2023 IEEE World Conference on Applied Intelligence and Computing

Design of Neural Network based Controller for

Switching Regulator in DC-DC Boost Converters
Deva Priya M D* Amutha C M.G.Umamaheswari
Electrical and Electronics Electrical and Electronics Electrical and Electronics
Engineering Engineering Engineering
Rajalakshmi Engineering College Rajalakshmi Engineering College Rajalakshmi Engineering College
Chennai, TamilNadu Chennai, TamilNadu Chennai, TamilNadu
[email protected] [email protected] [email protected]
2023 IEEE World Conference on Applied Intelligence and Computing (AIC) | 979-8-3503-1006-1/23/$31.00 ©2023 IEEE | DOI: 10.1109/AIC57670.2023.10263972

u.in

Abstract— The primary objective of switching regulators is A two-phase interleaved boost converter in fuel cell
to efficiently convert, regulate, and manage electrical power, applications is executed by incorporating an outer loop for
guaranteeing stable and effective operation of electronic systems voltage regulation and an inner loop for current tracking [5].
in diverse applications. In this research work a Conventional The utilization of an Arduino-based Proportional Integral (PI)
DC-DC Boost Converter (CBC) and a two-phase Interleaved controller in a two-phase interleaved boost converter
DC-DC Boost Converter (IBC) were employed in a Switching simplifies the system complexity, enhances stability, and
Regulator. A cascade controller based on Artificial Neural improves reliability [6]. The use of an interleaved boost DC-
Network (ANN) was implemented on both the CBC and IBC, DC converter minimizes the ripple current generated by
where the networks were trained using algorithms such as
Proton Exchange Membrane Fuel Cells (PEMFC) and
Levenberg Marquardt (LM), Bayesian Regularization (BR),
and Scaled Conjugate Gradient (SCG). Consequently, the
decreases the stress on the semiconductor components [7]. A
implementation of ANN resulted in reduced time domain photovoltaic application incorporates a three-level interleaved
parameters such as settling time, rise time, and peak time boost converter with the implementation of a Zero Voltage
compared to the Conventional Proportional Integral (PI) Switching (ZVS) technique [8]. A two-phase interleaved
Controller. boost converter based Microinverter is regulated by an MPPT
while the H-bridge is controlled using a PI controller [9]. A
Keywords— Conventional Boost Converter, PI Controller, Power Factor Correction (PFC) based interleaved boost
Artificial Neural Network, Levenberg-Marquardt Algorithm, converter is utilized in a renewable energy system, where both
Bayesian Regularization Algorithm, Scaled Conjugate Algorithm, PI and PID controllers are implemented and compared [10].
Interleaved Boost Converter.
Artificial Neural Network (ANN) based performance
I. INTRODUCTION comparison of a Boost Converter and a Buck-Boost Converter
in achieving maximum power point tracking for a solar PV
DC-DC converters regulate and convert DC power
system is discussed in the literature [11]. When compared to a
with high efficiency, ensuring that electronic systems and
conventional Proportional Integral (PI) controller, the
devices receive the necessary voltage levels for optimal
utilization of a Fractional Order Proportional Integral (FOPI)
performance and minimize power losses and maximize
controller with the Grey Wolf Optimization (GWO) algorithm
energy efficiency in the process. A DC/DC boost converter
yields superior performance in a DC-DC SEPIC configuration
employs a coupled inductor and an optimized MOSFET to
[12]. FOPI based controllers provide better performance when
achieve high efficiency and gain while stepping up lower DC
tuned with Generalized Hopfield Neural Network (GHNN) in
voltages to higher voltages and incorporates a Proportional
a closed-loop synchronous SEPIC configuration [14]. In the
Integral Derivative (PID) controller to maintain a constant
context of voltage regulation in DC-DC boost converters, an
output voltage, regardless of variations in the input source or
Artificial Neural Network (ANN) controller replaces a Linear
load conditions [1]. The inclusion of a boost converter enables
Quadratic Regulator (LQR) based controller [13]. Back
the utilization of a Maximum Power Point Tracking (MPPT)
Propagation (BP) neural network using a Proportional Integral
algorithm in photovoltaic (PV) systems that make use of the
Derivative (PID) controller is employed in Boost Converter
Peripheral Interface Controller. [2]. A Genetic Algorithm
[15]. A performance evaluation to compare the effectiveness
(GA) based tuning method has been developed and employed
of a PI controller and a Model Reference Control (MRC)
to optimize the parameters of a Fractional Order PID
based on a neural network is employed in a dual active bridge
controller for a DC-DC Boost converter [3]. A cascade control
converter [16]. ANN based voltage control is applied to DC
scheme is created by combining a Model Predictive Control
Microgrid over PI control using a Boost Converter [17]. ANN
(MPC) and a Proportional Integral (PI) controller to form a
based controller is designed and analyzed for temperature
current-mode controller for an interleaved DC-DC boost
control in literature [18]. The Neural Network-based
converter [4].

