International Journal of Power Electronics and Drive System (IJPEDS)

Vol. 14, No. 3, September 2023, pp. 1489∼1496

ISSN: 2088-8694, DOI: 10.11591/ijpeds.v14.i3.pp1489-1496 ❒ 1489

Application of machine learning controller in matrix

converter based on model predictive control algorithm
Yasoda Kailasa Gounder, Sowkarthika Subramanian
Department of Electrical and Electronics Engineering, Government College of Technology, Coimbatore, India

Article Info ABSTRACT

Article history: Finite control set model predictive control (FCS-MPC) algorithms are famous
in power converter for its easy implementation of constraints with cost function
Received Sep 30, 2022
than classical control algortihms. However computation complexity increases
Revised Dec 26, 2022 when swicthing state is high for converters such as matrix converter, multilevel
Accepted Jan 7, 2023 converters and this impose a serious drawback to compute multi-step prediction
horizon MPC algorithm which further increases the computation. To overcome
Keywords: the above said difficulty, machine learning based artificial neural network (ANN)
controller for matrix converter is proposed. The training data for ANN controller
Artificial neural network is derived from MPC algorithm and trained offline with an accuracy of 70.3%.
Finite control set model predictive The proposed ANN controller shows a similar and better performance than MPC
Current control controller in terms of total harmonic distortion (THD), peak overshoot during
Hardware in-loop dynamic change in reference current and dynamic change in load parameter and
Matrix converter less computation with less execution time. Further, ANN controller for matrix
Total harmonic distortion converter is tested in OPAL-RT using hardware in-loop (HIL) simulation and
showed that it outperforms MPC controller.

This is an open access article under the CC BY-SA license.

Corresponding Author:
Yasoda Kailasa Gounder
Department of Electrical and Electronics Engineering, Government College of Technology
Coimbatore 641013, Tamil Nadu, India
Email: [email protected]

1. INTRODUCTION
Model predictive control of matrix converter have gained a lot of interest in past years as it is simple
with good and fast dynamic response with inherent reactive power control. Mir et al. [1] proposed improvised
multi-objective finite control set model predictive control which relatively results in reduced total harmonic
distortion (THD) of source current by 5%. Model predictive control (MPC) at each step selects an optimal
switching state in accordance with pre-defined cost function. Though MPC offers several advantages such as
fast transient response, easy and straightforward constraints with implementation, it suffers when the number of
switching state is high. One such is matrix converter, as the valid switching state of three phase matrix converter
is 27. In each sampling time, the cost function is calculated for the 27 possible switching state and one optimal
switching state is selected to be applied in the next sampling time. This will increase the computation burden
of controller and the case would be worser if multi-horizon predictions are employed.
Supervised machine learning techniques is very popular in the application of power electronic convert-
ers as it reduces the complexity and process the results in a very short time [2]-[4]. ANN is aided in real time
modeling apprach for power electronic converters to model thermal stress, and switching loss, and reported
a unique performance in computation burden and resource utilisation of FPGA [5]. Khan et al. [6] applied
artificial neural network (ANN) to control voltage in DC-DC converter in DC microgrid applications and found

