G.P. 1/166/40. This research work is also
supported by the Deanship of Scientific Research, Majmaah University under the contract number RGP-2019-21.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A

Table A1. Training data for ANFIS.

Inputs Output Inputs Output Inputs Output

e ∆e d e ∆e d e ∆e d
15.1067 6.9273 0.929 −17.346 8.6269 0.4953 −17.0141 −15.9759 0
19.478 −5.8267 0.783 −16.8339 −17.4454 0 −17.4687 −18.9016 0
−17.9046 14.9559 0.6263 −11.6396 10.6189 0.4825 17.726 −6.1243 0.7414
Appl. Sci. 2019, 9, 5108 18 of 22

Table A1. Cont.

Inputs Output Inputs Output Inputs Output

e ∆e d e ∆e d e ∆e d
19.842 1.5756 0.9281 16.3544 −18.8754 0.4495 3.8258 −14.4903 0.2735
6.3532 −7.1651 0.4824 −11.7945 7.3804 0.4117 2.3933 −0.495 0.5412
−19.3181 21.0721 0.6084 15.0857 −0.2797 0.8154 −17.0422 −7.7043 0.0462
−10.6321 18.0453 0.65 −12.3108 13.3945 0.5486 16.9455 21.6783 1
2.2503 2.4075 0.6011 20.6047 10.3218 0.9929 5.8586 20.1759 0.9624
21.9603 5.8788 0.98 −7.2008 19.3786 0.7512 −7.1543 −21.4715 0.0192
22.3146 4.1781 0.9785 −14.5634 18.7643 0.6353 0.636 11.4172 0.7616
−16.4346 −14.0284 0 −11.948 −7.9602 0.0746 −4.7132 −11.0823 0.1572
22.5885 −9.5402 0.7647 5.5701 9.5398 0.8279 −20.3536 −3.7039 0.0496
21.944 −1.3957 0.9285 −1.2821 −14.5051 0.169 −12.484 2.2978 0.2883
−0.702 −12.9366 0.2081 −7.1203 −22.534 0.0112 −18.0807 21.2514 0.6214
14.4135 16.5268 1 15.8798 11.7156 1 −15.1724 −3.9483 0.1162
−17.1895 −14.6513 0 4.0927 0.0011 0.5888 −12.4823 23.1865 0.7169
−3.7555 −13.1558 0.1464 2.3867 −0.9637 0.5309 −3.9712 −9.5302 0.207
19.9553 −15.806 0.5831 20.0253 19.4267 1 −21.6166 9.6527 0.5962
14.026 −13.0721 0.5191 −10.2797 5.2736 0.3914 19.3304 7.9843 0.9713
22.0556 −3.0865 0.8941 12.3456 5.648 0.881 21.3498 1.8781 0.9556
7.4756 −9.0671 0.4655 12.179 17.2532 1 −0.4385 9.5091 0.6968
−22.2858 20.3222 0.5878 −5.7386 14.6635 0.6884 −0.5159 7.9933 0.6623
16.7582 −3.35 0.7822 3.2554 3.6826 0.6506 −7.7895 −15.4496 0.0555
20.8317 −15.1288 0.6142 −20.359 −15.2197 0 19.2026 −17.8553 0.527
8.5793 19.4343 0.9767 −21.4104 −12.4833 0 −6.2762 23.9559 0.8629
12.3715 23.0279 1 1.4783 18.5526 0.9049 −18.6623 −15.7862 0
11.6704 −2.9342 0.6893 13.4 −22.6236 0.3152 13.4521 −22.4352 0.3201
−5.1731 −18.6663 0.0589 20.8325 −0.4847 0.926 −5.2925 2.9376 0.4489
7.4629 −11.6129 0.4101 −17.7645 −15.9395 0 −12.3988 18.3296 0.6348
−15.783 −4.3814 0.102 3.3035 22.9767 0.9854 −4.6122 8.1204 0.5761
9.8902 4.555 0.8135 −1.4692 10.2093 0.6897 −19.3702 −14.8592 0
−22.472 −11.4138 0.0003 −23.4287 0.0226 0.0129 −17.6653 −6.292 0.0576
−10.7077 4.9365 0.3748 −7.8181 −1.3878 0.3002 21.2184 −1.8852 0.9034
−21.7838 10.1384 0.6263 −16.2152 −21.1383 0 21.8945 23.1186 1
−19.3377 −13.3562 0 14.1257 8.7347 0.9522 3.61 −16.4926 0.2287
15.526 −18.364 0.4431 −9.0617 −21.9633 0.0087 −21.1306 17.0651 0.6325
9.3518 −9.7596 0.4912 1.3696 −20.5706 0.0984 −12.7306 6.9487 0.4058
−8.7792 −8.6986 0.1207 −16.0489 1.0392 0.1982 −7.0484 −5.9389 0.2182
21.6107 −3.64 0.8731 4.8951 −19.357 0.1992 15.4173 −14.8357 0.5117
−22.3466 0.3772 0.0555 −11.3774 15.2711 0.5782 −23.2606 −3.4439 0.0104
−2.9403 −19.8952 0.0612 7.3958 15.2423 0.9372 −21.9349 −0.8629 0.0383
−5.6852 −11.4009 0.1292 9.0823 10.6771 0.9288 −15.8885 −18.2106 0
Appl. Sci. 2019, 9, 5108 19 of 22