979-8-3503-1006-1/23/$31.00 ©2023 IEEE 435

DOI: 10.1109/AIC.2023.71
Predictive Controller (NNPC) acts as an optimal controller for ΔIL = Vin/((1-D²)R) = 0.71 (5)
the buck converter [19,20]. L = D(1-D²)R/2f = 2µH (6)
II. CBC DESIGN AND CALCULATION ΔVo = (0.5% of Vo) = 0.25V (7)
C = DVo / (ΔVoxRF) = 2µF (8)
A Boost Converter can achieve exceptional efficiency
due to its ability to convert as much as 99% of input energy The following are values assigned to various
into usable output energy when properly designed. A CBC has parameters in the given terms: Input Voltage (Vin) is 12V,
two modes such as Mode 1 (switch ON; diode OFF) and Mode Switching Frequency (Fsw) is 50KHz, Duty Cycle (D) is 0.75,
2 (switch OFF; diode ON). PWM is utilized for turning on and Inductance (L) is 2 µH, Capacitance (C) is 2 µF, Load
off the switch. Fig. 1 depicts the basic CBC circuit. The two Resistance (RL) is 38.4Ω, and Output Voltage (Vout) is 48V.
modes of CBC are illustrated in Fig. 2 and Fig. 3. III. IBC DESIGN AND CALCULATION
This paper focuses on the study of Continuous Current
Mode (CCM) in the context of interleaving, which involves
synchronizing multiple smaller sources with relative phase
shifts to increase the effective pulse frequency. To perform
calculations, inductances are assumed to be equivalent with
equal duty cycles. Fig. 4 depicts the basic IBC circuit.

Fig. 1. Basic Conventional Boost Converter Circuit

Fig. 4. Basic Interleaved Boost Converter Circuit

A. Mode I-IBC
During this mode of operation, with the closure of
Fig. 2. Conventional Boost Converter -Mode 1 switch S1 and the opening of switch S2, diode D1 experiences
reverse biasing, whereas diode D2 undergoes forward biasing.
Inductor L1 receives energy from the input, leading to an
increase in inductor current iL1. Simultaneously, the load
receives energy from inductor L2, leading to a reduction in the
current flowing through inductor iL2. Fig. 5 illustrates the
Mode I circuit of IBC.

Fig. 3. Conventional Boost Converter -Mode 2

The following equations are used to find the output

current (Io), load resistance value (RL), duty cycle (d),
inductance (L), and capacitance value (C) in the given context:
Io = Po / Vo = 1.2A (1)
Ro= (Vo/Io) = (48/1.25) = 38.4kΩ (2) Fig. 5. Interleaved Boost Converter -Mode I.

Vo= (Vin/(1-D)) (3)

D = (36 / 48) = 0.75 (4)

436
B. Mode II-IBC
Fig 6 displays the IBC-Mode II circuit. n this operating
state, switches S1 and S2 remain unactivated, while diodes
D1 and D2 become forward-biased. Inductors L1 and L2
discharge, transferring energy to load, resulting in a decrease
in the currents flowing through inductors iL1 and iL2.

