INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH

H. V. P. Nguyen et al., Vol.12, No.4, December 2022

Comparative Efficiency Assessment of MPPT

Algorithms in Photovoltaic Systems

H. V. P. Nguyen*‡ , T. T. Huynh** , V. T. Nguyen*

* Faculty of Electrical Engineering, University of Science and Technology – The University of Danang,
54 Nguyen Luong Bang, Lien Chieu, Danang, Vietnam
** Faculty of Electronic and Telecommunication Engineering,
University of Science and Technology – The University of Danang, 54 Nguyen Luong Bang, Lien Chieu, Danang, Vietnam
([email protected], [email protected], [email protected])

‡
Corresponding Author; H.V.P. Nguyen, Faculty of Electrical Engineering,
University of Science and Technology – The University of Danang, 54 Nguyen Luong Bang, Lien Chieu, Danang, Vietnam
Tel: +84 917 981 779, [email protected]

Received: 26.09.2022 Accepted: 14.11.2022

Abstract- Algorithms of maximum power point tracking are widely used in most of photovoltaic systems to optimize the output
power which depends on ambient conditions such as solar irradiance and PV arrays’ temperature. In general, these techniques
can be classified into two categories: conventional algorithms such as Perturb and Observe (P&O) and Incremental Conductance
(INC), and artificial intelligence algorithms such as Artificial Neural Network (ANN). In this investigation, a comparison of
these algorithms is conducted to analyse, compare, and assess their performances when they are integrated in a PV power system
under dynamic changed conditions. The simulation results obtained from MATLAB/Simulink environment show that the
dynamic performances of intelligent MPPT controller are much better than those of traditional algorithms such as P&O and INC.
Under any circumstances that solar radiance varies rapidly or slowly, ANN MPPT algorithms always track MPP point correctly
within the varying duration to provide maximum output power. On the contrary, P&O and INC techniques only show
effectiveness under the case of slowly changing irradiance.
Keywords Photovoltaic, MPPT, P&O, INC, Artificial Neural Network.

1. Introduction drawbacks of conventional and intelligent algorithms were

also discussed in detail. However, these studies have-not
Photovoltaic (PV) panels change sunlight irradiance to exposed the comparisons of these algorithms’ efficiency in the
electric power. The amount of generated DC power depends same condition of changed irradiation.
on the luminosity of the sun and the ambient temperature. It
In this paper, a comparative investigation of efficiency
also varies corresponding to the increasing number of loads
assessment of popular MPPT techniques is presented. The
[1-2]. Under consistent radiation and temperature, a PV panel
investigation was simulated under different scenarios of solar
exhibits characteristic voltages and currents at a single point,
radiance to find out the advantages and drawbacks of each
called the peak power point, where the PV panel delivers its
algorithm applying to a sample PV system. The simulated
highest power. To provide the maximum output, a technique
models were performed in Matlab/Simulink environment to
to track the maximum power point (MPP) is essential for these
describe behaviours of models in details.
panels. The MPPT algorithm regulates the power converter to
capture the instant peak power of the PV system continuously.
Over the years, multiple MPPT techniques have been
introduced and implemented such as Perturb and Observe
(P&O) [3-5], incremental inductance (INC) [6-8], Fuzzy logic
[9-11], or neural networks [12-13]. In [14], the advantages and
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
H. V. P. Nguyen et al., Vol.12, No.4, December 2022

2. Fundamentals of PV Array a direct coupled system), the operating point of the system will
be the intersection of the PV cell array curve and the load
2.1. PV Characteristics curve. Thus, in a direct coupled system, the PV array must be
large enough to ensure that the required power can be supplied
PV cells all share a common characteristic that the to the load. This may come to a more costly built PV system.
voltage, current, and output power are quite small, so they To surpass this issue, a power electronic converter, named
cannot be used as a single unit in practical applications. the Maximum Power Point (MPP) Tracker is required to keep
Manufacturers often combine NS PV cells in series into a the operating point of the PV battery array at the MPP point
series, and the parallel combination of NP series in each [16]. The MPPT unit does this by controlling the voltage or
product is commercialized to create a battery source with a current of the PV cell membrane. If a proper MPPT technique
larger capacity [15-16]. Equation (1) describes is applied, the MPPT unit is able to identify and trace the MPP
mathematically for the coupled structure of a PV panel which
point of PV cells.
is similar to that of a PV cell:
Depending on practical applications, two types of DC/DC
𝑉𝑃𝑉 + 𝐼𝑃𝑉 𝑅𝑠 [18-20] power conversion circuits commonly are used to make
𝐼𝑃𝑉 = 𝐼𝑝ℎ − 𝐼0 [𝑒𝑥𝑝 ( ) − 1]
𝑛𝑉𝑡ℎ MPPTs: boost converters and buck converters. It showed that
𝑉𝑃𝑉 + 𝐼𝑃𝑉 𝑅𝑠 (1)
the usage of Arduino in DC/DC converters for MPPT
− conventional and intelligent algorithms provide low effective
𝑅𝑝
where: IPV and VPV are output current and voltage, cost [21]. In this study, the Boost converter was chosen for
respectively, of PV panels; I0 is the reverse saturated current investigation.
of the diode in the equivalent circuit model; RP and RS are
equivalent parallel and series resistance, accordingly, in the 3. Maximum Power Point Tracking Method
equivalent model of PV panels; Vth is the equivalent thermal
voltage. 3.1. Perturb and Observe (P&O) Technique