Journal homepage: http://ijpeds.iaescore.com

1490 ❒ ISSN: 2088-8694

ANN’s performance is better in terms of accuracy and computational burden. ANN has also been applied to
optimise the parameters of permanent magnet synchornous motor based model predictive control [7]. Akpolat
et al. [8] proved that ANN has also found application in reducing the number of sensor in the control of DC
microgrid and thus makes the system more reliable. Simonetti et al. [9] applied machine learning techniques
to reduce the computational burden for a cascaded H-bridge inverter instead of MPC control.
In recent years, supervised imitation learning of model predictive control is very popular which results
in less computation burden, improved total harmonic distortion in most of the test cases and same dynamic per-
formance as that of MPC. Supervised imitation learning is applied to three phase inverter with an output LC
filter which results in less harmonic distortion of output voltage in most of the test cases [10]. Time delayed
ANN is proposed and used as controller in grid-tied three level neutral point clamped transformerless inverter
[11]. Novak and Dragicevic [12] tested neural network controller for 2 step prediction horizon of three phase
inverter and proves that neural network controller outperforms MPC controller. Wang et al. [13] applied ANN
in place of MPC for power converters which results in less resource utilization of FPGA comparatively. Zaid
et al. [14] used ANN controller as end-to-end learning policy to control transformerless grid connected neutral
pointed clamped inverter and stated that ANN controller results with low harmonic distortion with enhanced
power quality and minimized leakage current than MPC. Abu-Ali et al. [15] tested deep learning controller
for permanent magnet synchronous motor drives and results with better torque transient response than conven-
tional MPC. Ahmed et al. [16] proposed recurrent neural network based predictive current control with better
dynamics, excellent control and tracking error. Akpolat et al. [17] ANN-MPC instead of proportional integral
(PI) controller for stabilization of DC microgrid and noted that ANN-MPC resulted in less instability issue and
oscillations in DC microgrid. Sabzevari et al. [18] proposed state-space recurrent neural network controller
for three phase power converter and showed that the control scheme is more robust compared to conventional
MPC. Sahu et al. [19] developed neural network based dicrete model predictive controller for induction motor
drive based on direct torque and flux and reported reduction of ripples in flux, torque and current compared
with conventional PI direct torque and flux control. Further, the authors used various intelligence techniques
in power converters [20]-[24]. The above literature motivated to train ANN controller for a matrix converter
which has higher switching possibilities and to test the performance of ANN controller for a matrix converter.
In this paper, a supervised machine learning based controller is developed and trained for matrix con-
verter of 27 valid switching states with the data collected from model predictive control. The main contributions
of the paper are as follows:
− Model free ANN based controller is tested for matrix converter in MATLAB simulation and hardware in-
loop (HIL) using RT-LAB.
− ANN based controller results in low THD in most of the test cases and dynamic performance is improved.
− Computation time and average switching frequency are compared which results in less computation burden.

2. MATRIX CONVERTER AND ITS SWITCHING STATE

Power circuit of matrix converter as shown in Figure 1 has a set of bidirectional switch with two power
transistor and anti-parallel diodes (S) which are connected to input supply through LC filter. To avoid abrupt
interruption of load current atleast one switch in the load phase must always be on as mentioned in (1).
Sxu + Syu + Szu = 1 ∀u ∈ a, b, c (1)
Based on the above restriction, 27 switching combinations are possible. Input and load voltages are related as
(2).
     
va(t) Sxa Sya Sza vex(t)
 vb(t)  =  Sxb Syb Szb  · vey(t)  (2)
vc(t) Sxc Syc Szc vez(t)
Where va (t), vb (t), and vc (t) forms the load voltage vector and vex (t), vey (t), and vez (t) forms the input
voltage vector. By KCL, input and load current are relates as (3).
     
iex(t) Sxa Sya Sza ia(t)
iey(t)  =  Sxb Syb Szb  ·  ib(t)  (3)
iez(t) Sxc Syc Szc ic(t)

Int J Pow Elec & Dri Syst, Vol. 14, No. 3, September 2023: 1489–1496
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1491

Where ia (t), ib (t), and ic (t) forms the output current vector and iex (t), iey (t), and iez (t) forms the input
current vector.