Table A1. Cont.

Inputs Output Inputs Output Inputs Output

e ∆e d e ∆e d e ∆e d
12.7448 14.4487 1 11.9113 −16.8065 0.4019 7.1575 4.2964 0.7486
14.1696 −22.5974 0.3312 −2.374 7.6611 0.6147 11.1227 −13.143 0.4589
−15.0301 20.585 0.6498 −19.9766 0.8926 0.1262 7.0918 −5.5383 0.5337
−0.4913 11.0559 0.7293 −13.0091 22.7028 0.7004 −2.3557 3.9833 0.5353
−2.6119 −0.5468 0.4315 19.8402 7.1516 0.9684 2.2564 −11.9133 0.2911
7.023 3.7692 0.7342 −16.6859 14.4159 0.6135 −9.7766 −10.0588 0.0695
10.0495 −12.6104 0.4462 15.6392 −2.2177 0.7844 11.7453 5.6204 0.8743
12.225 −1.9753 0.7213 1.8404 −3.2452 0.4695 −14.9302 −11.2665 0.004
−10.7508 22.2282 0.7312 23.8145 15.6151 1 8.9652 15.5701 0.9625
8.6257 2.2467 0.7359 −20.2476 −19.9934 0 −15.1915 23.1678 0.6696
7.4447 1.0145 0.6836 −2.7514 −17.6078 0.0975 −6.3127 11.0519 0.6028
−16.1946 −12.8835 0 −18.8807 −15.6773 0 6.0297 −7.4939 0.4682
−18.2881 −0.5329 0.1091 22.1711 −5.235 0.8498 13.4509 4.0353 0.8627
−0.0785 5.9549 0.6275 −23.7776 15.9062 0.6608 −20.106 −18.8271 0
22.0677 8.5985 0.9902 13.1957 14.5615 1 20.6105 19.5028 1
−7.6615 −5.0153 0.2249 15.2306 −21.0974 0.3825 13.2342 18.2234 1
4.0929 −6.363 0.4507 17.6973 −4.8356 0.7688 −0.634 15.2525 0.811
−13.257 23.4231 0.7069 −19.9471 1.29 0.147 −3.0788 −11.4851 0.1839
12.0608 −22.1885 0.2971 −4.8104 −3.9936 0.3089 −2.5544 4.5291 0.5429
−11.7554 18.4881 0.639 −11.5262 7.5293 0.4134 −9.2952 −22.9194 0.0042
0.2859 19.8378 0.9187 14.4033 6.1427 0.9103 0.4084 −3.5876 0.431
9.5557 14.2168 0.9666 −3.2921 −9.9848 0.2119 0.517 −8.9895 0.3161
18.7634 −19.2618 0.49 19.7111 −3.2807 0.8429 15.2461 −16.2487 0.4799
22.046 −11.4302 0.7128 −15.2713 −23.2566 0 14.1519 −15.4192 0.4746
2.2663 −7.9029 0.3777 −11.3375 23.2351 0.7386 −21.3612 21.2644 0.5845