The instantaneous power emitted from PV panels is P&O is a relatively simple and most widely used method.
determined by the formula: This technique considers the voltage variation according to the
cycle to find the working point with the largest capacity. The
𝑃=𝐼×𝑉 (2)
flowchart and characteristics of this algorithm are illustrated
With the relationship of current (I) and voltage (V) in
in Fig. 2 and Fig. 3, respectively [3]. If a change in voltage
equation (1), the I - V and P - V characteristics of PV panels
triggers power to increase, then the following variation will
are determined as shown in Fig. 1. It can be seen that the
stay in the same direction of increase or decrease. On the other
characteristic curve defines three special points namely short
hand, if the variation triggers the power to decrease, the
circuit, open circuit and maximum power points (MPP). These
subsequent deviation will opt to change in the opposite
special spots will be characterized by the short circuit current
direction. When the operating MPP point is identified on the
value of each panel (ISC), the parameter at MPP of each panel
P-V characteristic curve, the voltage deviation will happen
(VMPP voltage, IMPP current and PMPP capacity) and open circuit
around that maximum power operating point (MPP point) [3].
voltage of each panel (VOC).

Fig. 1. I-V and P-V characteristics of a PV panel.

2.2. MPPT Devices

The MPP point determined on the IPV - VPV and PPV - VPV
characteristic curves always changes under changing radiation
and temperature conditions. From there, we see a shift of the
MPP point once the solar irradiance or the working
temperature of the panel changes [15, 17]. In addition, when
the PV array is connected directly to the load (also known as

2062
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
H. V. P. Nguyen et al., Vol.12, No.4, December 2022

Fig. 2. P&O flow chart.

Fig.4. INC flowchart.
3.2. INC Technique
3.3. Artificial Neural Network (ANN) Algorithm
The INC technique uses the incremental total inductance
of the solar cell to identify the optimal power point is shown ANN was born from the idea of simulating the human
in Fig. 4 and Fig. 5. This method is relied on the following brain [22-25]. Like humans, ANNs are learned by experience,
characteristic: the slope of the curve pin is zero at the MPPT saving those experiences and using them in the right
point, this value is positive when locating to the left of the situations.
MPP point, negative when positioning to the right of the MPP
point [6-7]. This technique evaluates the instant inductance
value (I/V) with the increment inductance (∆I/∆V) to
determine the working point with the maximum capacity.

Fig. 5. P-V characteristics of INC algorithm.

Supervised learning is a group of algorithms that predict
the output of new input based on a number of known pairs of
data. These data pairs are called as data-label. It is the
Fig. 3. P-V characteristics of P&O algorithm. common group of machine learning techniques.
At the MPP point, the reference voltage Vref = VMPP. This Mathematically, supervised learning includes an input set
INC algorithm has the advantage of giving good results in the of “n” variables of X = {x1, x2, … ,xn} together with a
case of a correspondingly large increase in inductance value comparable set of “n” labels of Y = {y1, y2, … , yn}. Pairs of
in sudden weather conditions. known data (xi, yi) are called training values. Based on these
quantities, it is necessary to generate a function mapping a
single element of input variables to a approximately matching
element in set Y as follows:

2063
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
H. V. P. Nguyen et al., Vol.12, No.4, December 2022