Figure 1. Power circuit of matrix converter

3. CONVENTIONAL FINITE CONTROL SET MODEL PREDICTIVE CONTROL OF MATRIX

CONVERTER
Model predictive control uses discretized system model to predict the future behaviour of the variables
for a certain time horizon. The predicted variables are used in all possible 27 switching state and one switching
state is selected to be applied at the next sampling time based on the minimisation of cost function. The
objectives of finite control set-model predictive control (FCS-MPC) are to control output current, control of
input current with less harmonics and unity power factor.
Filter current and capacitor voltage differential equations are used to predict the future behaviour of
source current is (k + 1).
1
" # " #  " #
dvf (t) 0 −1 
0 C

dt Cf vf (t) vs (t)
dis (t) = −1 −Rf + 1 f
(4)
L L
is (t) Lf 0 ii (t)
dt f f

The (4) is discretized using euler forward disctretization method and source current is (k + 1) is predicted.
Reactive power Q(k + 1) is predicted with (5).

Q(k + 1) = vsβ (k + 1)isα (k + 1) − vsα (k + 1)isβ (k + 1) (5)

Where α and β are the real and imaginary parts of the vector. The source current is (k + 1) is predicted from
(4) using the discretised model and the source voltage vs (k + 1) = vs , as source voltages are low frequency
components. The load current is predicted in (6).
RTs Ts
io (k + 1) = (1 − )io (k) + (vo (k) − e(k)) (6)
L L
Where Ts is the sampling time. The FCS-MPC algorithm of matrix converter are detailed below in steps.
− The controlled variables such as filter capacitor voltage vf (k), source current is (k), source voltage vs (k),
and output current io (k) are measured at an instant k.
− The system model is discretized from (4) to predict is (k + 1) and Q(k + 1).
− Load current io (k + 1) is predicted as in (6).

Application of machine learning controller in matrix converter based ... (Yasoda Kailasa Gounder)
1492 ❒ ISSN: 2088-8694

− Cost function C to track the reference load current and to maintain unity power factor at source side is:

C = (i∗oα − ipoα ) + (i∗oβ − ipoβ ) + λ(Q∗ − Qp ) (7)

where the superscript p represents the predicted quantity, λ is the weighing factor. Q∗ is maintained at zero
to achieve unity power factor.
− Cost function is evaluated for all 27 possible switching state and the optimum switching state that minimises
the cost function is selected to be applied at next sampling instant.

4. NEURAL NETWORK BASED MODEL PREDICTIVE CONTROLLER

4.1. ANN based controller
In general, ANN controller can be trained and tested with a given set of data. Model predictive control
scheme uses this feature of ANN to imitate as that of MPC controller. Initially model is made to run under
model predictive control scheme and the training datas are generated. With the generated data, ANN is trained
to predict next optimal switching state. Input data generated from MPC scheme to train ANN are desired output
reference current i∗o , filter voltage vf , filter current if , source voltage vs , source current is , output voltage vo ,
output current io , and the previous optimum switching state Sopt−prev . Output data from MPC scheme to train
ANN is optimum switching state at next instant k+1 Sopt(k+1) . Training conditions for ANN based controller
are listed in Table 1 with a total of 100001 samples. Of this 80% and 20% samples are used for training and
testing phases respectively. Samples are trained until it reaches an accuracy of 70.3% with 50.03 epochs. Even
with minimal samples, ANN is trained perfectly to achieve the performance of MPC controller.
ANN is trained to work with shallow layer (i.e only one hidden layer) initially. But the performance is
poor as compared with MPC controller as the number of possible output states are 27. Hence two hidden layers
are used in this paper with 45 and 15 units respectively. Input and output layer has 15 and 27 units respectively.

Table 1. Training parameters for ANN controller

i∗o (A) R (Ω) L (mH) Ts (µs)
14 3.33 10 10
12 5 15 10
10 5 15 10
8 5 15 10
8 10 30 10
7 10 30 10