Fig. 8. Interleaved Boost Converter -Mode IV

E. IBC Parameters
The following criteria are used to design the IBC
parameters: output power = 60 W, input voltage = 12 V,
Output Voltage = 48 V, Switching frequency, f = 50 kHz,
Fig. 6. Interleaved Boost Converter -Mode II Load Resistance = 384 Ω, ΔiL1 = 0.5 A, ΔvC1 = 0.026 V,
duty cycle = 0.7, Output Current = 2 A, ΔvC2 = ΔVo = 0.25
C. Mode III-IBC V, L = 2µH, C = 2µF.
The circuit for IBC-Mode III was illustrated in Fig 7. IV. CASCADE CONTROL SIMULATIONS WITH PI CONTROLLER
In this operational state, with switch S1 open and switch S2
closed, diode D1 becomes forward-biased while diode D2 The suggested PI controller for the CBC and IBC has
becomes reverse-biased. Inductor L1 discharges, resulting in a dual-loop configuration. The initial loop oversees the
a decrease in its current iL1, which supplies energy to the load. regulation of the inductor current, whereas the secondary loop
manages the voltage across the capacitor and establishes the
reference current for the first loop. The two PI controllers are
linked in a cascade fashion to enable the closed-loop operation
of CBC and IBC. The converter has a 48V reference voltage
as input, along with the output signal from the converter's load
A. CBC simulation with PI
The 12V input voltage is utilized in a closed-loop
CBC, where the 48V output voltage (Vo) is compared to a
reference voltage for precise regulation. This is achieved
through an external voltage loop and an internal current loop.
The duty cycle of the PWM signal in the boost converter is
Fig. 7. Interleaved Boost Converter -Mode III continuously modified by the PI controller to regulate the
output voltage. The PI controller compares the measured
D. Mode IV-IBC output voltage with the desired reference voltage, and by
During this mode of operation for IBC, when switches regulating the error between them, it maintains a stable output
S1 and S2 are simultaneously open, both diodes D1 and D2 voltage that closely tracks the reference voltage.
become forward-biased. Consequently, inductors L1 and L2 B. IBC simulation with PI
discharge and provide energy to load, resulting in a reduction
of the currents flowing through inductors iL1 and iL2. The The IBC employs a closed-loop control system
circuit diagram of Mode IV IBC is illustrated in Fig. 8. comprising an outer voltage loop and an inner current loop.
Both of these loops utilize PI controllers with a 180-degree
phase shift. The IBC is powered by a 12V input, while the
reference voltage is established at 48V. The PI controller
implemented in the two-phase interleaved boost converter
continuously modulates the duty cycles of the PWM signals
for each phase. By monitoring and regulating the discrepancy
between the measured output voltage and the desired
reference voltage, the PI controller guarantees a stable output
voltage that accurately tracks the reference. The interleaved
operation enhances power distribution and efficiency within
the converter.

