Voltage Tracking of a Multi-Input Interleaved

Buck-Boost DC-DC Converter Using Artificial

Neural Network Control
Yonis. M. Buswig 1, Al-Khalid bin Hj Othman 1, Norhuzaimin bin Julai 1, Sim Sy Yi 2, Wahyu Mulyo Utomo 2
and Alvin John Lim Meng Siang 2
1
Universiti Malaysia Sarawak, 94300, Kota Samarahan, Sarawak.
2
Universiti Tun Hussein Onn Malaysia, 86400, Parit Raja, Johor.
[email protected]

Abstract—This paper proposes an artificial neural network controller and neural network control techniques for DC
(ANN) voltage tracking of multi-input interleaved buck-boost converter have been described for in [4]-[11].
DC-DC converter. A back-propagation algorithm topology is
implemented in this paper. The control unit is implemented to II. MULTI-INPUT INTERLEAVED BUCK-BOOST CONVERTER
ameliorate the performance of the proposed multi-input
converter during transient dynamic response and steady-state
operation mode. The neural network controller unit design, The circuit outline of the proposed multi-input interleaved
which is adaptive against output voltage command tracking and buck-boost converter as described in Figure 1. It had been
reference voltage variations is proposed. The proposed design intentionally supposed that Switch(1), Switch(2), Switch(3) and
has been verified through the MATLAB software. The Switch(4) signified as S(1), S(2), S(3) and S(4) separately while
simulation outcomes emphasized the validity and reliability of Diode (1), Diode (2), Diode (3) and Diode (4) indicated as D(1),
the proposed neural network technique, which would be a D(2), D(3) and D(4) individually. The gate pulses of the switches
promising an efficient control method that ensures multi-input for the proposed converter have been purposely assumed that
converter suitable for electric vehicle and renewable energy the duty cycle of S(1), S(2), S(3) and S(4) described as d(1), d(2),
application systems
d(3), and d(4) respectively. A circuit diagram of the proposed
Index Terms—Multi-Input Converter; Algorithm Back- multi-input interleaved buck-boost DC-DC converter
Propagation; Artificial Neural Network Control; Tracking designed is shown in Figure 1. Switching pulses pattern of
Voltage Variations. switches S1, S2, S3 and S4 are described in Figure 2.

I. INTRODUCTION

These days, multi-input DC control supplies are broadly

utilized as a part of many advanced applications in
comparison to the single port electronic system, for example,
electric vehicle and furthermore the sustainable power source
applications [1].
Thus, DC-DC multi-input converter is broadly utilized by
change over a DC voltage created from various input voltage
supplies to an alternate direct voltage amount to supply the
direct voltage source amount demand of the electrical load.
Figure 1: Circuit diagram of the Multi-Input Interleaved buck-boost
Furthermore, the multi-input converter is likewise an DC-DC Converter
imperative application for the power control of the renewable
energy system, for example, solar panel and wind turbine.
Because of these reasons, multi-input converter applications
will go to a more potential market in sustainable power source
applications [2].
Fundamentally, the multi-input converter comprises of the
power semiconductor system which worked as an electrical
converter with a specific end goal to transform the electrical
energy from various power supplies. The operation processes
of the switching system cause the innately nonlinear normal
for the multi-input converters. Because of this undesirable
nonlinear characteristic, the multi-input converters require a
controller system with an excellent level of dynamic reaction
[3].
With a specific end goal to handle this issue and enhance
the dynamic reaction of multi-input DC-DC converters, a few
Figure 2: Switching pattern and duty cycles of switches S(1), S(2), S(3) and
knowledge controller procedures, for example, fuzzy logic S(4)

e-ISSN: 2289-8131 Vol. 10 No. 1-12 29

Journal of Telecommunication, Electronic and Computer Engineering

T( 0)  ( d1 (t )  d1 (t ) ) T (1) aim  f m
( nim ) (9)
T(1)  d12 (t ) T (2)
In Equation (9), f is defined as activation function model
T( 2)  (d 2 (t )  d12 (t ) ) T (3)
of the current neuron. This structure, the main activation
T(3)  d3 (t ) T  (1  (d1 (t )  d 2 (t )  d12 (t ) )) T (4) function for the output layer and the hidden layer are unity
and a tangent hyperbolic function respectively. The activation
Therefore, can write the following equations for a multi- function of the hidden layer is presented as
input interleaved buck-boost converter based upon Figure 2
and volt-second balance equation of the inductors. 2
f m (nim )  1 (10)
1  e  2 ni
m