5. RESULTS AND DISCUSSION

ANN controller is trained with input and output parameters for conditions mentioned in Table 1. The
controller is tested for the following cases.
− Dynamic performance analysis with respect to change in load and reference current.
− Harmonic analysis for different loads.
− Computational burden of ANN and MPC controller
The nominal parameters used in simulation are source voltage: 230 V, Ts of predictive algorithm: 10 µs, Lf :
400 µH, Rf : 0.5 Ω, Cf : 61 µF, R: 0.5 Ω, and L: 30 mH. The weighing factor λ as in (7) that results in reduction
of input current THD and load current THD is 0.0045V − 1 [25].
5.1. Harmonic analysis for different loads
Performance of ANN controller and MPC controller are listed in Table 2 for different load conditions.
ANN controller is trained with only 6 conditions as in Table 1 and is able to achieve similar performance as
that of MPC controller for the un-trained cases. Normally in model predictive control method, higher step
prediction horizon increases computation burden but results with good performance. Hence data from MPC
2-step prediction horizon is used as training data to train ANN controller. Total harmonic distortion of MPC
and ANN controller are almost similar and in most of the cases, ANN controller’s source current harmonic is
less on compared with MPC controller. The disadvantage of ANN controller is a slight increase in average
switching frequency than MPC controller.

Int J Pow Elec & Dri Syst, Vol. 14, No. 3, September 2023: 1489–1496
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1493

Table 2. Performance comparison of MPC and ANN controller

i∗o R L MPC-1 Step MPC-2 Step ANN
THD fsw(avg) THD fsw(avg) THD fsw(avg)
A Ω mH % kHz % kHz % kHz
12 5 18 2.23 12.2 2.09 12.4 1.85 12.4
9.5 8 25 1.91 12.8 1.98 13.02 1.58 13.05
11 5 20 2.49 13.2 2.52 13.6 2.06 13.6
9 5 20 4.24 12.19 3.27 12.58 3.05 12.9
8 7 23 3.75 12.04 2.47 12.3 2.28 12.1
9 7 23 2.44 12.5 2.24 12.9 2.12 12.7
10 6 20 2.40 12.8 2.33 13.2 2.21 13.2
6 12 25 33.28 10.5 3.31 9.6 2.42 9.7

5.2. Dynamic performance comparison of ANN and MPC controller

Performance of MPC and ANN controller are compared for sudden change in reference load current
and for sudden change in load parameters. Figures 2 and 3 shows the dynamic performance in source and load
current of ANN and MPC controller for sudden change in reference load current from 14 A to 9 A at 0.06
seconds. Similar performance is noted in both controller in terms of its dynamic response. ANN controller’s
peak overshoot at 0.06 seconds is quite less when compared with MPC controller as in Figure 2. Also in terms
of harmonic analysis for source current, THD of MPC controller is 2.17% and ANN controller is 1.47% when
reference load current is 14 A and 3.28% for MPC controller and 3.12% for ANN controller when reference
load current is 9 A. In most of the cases, ANN controller outperforms MPC controller in harmonic performance
of source current. Figures 4 and 5 shows the dynamic performance in source and load current of ANN and MPC
controller for sudden change in R=5 Ω and L=20e−3 H to R=2.5 Ω and L=10e−3 H at 0.06 seconds. Similar
performance is noted in both controller in terms of its dynamic response. Also in terms of harmonic analysis
for source current, THD of MPC controller is 2.17% and ANN controller is 1.47% when reference R=5 Ω and
L=20e−3 H and 3.35% for MPC controller and 3.13% for ANN controller when R=2.5 Ω and L=10e−3 H. The
above results confirms that ANN controller can be a good alternative for MPC controller as the performance
is similar and comparable. ANN controller is trained with Q∗ = 0 as in (7) to maintain unity power factor.
Source voltage and source current in Figure 6 are in-phase with the trained ANN controller.