437
 V. CASCADE CONTROL SIMULATIONS WITH ANN
Here, using LM algorithm, BR Algorithm, and SCG TABLE II. IBC RESULTS
Algorithm, the ANN for CBC and IBC are trained in
Control Settling Time Rise Time Peak Time
Matlab/Simulink. The suggested ANN controller for the CBC Methods (sec) (sec) (sec)
and IBC also has a dual-loop configuration.
PI 0.73 0.26 0.72
A. CBC with ANN LM 0.42 0.14 0.41
BR 0.65 0.26 0.65
The CBC in a closed-loop system utilizes inner and
outer PI controllers that simulate neural networks to obtain the SCG 0.31 0.13 0.19
reference voltage. The supply is a DC voltage, while a
reference voltage of 48V is used as a standard to compare the TABLE I and TABLE II illustrate that the ANN based
attained voltage. The supply is 12V, and the simulation controller provides better settling time, rise time, and peak
resulted in an output voltage of 48V. The inclusion of a neural time when compared with the conventional PI controller.
network tool in boost converter simulation improves the Based on the obtained results, the Scaled Conjugate Gradient
process by facilitating precise modeling, advanced control (SCG) algorithm employed in the Artificial Neural Network
design, optimization, and performance evaluation. By (ANN) based controller demonstrated improved performance
harnessing the capabilities of neural networks, the tool when compared to other ANN based algorithms and standard
effectively addresses the intricacies and nonlinearities Proportional Integral (PI) controller in both CBC and IBC.
inherent in the converter system, resulting in enhanced
analysis and design that is both efficient and effective.
B. Variations in Input Voltage
B. IBC with ANN
A 40% increase and a 20% decrease in the supply
The IBC in a closed-loop system utilizes inner and voltage were performed for 0 to 4 seconds. The output voltage
outer PI controllers that simulate neural networks to attain the waveform of the DC-DC CBC for constant reference voltage
voltage. The input voltage for this system is 12V, and the is obtained by utilizing both the conventional PI controller and
reference voltage is 48V, resulting in an output voltage of 48V ANN, while varying the input voltage.
after the simulation. The utilization of neural networks in a
two-phase interleaved boost converter offers precise nonlinear TABLE III. SETTLING TIME FOR INPUT VOLTAGE CHANGES
modeling, adaptability, advanced control design, optimization
capabilities, and improved performance evaluation and
Change in ANN
analysis. These features collectively enhance the efficiency, Input Voltage PI
performance, and optimization of the converter system. LM BR SCG
0% á 0.81 0.59 0.57 0.40
VI. SIMULATION RESULTS 25%á 1.23 1.01 0.89 0.84
The findings are tabulated for both the CBC and Two- 50%á 1.80 1.78 1.66 1.57
phase IBC. Comparing the results, it can be seen that using a 75%á 2.89 2.41 2.23 1.94
neural network in the IBC and CBC leads in shorter settling, 100%á 3.40 3.01 2.98 2.63
rising, and peak times than when using a conventional PI
controller. A description of output voltage waveforms was TABLE III illustrates the settling time obtained when
also provided in this section. the supply voltage was increased from 0% to 100% at a time
period of 0 to 4 seconds. The waveform obtained with an
A. Comparison of Time Domain Responses increase and decrease in supply voltage is depicted in Fig. 10.
The comparison of time domain response for CBC and C. Variations in Setpoint
IBC using a combination of neural network-based controllers
and PI controllers have been incorporated. The use of a neural A 40% increase and a 20% decrease in the reference
network controller has been found to decrease the settling voltage were performed for 4 to 8 seconds. Using the common
time, rising time, and peak time in both CBC and IBC as PI controller and ANN to the DC-DC CBC, the output voltage
compared to utilizing only a PI controller. waveform for reference voltage variation with constant input
voltage is obtained.
TABLE I. CBC TIME DOMAIN RESPONSE COMPARISON
TABLE IV. SETTLING TIME FOR SET[POINT VARIATIONS
Control Settling Time Rise Time Peak Time
Methods (sec) (sec) (sec) Setpoint
Variation PI ANN
PI 0.80 0.26 0.79 LM BR SCG
LM 0.59 0.27 0.59 0% á 4.72 4.6 4.42 4.31
BR 0.57 0.17 0.56 25%á 5.40 5.11 4.96 4.85

SCG 0.40 0.13 0.40 50%á 5.97 5.60 5.53 5.42

438
 75%á 6.51 6.29 6.08 5.89
100%á 7.21 6.89 6.55 6.39
TABLE IV illustrates the settling time obtained when
the reference was increased from 0% to 100% at a time period
of 4 to 8 seconds. The waveform obtained with an increase
and decrease in reference voltage is depicted in Fig. 11.
VII. CONCLUSIONS
This paper utilizes a combination of a PI controller and
ANN to model closed-loop CBC and IBC. The required
output voltage produced using a standard PI controller has a
higher settling time than an ANN. The time domain response
comparison of the PI controller and ANN based controllers is
illustrated in Fig. 9 and the numerical results were illustrated
in TABLE 1 and TABLE 2. Hence, it has been verified that
ANN exhibits a shorter settling time, as replacing the inner PI
controller with a block of neural network algorithms results in
a reduced settling time.

Fig. 10. Input Voltage Variations for CBC with PI and ANN.

Fig. 9. Time Domain Response Comparison of PI, LM, BR and SCG

Controllers.