T0  T1  T2  T3  T (5)
Updating of the connection weight value and bias value
Suppose that the current had reached a steady state. Thus, parameters are presented by
by analyzed can be obtained:
F (k )
wijm (k  1)  wijm (k )   (11)
V1 (d1  d3,4 )T  (V1  V2 ) d3,4T  Vo (1  d1 )T  V2 (d 2  d3,4 )T  V0 (1  d 2 )T  0 (6) wijm

Equation (6) in principle can be solved to find voltage F (k )

output transfer ratio Vo. After several algebraic manipulations bim (k  1)  bim (k )   (12)
bim
and solving, the expression for the output voltage shown as
the following Equation (7).
In Equations (11) and (12), k is defined as sample time, α
V1d1 V2 d 2 is defended as learning rate, and F defined as performance
Vo   (7) index function of the network.
(1  d1 )  (1  d 2 ) (1  d1 )  (1  d 2 )
The training operation reduces and minimizes the error
III. THE PROPOSED STRUCTURE OF ARTIFICIAL NEURAL
output percentage of the network through an optimization
NETWORK CONTROLLER
process [12]. Mostly, in learning mode technique of the ANN
controller an appropriate training data input-output mapping
The structure of proposed neural network controller of a
data of a plant system is desired. The online learning back-
multi-input converter is as described in Figure 3. The network
propagation algorithm is designed and developed [3].
of the suggested ANN controller has a 1-5-1 neurons network
Therefore, the performance index is the sum of square error
structure [3].
as a function of sampling time and is presented by

1
F (k )   ei2 (k )
2 i
(13)

ei (k )  t i (k )  ai (k ) (14)

In Equation (14), ti defined as target signal and ai defined

as output signal on the last layer.

The gradient descent method of the performance index

against to the connection weight values is presented by:

F F  nim
 (15)
 wijm  nim  wijm

Figure 3: The proposed structure of the neural network controller system

The sensitivity case of parameter for the network is defined
The main weight connections value parameters between jth as
and ith neuron at mth layer is presented by wij, while the bias
value parameters of this layer at ith neuron are presented by
F
bmi. The main transfer function model of the network of the s im  (16)
neuron in mth layer is defined as  nim

S m1 F  aim
nim   wijm a mj1  bim
j 1
(8) sim 
 aim  nim
(17)

The total output function of this neuron at mth layer is given

by In gradient processes the transfer function again to the

30 e-ISSN: 2289-8131 Vol. 10 No. 1-12

 Voltage Tracking of a Multi-Input Interleaved Buck-Boost DC-DC Converter Using Artificial Neural Network Control

connection weight value parameter is given by

Source Source
1 2
 nim
 aim 1 (18) Control Unit
 wij
m
Multi-
Artificia
Pulse Input
By substituting Equations (16) and (18) through Equation l Neural Width Loa
Interleave
Networ Modulatio d
(11). Consequently, the updating connection parameter is d
k n
given by DC-DC
Control Technique
Buck-
(ANN) (PWM)
(Control Signal) Boost
wijm 1 ( k  1)  wim 1 ( k )   s mj ( k ) aim 1 ( k ) (19)
(Error Signal)
Feedback Signal
_
(Output Voltage Signal)
The same processes will be implemented in order to update -
the bias parameter as given in Equation (20)
+
Reference
bim 1 ( k  1)  bim 1 ( k )   s mj ( k ) (20) value
(voltage)