Figure 2. Source and load current for sudden change in reference load current from 14 A to 9 A at 0.06
seconds for R=5 Ω and L=20e−3 H in ANN controller

Figure 3. Source and load current for sudden change in reference load current from 14 A to 9 A at 0.06
seconds for R=5 Ω and L=20e−3 H in MPC controller

Application of machine learning controller in matrix converter based ... (Yasoda Kailasa Gounder)
1494 ❒ ISSN: 2088-8694

Figure 4. Source and load current for sudden change in R=5 Ω and L=20e−3 H to R=2.5 Ω and L = 10e−3 H
in ANN controller

Figure 5. Source and load current for sudden change in R=5 Ω and L=20e−3 H to R=2.5 Ω and L=10e−3 H in
MPC controller

Figure 6. Source and load current in ANN controller for Q∗ = 0

5.3. Computational burden of ANN and MPC controller

MPC controller calculates (5)-(7) for all possible 27 switching states and selects an optimal switching
state for every instant. For a two step prediction horizon, MPC controller calculates (5)-(7) for 27*27 times.
Whereas ANN controller has 45 units and 15 units in two hidden layers in which computational burden is
quite lesser than MPC controller and shows a similar performance as that of MPC controller. To calculate
computational burden, MATLAB profiling tools are used. ANN and MPC controller is simulated for 0.5
seconds in simulation time. The simulation execution time of whole MPC 2-step prediction horizon, MPC
1-step prediction horizon and ANN model are 24.27 seconds, 8.98 seconds, and 6.16 seconds respectively, of
which the execution time of MPC 2-step controller, MPC 1-step controller and ANN are 14.62 seconds, 3.98
seconds, and 0.99 seconds respectively. This proves that ANN controller’s computation burden is comparitively
less when compared to MPC controller.

Int J Pow Elec & Dri Syst, Vol. 14, No. 3, September 2023: 1489–1496
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1495

5.4. HIL validation of the proposed model

Further the model has been validated in hardware in-loop (HIL) simulation through OPAL-RT and
presented in Figure 7. δI in source current for sudden change in reference load current as in Figure 7(a) is 10.6
A for ANN controller whereas δI in source current for sudden change in reference load current as in Figure
7(b) is 30.2 A for MPC controller. Dynamic performance of ANN controller is best in terms of peak overshoot.

(a) (b)

Figure 7. Results from HIL simulation - source current for sudden change in reference load current from 14 A
to 9 A for R=5 Ω and L=20e−3 H in (a) ANN controller and (b) MPC controller

6. CONCLUSION AND FUTURE SCOPE

This paper evaluates the performance of machine learning based controller for matrix converter using
model predictive control algorithm. A similar dynamic performance with sudden change in reference load
current and with sudden change in load parameters are noted for ANN controller. The main advantage of
ANN controller is reduction in harmonics in most of the tested cases, reduced peak overshoot during dynamic
performance assessment and less computation burden. But ANN controllers suffers from slight increment in
average switching frequency which may increase swicthing loss comparatively. The future improvement can
be made by testing other advanced machine learning techniques in power converters.