Fig. 11. Setpoint Variations for CBC with PI and ANN

REFERENCES
[1] C. B. Ali, A. H. Khan, K. Pervez, T. M. Awan, A. Noorwali and S. A.
Shah, "High Efficiency High Gain DC-DC Boost Converter Using PID
Controller for Photovoltaic Applications," 2021 International Congress
of Advanced Technology and Engineering (ICOTEN), Taiz, Yemen,
2021, pp. 1-7, doi: 10.1109/ICOTEN52080.2021.9493476.
[2] Z. A. Ghani, E. Onn, M. Y. Zeain, Apriana Toding, F. A. Azidin, H.
Othman, Z. Zakaria, H. S. Lago, “Peripheral interface controller-based
maximum power point tracking algorithm for photovoltaic DC to DC
boost controller,” TELKOMNIKA Telecommunication, Computing,
Electronics and Control, Vol. 18, No. 1, February 2020, pp. 240~250,
ISSN: 1693-6930, accredited First Grade by Kemenristekdikti, Decree
No: 21/E/KPT/2018 DOI: 10.12928/TELKOMNIKA.v18i1.12730.

439
[3] Pereira, L.F.d.S.C.; Batista, E.; de Brito, M.A.G.; Godoy, R.B., “A [13] R. Panahidoost, H. Mirshekali, R. Dashti, R. Samsami, M. H. Rezaei
Robustness Analysis of a Fuzzy Fractional Order PID Controller Based and H. R. Shaker, "Optimized Artificial Neural Network-Based
on Genetic Algorithm for a DC-DC Boost Converter,” Electronics Control Strategy For Boost Converters," 2022 International
2022, 11, 1894. https://doi.org/ 10.3390/electronics11121894 Conference on Electrical, Computer, Communications and
[4] Y. Liang, Z. Liang, D. Zhao, Y. Huangfu, L. Guo and B. Zhao, "Model Mechatronics Engineering (ICECCME), Maldives, Maldives, 2022,
predictive control of interleaved DC-DC boost converter with current pp. 1-5, doi: 10.1109/ICECCME55909.2022.9987744.
compensation," 2019 IEEE International Conference on Industrial [14] S. Sridhar and M.G.Umamaheswari, "Design of Generalized Hopfield
Technology (ICIT), Melbourne, VIC, Australia, 2019, pp. 1701-1706, Neural Network based Fractional Order PI controller for Average
doi: 10.1109/ICIT.2019.8843692. Current mode control of Synchronous SEPIC Converter," 2020 IEEE
[5] H. Zhang, R. Ma, C. Han, R. Xie, B. Liang and Y. Li, "Advanced 7th Uttar Pradesh Section International Conference on Electrical,
Control Design of Interleaved Boost Converter for Fuel Cell Electronics and Computer Engineering (UPCON), Prayagraj, India,
Applications," IECON 2020 The 46th Annual Conference of the IEEE 2020, pp. 1-6, doi: 10.1109/UPCON50219.2020.9376499.
Industrial Electronics Society, Singapore, 2020, pp. 5000-5005, doi: [15] H. Xi and Q. Wang, "Design of Back Propagation Neural Network PID
10.1109/IECON43393.2020.9254532. Control for Boost Converter," 2021 IEEE Sustainable Power and
[6] K. Alam and A. Hoque, "Design and Analysis of Closed Loop Energy Conference (iSPEC), Nanjing, China, 2021, pp. 3889-3893,
Interleaved Boost Converter with Arduino based Soft PI Controller for doi: 10.1109/iSPEC53008.2021.9735583.
Photovoltaic Application," 2019 IEEE International Conference on [16] M. Bindi et al., "Comparison Between PI and Neural Network
Electrical, Computer and Communication Technologies (ICECCT), Controller for Dual Active Bridge Converter," 2021 IEEE 15th
Coimbatore, India, 2019, pp. 1-5, doi: International Conference on Compatibility, Power Electronics and