IV. SIMULATION RESULTS Figure 4: Schematic diagram of the proposed ANN controller system

A Simulink-MATLAB software has been used to verify the [GS1] Inductor 1 D3

g
efficiency of the proposed ANN control system. Schematic Discrete,
Ts = 1e-005 s. S
D
diagram of the proposed ANN controller system for the multi- powergui L1
IGBT 1
input interleaved buck-boost DC-DC converter is described S1
Inductor 2
VOLTAGE
D1
in Figure 4. The parameters of the multi-input DC-DC S1 [GS1] SOURCE 1
Fs f _sw
converter are shown in Table 1. S2 [GS2]
L2 D4
Switching
Frequency
[GS2] [GS4] [GS3] C R +v
Table 1 S3 [GS3] g -
Capacitor Resistor
Parameters for Proposed Multi-Input Converter D S
D
S4

D
g

g
[GS4]
IGBT 4 IGBT 3
IGBT 2
Symbol Parameter Values PWM VOLTAGE S2
S4 S3
D2
L1 & L2 Inductance 0.5 mH
S

S
GENERATOR BLOCK SOURCE 2

C Capacitance 100 µF
R Resistor 10 Ω
V1 First input voltage 12 V Voltage Reference
Sum1
V2 Second input voltage 24 V Value

S(1) & S(2) Switching frequency 10 kHz Control signal

Control Signal Error v alus Vref
S(3) & S(4) Switching frequency 20 kHz
d(1) & d(2) Duty cycle for main switches 75% ANN
Output Voltage signal

d(3) & d(4) Duty cycle for interleaved switches 50% controller munit

Vo Output voltage 20 V- 60 V
Figure 5: Simulation design of the proposed multi-input interleaved buck-
The pulse width modulation switching technique with a boost converter
fixed switching frequency for the gate pulse has been selected 80
for verification simulation results study. All the results have ANN
been analyzed with continues current mode (CCM) under 70

steady-state, transient conditions, and power flow

60
management control. The simulation circuit of proposed
multi-input interleaved buck-boost converter with using
50
Output Voltage

MATLAB Simulink is shown in Figure 5.

With a specific end goal to confirm the viability and 40
effectiveness of proposed controller, the comparison between
ordinary PID controller and the proposed ANN controller has 30
been done in this investigation of simulation results.
The simulation outcomes showed that the proposed ANN 20

controller has the capability of tracking the voltage reference

10
variations in case of stepping-up and stepping-down, and
displays a very low rise time and negligible overshoot. 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time

Figure 6: The dynamic response of the proposed converter when the output
voltage is ordered to tracking voltage reference values 40V, 20V, and 60V.

e-ISSN: 2289-8131 Vol. 10 No. 1-12 31

Journal of Telecommunication, Electronic and Computer Engineering

80 V. CONCLUSION
70 ANN ANN control technique for multi-input interleaved buck-
PID boost converter has been discussed in this paper. To promote
60
and develop performance of the voltage tracking of a multi-
50 input interleaved buck-boost converter, a back-propagation
Output Voltage

algorithm has been developed. It is noticeable that ANN

40 controller is efficient in decreasing the overshoot, reducing
the settle time and also has a fast dynamic response to track
30
the desired output voltage of the system. Additionally, based
20
on the simulation outcomes, the implementation of the back-
propagation technique is an appropriate solution for the
10 output power regulation for multi-input converter supplied
with different input voltage sources such as solar panel and
0
1.96 1.98 2 2.02 2.04 2.06 2.08
wind turbine in renewable energy system applications.
Time
Figure 7: The dynamic response of the proposed converter when the output ACKNOWLEDGEMENT
voltage is ordered to step-down from 40V to 20V

80 The authors would like to gratitude research and innovation

ANN management centre (RIMC) UNIMAS for valuable supports
70
PID during conducting this research which funded by a Special
Short Term Grant (SpSTG), University Malaysia Sarawak.
60
Grant number: F02/SpSTG/1582/2017.
50
Output Voltage