REFERENCES
[1] T. N. Mir, B. Singh, and A. H. Bhat, “Improvised multi-objective model predictive control of matrix converter using fuzzy logic and
space vectors for switching decisions,” IET Power Electronics, vol. 13, no. 4, pp. 758–764, 2020, doi: 10.1049/iet-pel.2018.5873.
[2] B. K. Bose, “Artificial neural network applications in power electronics,” IECON’01. 27th Annual Conference of the IEEE Industrial
Electronics Society, 2001, vol. 3, pp. 1631-1638, doi: 10.1109/IECON.2001.975533.
[3] K. T. Chau and C. C. Chan, “Real-time implementation of an on-line trained neural network controller for power electron-
ics converters,” Proceedings of 1994 Power Electronics Specialist Conference - PESC’94, 1994, vol. 1, pp. 321-327, doi:
10.1109/PESC.1994.349714.
[4] B.-R. Lin and R. G. Hoft, “Neural networks and fuzzy logic in power electronics,” Control Engineering Practice, vol. 2, no. 1, pp.
113–121, 1994, doi: 10.1016/0967-0661(94)90579-7.
[5] H. Bai, C. Liu, E. Breaz, and F. Gao, “Artificial neural network aided real-time simulation of electric traction system,” Energy and
AI, vol. 1, 2020, doi: 10.1016/j.egyai.2020.100010.
[6] H. S. Khan, I. S. Mohamed, K. Kauhaniemi, and L. Liu, “Artificial neural network-based voltage control of DC/DC con-
verter for DC microgrid applications,” 2021 6th IEEE Workshop on the Electronic Grid (eGRID), 2021, pp. 1-6, doi:
10.1109/eGRID52793.2021.9662132.
[7] L. Liao, L. Feng, Y. Wen, and K. Du, “Application of neural network in parameters optimization of permanent magnet synchronous
motor model predictive control,” 2021 IEEE International Conference on Predictive Control of Electrical Drives and Power Elec-
tronics (PRECEDE), 2021, pp. 744-747, doi: 10.1109/PRECEDE51386.2021.9681029.
[8] A. N. Akpolat et al., “Sensorless control of DC microgrid based on artificial intelligence,” IEEE Transactions on Energy Conversion,
vol. 36, no. 3, pp. 2319-2329, 2021, doi: 10.1109/TEC.2020.3044270.
[9] F. Simonetti, G. D. Di Girolamo, A. D’Innocenzo, and C. Cecati, “Machine learning for model predictive control of cascaded H-
bridge inverters,” IEEE 21st Mediterranean Electrotechnical Conference (MELECON), 2022, pp. 1241-1246, doi: 10.1109/MELE-
CON53508.2022.9842941.
[10] I. S. Mohamed, S. Rovetta, T. D. Do, T. Dragicević, and A. A. Z. Diab, “A neural-network-based model predictive control of three-
phase inverter with an output LC filter,” in IEEE Access, vol. 7, pp. 124737-124749, 2019, doi: 10.1109/ACCESS.2019.2938220.

Application of machine learning controller in matrix converter based ... (Yasoda Kailasa Gounder)
1496 ❒ ISSN: 2088-8694