10.1109/ICECCT.2019.8869408. Power Engineering (CPE-POWERENG), Florence, Italy, 2021, pp. 1-
[7] A. Belkheir, B. Amar, B. Ouahid, B. Rabhi, Z. Laid and B. M. Fouad, 6, doi: 10.1109/CPE-POWERENG50821.2021.9501168.
"Modeling and control of an interleaved boost DC-DC converter [17] H. S. Khan, I. S. Mohamed, K. Kauhaniemi and L. Liu, "Artificial
applied for PEM fuel cell," 2021 IEEE 1st International Maghreb Neural Network-Based Voltage Control of DC/DC Converter for DC
Meeting of the Conference on Sciences and Techniques of Automatic Microgrid Applications," 2021 6th IEEE Workshop on the Electronic
Control and Computer Engineering MI-STA, Tripoli, Libya, 2021, pp. Grid (eGRID), New Orleans, LA, USA, 2021, pp. 1-6, doi:
80-85, doi: 10.1109/MI-STA52233.2021.9464426. 10.1109/eGRID52793.2021.9662132.
[8] N. Shanthi, P. Nivethitha, S. Sindhuja, M. Hilasini and K. [18] S. Halder, A. Khan, U. N. Thakur and S. Saha, "Design and Analysis
Divyabharathi, "High Efficient Interleaved Boost Converter for of Temperature Control System using Conventional PI and Advanced
Photovoltaic Applications," 2018 International Conference on ANN Controllers," 2018 International Conference on Computing,
Computation of Power, Energy, Information and Communication Power and Communication Technologies (GUCON), Greater Noida,
(ICCPEIC), Chennai, India, 2018, pp. 305-309, doi: India, 2018, pp. 4-8, doi: 10.1109/GUCON.2018.8675066.
10.1109/ICCPEIC.2018.8525145. [19] S. Saadatmand, P. Shamsi and M. Ferdowsi, "The Voltage Regulation
[9] A. Boyar and E. Kabalci, "Design and Analysis of a Two-Phase of a Buck Converter Using a Neural Network Predictive Controller,"
Interleaved Boost Converter Based Microinverter," 2018 IEEE 18th 2020 IEEE Texas Power and Energy Conference (TPEC), College
International Power Electronics and Motion Control Conference Station, TX, USA, 2020, pp. 1-6, doi:
(PEMC), Budapest, Hungary, 2018, pp. 1067-1072, doi: 10.1109/TPEC48276.2020.9042588.
10.1109/EPEPEMC.2018.8521982. [20] M. Y. Bote-Vazquez, J. Ramirez-Hernandez, L. Hernandez-Gonzalez,
[10] S. Gomathi, D. Arulanantham, R. Thumma, S. Vimala, S. Jenifer E. D. Delgado-Piña and O. U. Juarez-Sandoval, "Artificial Neural
Rayen and T. D. Subha, "PFC based Three Stage Interleaved Boost Network-Based Voltage Control in a DC-DC Converter using a
Converter for Renewable Energy System," 2022 6th International Predictive Model," 2022 IEEE International Autumn Meeting on
Conference on Trends in Electronics and Informatics (ICOEI), Power, Electronics and Computing (ROPEC), Ixtapa, Mexico, 2022,
Tirunelveli, India, 2022, pp. 289-295, doi: pp. 1-6, doi: 10.1109/ROPEC55836.2022.10018649.
10.1109/ICOEI53556.2022.9777218.
[11] O. K. Islam, M. S. Ahmed, K. Rahman and T. Tahsin, "A
Comprehensive Comparison Between Boost and Buck-Boost
Converters in Solar MPPT With ANN," 2020 Emerging Technology in
Computing, Communication and Electronics (ETCCE), Bangladesh,
2020, pp. 1-6, doi: 10.1109/ETCCE51779.2020.9350867.
[12] K. C. and U. M. G., "Design of Gray Wolf Optimizer Algorithm-Based
Fractional Order PI Controller for Power Factor Correction in SMPS
Applications," in IEEE Transactions on Power Electronics, vol. 35, no.
2, pp. 2100-2118, Feb. 2020, doi: 10.1109/TPEL.2019.2920971.

440