References
1. Sarif, M.S.M.; Pei, T.X.; Annuar, A.Z. Modeling, design and control of bidirectional DC-DC converter using
state-space average model. In Proceedings of the IEEE Symposium on Computer Applications & Industrial
Electronics (ISCAIE), Penang, Malaysia, 28–29 April 2018; pp. 416–421.
2. Jiang, W.; Chincholkar, S.H.; Chan, C. Investigation of a Voltage-Mode Controller for a dc-dc Multilevel
Boost Converter. IEEE Trans. Circuits Syst. II Express Briefs 2018, 65, 908–912. [CrossRef]
3. MShahid, A.; Yasin, A.R.; Ahmad, S. Practical sliding mode controller in Dc-Dc converters for use in hybrid
vehicles. In Proceedings of the 19th International Multi-Topic Conference (INMIC), Islamabad, Pakistan,
5–6 December 2016; pp. 1–7.
4. Wang, C. A Novel ZCS-PWM Flyback Converter with a Simple ZCS-PWM Commutation Cell. IEEE Trans.
Ind. Electron. 2008, 55, 749–757. [CrossRef]
5. Iqbal, H.K.; Abbas, G. Design and analysis of SMC for second order DC-DC flyback converter. In Proceedings
of the 17th IEEE International Multi Topic Conference 2014, Karachi, Pakistan, 8–10 December 2014;
pp. 533–538.
6. Tamyurek, B.; Torrey, D.A. A Three-Phase Unity Power Factor Single-Stage AC–DC Converter Based on an
Interleaved Flyback Topology. IEEE Trans. Power Electron. 2011, 26, 308–318. [CrossRef]
Appl. Sci. 2019, 9, 5108 20 of 22

7. Ohsato, T.; Satoh, N.; Sekiya, H. A flyback converter using power-MOSFETs to achieve high-frequency
operation beyond 10MHz. In Proceedings of the IEEE 3rd International Future Energy Electronics Conference
and ECCE Asia, Kaohsiung, Taiwan, 3–7 June 2017; pp. 1101–1105.
8. Xie, X.; Li, J.; Peng, K.; Zhao, C.; Lu, Q. Study on the Single-Stage Forward-Flyback PFC Converter With QR
Control. IEEE Trans. Power Electron. 2016, 31, 430–442. [CrossRef]
9. Hwu, K.; Jiang, W. Isolated step-up converter based on flyback converter and charge pumps. IET Power
Electron. 2014, 7, 2250–2257. [CrossRef]
10. Nymand, M.; Andersen, M.A.E. High-Efficiency Isolated Boost DC–DC Converter for High-Power
Low-Voltage Fuel-Cell Applications. IEEE Trans. Ind. Electron. 2010, 57, 505–514. [CrossRef]
11. Yousefzadeh, V.; Shirazi, M.; Maksimovic, D. Minimum Phase Response in Digitally Controlled Boost and
Flyback Converters. In Proceedings of the APEC—Twenty-Second Annual IEEE Applied Power Electronics
Conference and Exposition, Anaheim, CA, USA, 25 February–1 March 2007; pp. 865–870.
12. Verma, S.; Singh, S.K.; Rao, A.G. Overview of control Techniques for FC-FC converter. Res. J. Eng. Sci. 2013,
2, 18–21.
13. Tan, S.; Lai, Y.M.; Tse, C.K.; Martinez-Salamero, L.; Wu, C. A Fast-Response Sliding-Mode Controller for
Boost-Type Converters with a Wide Range of Operating Conditions. IEEE Trans. Ind. Electron. 2007,
54, 3276–3286. [CrossRef]
14. Salimi, M.; Soltani, J.; Zakipour, A.; Hajbani, V. Sliding mode control of the DC-DC flyback converter with