[11] A. Bakeer, I. S. Mohamed, P. B. Malidarreh, I. Hattabi, and L. Liu, “An artificial neural network-based model predictive con-
trol for three-phase flying capacitor multilevel inverter,” in IEEE Access, vol. 10, pp. 70305-70316, 2022, doi: 10.1109/AC-
CESS.2022.3187996.
[12] M. Novak and T. Dragicevic, “Supervised imitation learning of finite-set model predictive control systems for power electronics,”
in IEEE Transactions on Industrial Electronics, vol. 68, no. 2, pp. 1717-1723, 2021, doi: 10.1109/TIE.2020.2969116.
[13] D. Wang et al., “Model predictive control using artificial neural network for power converters,” in IEEE Transactions on Industrial
Electronics, vol. 69, no. 4, pp. 3689-3699, 2022, doi: 10.1109/TIE.2021.3076721.
[14] S. A. Zaid et al., “From MPC-based to end-to-end (E2E) learning-based control policy for grid-tied 3L-NPC transformerless in-
verter,” in IEEE Access, vol. 10, pp. 57309-57326, 2022, doi: 10.1109/ACCESS.2022.3173752.
[15] M. Abu-Ali, F. Berkel, M. Manderla, S. Reimann, R. Kennel, and M. Abdelrahem, “Deep learning-based long-horizon MPC: robust,
high performing, and computationally efficient control for PMSM drives,” in IEEE Transactions on Power Electronics, vol. 37, no.
10, pp. 12486-12501, 2022, doi: 10.1109/TPEL.2022.3172681.
[16] H. M. Ahmed, I. Jlassi, A. J. M. Cardoso, and A. Bentaallah, “Model-free predictive current control of synchronous reluctance
motors based on a recurrent neural network,” in IEEE Transactions on Industrial Electronics, vol. 69, no. 11, pp. 10984-10992,
2022, doi: 10.1109/TIE.2021.3120480.
[17] A. N. Akpolat et al., “Dynamic stabilization of DC microgrids using ANN-based model predictive control,” in IEEE Transactions
on Energy Conversion, vol. 37, no. 2, pp. 999-1010, 2022, doi: 10.1109/TEC.2021.3118664.
[18] S. Sabzevari, R. Heydari, M. Mohiti, M. Savaghebi, and J. Rodriguez, “Model-free neural network-based predictive control for
robust operation of power converters,” Energies, vol. 14, no. 8, p. 2325, 2021, doi: 10.3390/en14082325.
[19] A. Sahu, K. B. Mohanty, and R. N. Mishra, “Development and experimental realization of an adaptive neural-based discrete model
predictive direct torque and flux controller for induction motor drive,” Applied Soft Computing, vol. 108, p. 107418, 2021, doi:
10.1016/j.asoc.2021.107418.
[20] S. Reese and D. Maksimovic, “An approach to DC-DC converter optimization using machine learning-based component mod-
els,” 2022 IEEE 23rd Workshop on Control and Modeling for Power Electronics (COMPEL), 2022, pp. 1-8, doi: 10.1109/COM-
PEL53829.2022.9830017.
[21] M. I. F. M. Hanif, M. H. Suid, and M. A. Ahmad, “A piecewise affine PI controller for buck converter generated DC motor,”
International Journal of Power Electronics and Drives, vol. 10, no. 3, pp. 1419–1426, 2019, doi: 10.11591/ijpeds.v10.i3.pp1419-
1426.
[22] K. Premkumar, M. Vishnupriya, T. Thamizhselvan, P. Sanjeevikumar, and B. V. Manikandan, “PSO optimized PI controlled DC-DC
buck converter-based proton-exchange membrane fuel cell emulator for testing of MPPT algorithm and battery charger controller,”
International Transactions on Electrical Energy Systems, vol. 31, no. 2, p. e12754, 2020, doi: 10.1002/2050-7038.12754.
[23] M. A. Khazain, N. M. Hidayat, K. Burhanudin, and E. Abdullah, “Boost converter of maximum power point tracking (MPPT) using
particle swarm optimization (PSO) method,” 2021 IEEE 12th Control and System Graduate Research Colloquium (ICSGRC), 2021,
pp. 281-286, doi: 10.1109/ICSGRC53186.2021.9515228.
[24] L. Wei and S. Jing, “Improvement of fuzzy controller with maximum power point tracking technology,” Proceedings of 2013 2nd In-
ternational Conference on Measurement Information and Control ICMIC, 2013, pp. 1246–1250, doi: 10.1109/MIC.2013.6758185.
[25] R. Vargas, J. Rodriguez, U. Ammann, and P. W. Wheeler, “Predictive current control of an induction machine fed by a matrix
converter with reactive power control,” in IEEE Transactions on Industrial Electronics, vol. 55, no. 12, pp. 4362-4371, 2008, doi:
10.1109/TIE.2008.2006947.

BIOGRAPHIES OF AUTHORS

Yasoda Kailasa Gounder received the B.E. degree in Electrical Engineering at Coim-
batore Institute of Technology, Coimbatore, India, M.E. degree in Power Electronics and Drives
from Alagappa Chettiar College Engineering and Technology, Karaikudi, India and Ph.D. from Anna
University, Chennai, India. Currently, she is working as an assistant professor (senior grade) in De-
partment of Electrical Engineering at Government College of Technology, Coimbatore, India. Her
research interests are wind energy conversion systems, power electronics, and microgrids. She can
be contacted at email: [email protected].

Sowkarthika Subramanian received the B.E. degree in electrical engineering from Ku-
maraguru College of Technology and M.E. degree in power electronics and drives in PSG College of
Technology. She is currently working as assistant professor in Government College of Technology.
Her research interests include AI based power electronics applications to renewable energy systems.
She can be contacted at email: [email protected].

Int J Pow Elec & Dri Syst, Vol. 14, No. 3, September 2023: 1489–1496