REFERENCES
40
[1] Y. M. Buswig, W. M. Utomo, Z. A. Haron, and A. A. Bakar, “Power
30 sharing analysis of a new modified multi-input interleaved boost
converter based on h-bridge cells,” ARPN Journal of Engineering and
Applied Sciences, pp. 3943–3950. 2016.
20
[2] Y. M. Buswig, W. M. Utomo, Z. A. Haron, and S. S. Yi, “Multi-Input
Boost Converter for Hybrid PV and Wind Generator Systems,” In
10 Advanced Materials Research, pp. 619–624. 2014.
[3] W. M. Utomo, S. S. Yi, M. Y. Yonis, and M.D.Zarafi.,“Voltage
0 Tracking of a DC-DC Flyback Converter Using Neural Network
3.9 3.95 4 4.05 4.1 4.15 4.2
Time
Control, International Journal of Power Electronics and Drive System
Figure 8: The dynamic response of the proposed converter when the output (IJPEDS), vol. 2, pp. 35–42, March 2012
voltage is ordered to step-up from 20V to 50V [4] K.-H. Cheng, “Fuzzy – Neural Sliding-Mode Control for DC– DC
Converters Using Asymmetric Gaussian Membership Functions,”
IEEE Trans. Industrial Electronics, vol. 54, pp. 1528–1536, 2007.
As observed that the output voltage dynamic transient [5] C.-F. Hsu, “Intelligent Control for DC -DC Power Converter with
response of the multi-input converter with reference voltage Cecurrent Fuzzy Neural Network Approach,” in Conf. Rec. 2006 IEEE
lower than the two input voltage sources (V1 and V2) as in International Conference, pp. 16–21.
the case of the buck converter (the output voltage 20 V) and [6] P. Mattavelli, “General-purpose fuzzy controller for DC–DC
Converters,” IEE E Trans. Power Electronics, vol. 12, pp.79–86,1997
higher than the two input voltage sources (V1 and V2) as in [7] T. Gupta, “Implementation of a fuzzy controller for DC -DC converters
the case of the boost converter (the output voltage at 60V) and using an inexpensive 8-b microcontroller,” IEEE Trans. Industrial
as described in Figure 7 and Figure 8 respectively. Electronics, vol. 44, pp.661–669.1997
The ANN proposed controller provides a preferable [8] J. Mahdavi, “Application of neural networks and State -space
averaging to DC /DC PWM converters in sliding-mode operation,”
performance than PID-Controller such as eliminating IEEE Trans. Mechatronics, vol. 10, pp. 60 – 67, 2005.
overshoot and achieve required output voltage as described in [9] W.M. Utomo, “Online Learning Neural Network Control of Buck -
Figure 7 and Figure 8. Moreover, the settle time of the Boost Converter,” Eighth International Conference on Information
proposed ANN controller also faster than PID-Controller. Technology: New Generations, pp.485–489,2011.
[10] R. Leyva, “Identification and control of power converters by means of
Figure 7 and Figure 8 show the change of reference voltage neural networks,” IEEE Trans. Industrial Applications, vol. 44, pp.
of output voltage transient response for the multi-input 735–742, 1997.
converter. In Figure 7, the output voltage stepping-down from [11] W.M. Utomo, “Voltage Tracking of a DC-DC Buck-Boost. Converter
40 to 20 V. In Figure 8, the output voltage stepping-up from Using Neural Network Control,” IEEE International Journal of
Computer Technology and Electronics Engineering (IJCTEE), vol. 3,
20 V to 60 V. It is obvious from Figure 6 and Figure 7, and pp. 108–113, 2011.
Figure 8 the proposed ANN controller has a preferable [12] W.M. Utomo, N. M. Zin, Z. A. Haron, “Speed Tracking of Field
transient dynamic response than the PID-Control under Oriented Control Permanent Magnet Synchronous Motor Using Neural
variations of the reference voltage-step. Network,” International Journal of Power Electronics and Drive
Systems, vol. 4, pp. 290–298, 2014.

32 e-ISSN: 2289-8131 Vol. 10 No. 1-12