zero steady-state error. In Proceedings of the 4th Annual International Power Electronics, Drive Systems and
Technologies Conference, Tehran, Iran, 13–14 February 2013; pp. 158–163.
15. He, J.; Xu, J.; Yan, T. Peak current-controlled CRM flyback PFC converter with square of line
voltage-compensated primary current envelope. Electron. Lett. 2015, 51, 684–686. [CrossRef]
16. Chandranadhan, V.R.; Renjini, G. Average current mode control of improved bridgeless flyback rectifier with
bidirectional switch. In Proceedings of the International Conference on Circuits, Power and Computing
Technologies [ICCPCT-2015], Nagercoil, India, 19–20 March 2015; pp. 1–6.
17. Park, J.; Moon, Y.J.; Jeong, M.G.; Kang, J.G.; Kim, S.H.; Gong, J.C.; Yoo, C. Quasi-Resonant (QR) Controller
with Adaptive Switching Frequency Reduction Scheme for Flyback Converter. IEEE Trans. Ind. Electron.
2016, 63, 3571–3581. [CrossRef]
18. Chen, K.; Liang, T. Design of Quasi-resonant flyback converter control IC with DCM and CCM operation.
In Proceedings of the International Power Electronics Conference, Hiroshima, Japan, 18–21 May 2014;
pp. 2750–2753.
19. Dogra, A.; Pal, K. Designing and Tuning of PI Controller for Flyback Converter. Int. J. Eng. Trends
Technol. (IJETT) 2014, 13, 117–122. [CrossRef]
20. Modak, S.; Panda, G.K.; Saha, P.K.; Das, S. Design of Novel Fly-Back Converter Using PID Controller. Int. J.
Adv. Res. Electr. Electron. Instrum. Eng. 2015, 4, 289–297.
21. Priyadarshini, D.; Rai, S. Design, Modelling and Simulation of a PID Controller for Buck Boost and Cuk
Converter. Int. J. Sci. Res. (IJSR) 2014, 3, 1226–1229.
22. Yilmaz, U.; Kircay, A.; Borekci, S. PV system flyback converter-controlled PI control to charge battery under
variable temperature and irradiance. In Proceedings of the Electronics, Palanga, Lithuania, 19–21 June 2017;
pp. 1–6.
23. Chen, T.H.; Lin, W.L.; Liaw, C.M. Dynamic modelling and controller design of flyback converter. IEEE Trans.
Aerosp. Electron. Syst. 1999, 35, 1230–1239. [CrossRef]
24. Garcia-Rodriguez, L.A.; Williams, E.; Balda, J.C.; Gonzalez-Llorente, J.; Chiacchiarini, H. Control of
a flyback converter operating in BCM using the natural switching surface. In Proceedings of the
IEEE 6th International Symposium on Power Electronics for Distributed Generation Systems (PEDG),
Aachen, Germany, 22–25 June 2015; pp. 1–8.
25. Chen, X.; Jiang, T.; Zhao, S.; Zeng, H.; Zhang, J. Evaluation of primary side control schemes for flyback
converter with constant current output. In Proceedings of the Twenty-Eighth Annual IEEE Applied Power
Electronics Conference and Exposition (APEC), Long Beach, CA, USA, 17–21 March 2013; pp. 1859–1863.
26. Singh, A.; Londhe, P.S. Design of signed distance method based fuzzy logic controller for TITO process.
In Proceedings of the Recent Developments in Control, Automation & Power Engineering (RDCAPE),
Noida, India, 26–27 October 2017; pp. 13–17.
Appl. Sci. 2019, 9, 5108 21 of 22

27. Fahmizal; Kuo, C. Development of a fuzzy logic wall following controller for steering mobile robots.
In Proceedings of the International Conference on Fuzzy Theory and Its Applications (iFUZZY), Taipei, Taiwan,
6–8 December 2013; pp. 7–12.
28. Deepa, K.; Jeyanthi, R.; Mohan, S.; Kumar, M.V. Fuzzy based flyback converter. In Proceedings of the
International Conference on Advances in Electrical Engineering (ICAEE), Vellore, India, 9–11 January 2014;
pp. 1–4.
29. Subbarao, M.; Babu, C.S.; Satyanarayana, S.; Kumar, S.L.V.S. Fuzzy controlled single stage AC/DC converter
with PFC to drive LEDs. In Proceedings of the International Conference on Smart Electric Grid (ISEG),
Guntur, India, 19–20 September 2014; pp. 1–5.
30. Kumar, P.R.; Kumar, S.L.V.S. Comparative Study of PID Based VMC and Fuzzy Logic Controllers for Flyback
Converter. Int. J. Instrum. Control Autom. (IJICA) 2011, 1, 29–36.
31. Hamdan, H.; Garibaldi, J.M. Adaptive neuro-fuzzy inference system (ANFIS) in modelling breast cancer
survival. In Proceedings of the International Conference on Fuzzy Systems, Barcelona, Spain, 18–23 July 2010;
pp. 1–8.
32. Hosseini, S.A.; Afrakoti, I.E.P. Evaluation of a new neutron energy spectrum unfolding code based on an
Adaptive Neuro-Fuzzy Inference System (ANFIS). J. Radiat. Res. 2018, 59, 436–441. [CrossRef]
33. Zhai, J.; Zhou, J.; Zhang, L.; Zhao, J.; Hong, W. Dynamic Behavioral Modeling of Power Amplifiers Using
ANFIS-Based Hammerstein. IEEE Microw. Wirel. Compon. Lett. 2008, 18, 704–706. [CrossRef]
34. Farayola, A.M.; Hasan, A.N.; Ali, A. Curve fitting polynomial technique compared to ANFIS technique for
maximum power point tracking. In Proceedings of the 8th International Renewable Energy Congress (IREC),
Amman, Jordan, 21–23 March 2017; pp. 1–6.
35. Mohammed, S.S.; Devaraj, D.; Ahamed, T.P.I. Maximum power point tracking system for standalone
solar PV power system using Adaptive Neuro-Fuzzy Inference System. In Proceedings of the Biennial
International Conference on Power and Energy Systems: Towards Sustainable Energy (PESTSE), Bangalore,
India, 21–23 January 2016; pp. 1–4.
36. Chu, Y.; Yuan, L.; Chiang, H. ANFIS-based maximum power point tracking control of PV modules with
DC-DC converters. In Proceedings of the 18th International Conference on Electrical Machines and Systems
(ICEMS), Pattaya, Thailand, 25–28 October 2015; pp. 692–697.
37. Tarek, B.; Said, D.; Benbouzid, M.E.H. Maximum Power Point Tracking Control for Photovoltaic System
Using Adaptive Neuro-Fuzzy ANFIS. In Proceedings of the Eighth International Conference and Exhibition
on Ecological Vehicles and Renewable Energies (EVER), Monte Carlo, Monaco, 27–30 March 2013; pp. 1–7.
38. Noman, A.M.; Addoweesh, K.E.; Alolah, A.I. Simulation and Practical Implementation of ANFIS-Based
MPPT Method for PV Applications Using Isolated Ćuk Converter. Int. J. Photoenergy 2017, 2017, 3106734.
[CrossRef]
39. Al-Nussairi, M.K.; Bayindir, R.; Hossain, E. Fuzzy logic controller for Dc-Dc buck converter with constant
power load. In Proceedings of the IEEE 6th International Conference on Renewable Energy Research and
Applications (ICRERA), San Diego, CA, USA, 5–8 November 2017; pp. 1175–1179.
40. Feng, G. A Survey on Analysis and Design of Model-Based Fuzzy Control Systems. IEEE Trans. Fuzzy Syst.
2006, 14, 676–697. [CrossRef]
41. Swathy, M.K.; Jantre, M.S.; Jadhav, M.Y.; Labde, M.S.M.; Kadam, M.P. Design and Hardware Implementation
of Closed Loop Buck Converter Using Fuzzy Logic Controller. In Proceedings of the Second International
Conference on Electronics, Communication and Aerospace Technology (ICECA), Coimbatore, India,
29–31 March 2018; pp. 175–180.
42. Bendaoud, K.; Krit, S.; Kabrane, M.; Ouadani, H.; Elaskri, M.; Karimi, K.; Elbousy, H.; Elmaimouni, L.
Fuzzy logic controller (FLC): Application to control DC-DC buck converter. In Proceedings of the International
Conference on Engineering & MIS (ICEMIS), Monastir, Tunisia, 8–10 May 2017; pp. 1–5.
43. de Martins, J.K.E.C.; de Araújo, F.M.U. Nonlinear system identification based on modified ANFIS.
In Proceedings of the 12th International Conference on Informatics in Control, Automation and Robotics
(ICINCO), Colmar, France, 21–23 July 2015; pp. 588–595.
44. García, P.; García, C.A.; Fernández, L.M.; Llorens, F.; Jurado, F. ANFIS-Based Control of a Grid-Connected
Hybrid System Integrating Renewable Energies, Hydrogen and Batteries. IEEE Trans. Ind. Inform. 2014,
10, 1107–1117. [CrossRef]
Appl. Sci. 2019, 9, 5108 22 of 22

45. Al-Hmouz, A.; Shen, J.; Al-Hmouz, R.; Yan, J. Modeling and Simulation of an Adaptive Neuro-Fuzzy
Inference System (ANFIS) for Mobile Learning. IEEE Trans. Learn. Technol. 2012, 5, 226–237. [CrossRef]
46. Hiremath, S.; Patra, S.K. Transmission rate prediction for Cognitive Radio using Adaptive Neural Fuzzy
Inference System. In Proceedings of the 5th International Conference on Industrial and Information Systems,
Mangalore, India, 29 July–1 August 2010; pp. 92–97.
47. Sowmmiya, U.; Uma, G. ANFIS-based sensor fault-tolerant control for hybrid grid. IET Gener. Transm.
Distrib. 2018, 12, 31–41. [CrossRef]
48. Ying, L.; Pan, M. Using adaptive network based fuzzy inference system to forecast regional electricity loads.
Energy Convers. Manag. 2008, 49, 205–211. [CrossRef]
49. Abbas, G.; Muazzam, H.; Farooq, U.; Gu, J.; Asad, M.U. Comparative Analysis of Analog Controllers for
DC-DC Buck Converter. J. Autom. Control Eng. 2015, 3. [CrossRef]

© 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 ([Link]