using Matlab/Simulink to simulate the PV panel parameters

𝑦̂ = ∫(𝑥𝑖 ) (3) provided by the manufacturer.
The purpose of this mapping is to estimate the function f Back propagation (BP) algorithm is used to train the ANN
properly under any circumstance when a new data x is with Bayesian Regularization optimization method. The
available, the corresponding label y can be deduced from the supervised learning aims to provide it with some combination
function. of desired suggestions and associated values of the inputs.
This research study introduces and uses a regular neural Firstly, the weights are usually assigned with random values.
network for implementation because of its simple and Supervised learning is then operated to properly tune the
accessible calculation method. weights for reducing the difference between each requested
output and the result from the ANN for each relevant input.
An ANN usually organizes neurons into layers, and each
layer is responsible for a specific task. ANN usually has 3
layers: input, hidden and output layers [22]. The first layer
gives the network with the necessary data. The number of
neurons in this layer depends on input parameters provided to VPV
the network, and these input parameters are assumed to be in
vector form. The hidden layer contains hidden neurons that D
help connect input values to output values. A neural network
may have one or multiple hidden layers that are primarily IPV
responsible for processing the neurons of the input layer and
delivering the information to the neurons of the output layer.
These neurons are suitable for classifying and identifying the
relationship between input parameters and output parameters. Input Hidden Output
The output layer contains output neurons to transfer the output Layer (IL) Layer (HL) Layer (OL)
information of computations from the ANN to the user. An
ANN can be built to have multiple output parameters. W1
W2
The amount of neurons of the input and output layers is z Activation y
decided by the problem whereas the parameters of hidden =
Wn-1
function
layers are decided by the user [12-13]. However, choosing the Wn
type and quantity of input parameters has a great influence on
the quality of the network. In such a model, a neuron point is Fig. 6. Multilayer straight-propagation ANN structure.
a handling node that first linearly balances inputs, then builds
the summation using the nonlinear activation function (AF) 4. Simulation Results and Discussions
and finally, sends the outcomes to subsequent neurons. The
model of an ordinary neuron is given by equation (4) as The ANN algorithm is implemented using the “nnstart”
follows: tool in Matlab/Simulink with training data taken from the
𝑛 parameters of the solar panels. The PV system used in the
simulation here is the Average Model of 100 kW PV Array in
𝑧 = ∑ 𝑤𝑛 𝑥𝑛 + 𝛼 (4)
MATLAB/Simulink which is set up with an output power of
𝑛=1 100.7kW, an open-circuit voltage of 64.2 V and consists of 66
where z is the AF argument and x1, x2, x3,…,xn are N input parallel series, each consisting of 5 PV panels in series. The
signals, and w1, w2,…, wn are the weights of the synapses ANN training results are shown in Fig. 7. The details of
involved. simulation parameters are tabulated in Table 1.
Various activation functions are introduced such as tanh,
liner and sigmoid functions; in this study using the sigmoid
function [26] as follows:
1
𝑦 = (5)
1 + 𝑒 −𝑧
The structure of a multilayer straight-propagation ANN
counted in the investigation is depicted in Fig. 6, where the
input layers’ neurons act as buffers to deliver input signals
(VPV, IPV) or environmental conditions such as radiation and
temperature or a combination of the above [7]. There is a
neuron in the output layer that provides the VMPP value
according to the MPP or the duty cycle manipulated to control
the power converter to work closely to the MPP point. In the Fig. 7. ANN training results.
proposed ANN method, the group uses the input pair (VPV, IPV)
and the output is a duty cycle. The training data was obtained The solar radiation in the simulation scenario is
programmed to change rapidly as shown in Fig. 8. The rapid

2064
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
H. V. P. Nguyen et al., Vol.12, No.4, December 2022

change in radiation can be caused by transient shading effects

due to clouds and trees. Thereby showing the response of
ANN algorithms to environmental conditions. Simulation
with variable radiation assumptions as shown in Fig. 8 with
constant temperature of 25°C. In this radiation assumption is
divided into the following scenarios:
- Scenario 1: radiation increases rapidly with small
change (200W/m2 within 0.5s).
- Scenario 2: radiation increases rapidly with a large
change (from 200W/m2 to 1000W/m2 within 1s).
- Scenario 3: radiation increases slowly with a large Fig. 9. PV output power according to MPPT algorithms.
change (from 200W/m2 to 1000W/m2 within 10s).
In this paper, the signal tracking along the peak power line
Table 1. Simulation parameters ηMPPT is used to evaluate the efficiency of various MPPT
PV cell characteristics techniques. The tracking factor is defined as follows:
Voc (V) 64.2 𝑡2
ISC (A) 5.96 ∫𝑡1 𝑃𝑑𝑡
VMPP (V) 54.7
𝜂𝑀𝑃𝑃𝑇 = 𝑡2
(6)
IMPP (A) 8.58
∫𝑡1 𝑃𝑚𝑎𝑥 𝑑𝑡
- 66 series connected in In which t1, t2 are the starting and ending points of the
parallel survey process. P is the values obtained through MPPT
Details PV array algorithms, Pmax is the maximum output power that can be
- 5 cell connected in
series in each achieved by the PV system. We can calculate the Pmax value
Maximum output power PMPP (kW) 100.7 corresponding to each specific radiation, thereby constructing
Configuration of MPPT Boost Circuit the Pmax characteristic curve according to the survey scenario.
Switching frequency fsw 20 kHz
4.1. Scenario 1: Simulation When Radiation Increases
Sample frequency fs 10 kHz
Rapidly with Minor Changes
Inductane value 5 mH
Output Capacitance 12 mF
The irradiance changes slightly from 200 W/m2 to 400
Initial pulse width value 0.5 W/m2 in 0.5s with the consistent ambient temperature of 25°C.
Step of pulse width variation Dstep 3×10-4 The output power of the received algorithms is as shown in
Fig. 10. As can be seen in Fig. 10, the conventional P&O and
INC algorithms cannot track the MPP point correctly when the
irradiance changes rapidly with small amount from 5s to 5.5s.
Therefore, the output power of PV system suddently increase
at 5.5s when the radiance stops fluctuating. On the other hand,
the ANN algorithm always follow the MPP point in duration
from 5s to 5.5s. It is obvious that the ANN MPPT algorithm
trace the maximum power line better than classical algorithms
such as INC and P&O.

Fig. 8. Solar irradiation in each scenario.

Based on the ability to trace the MPP point of the ANN,
P&O and INC algorithms in the respective scenarios, thereby
giving the MPP tracking performance. The obtained
simulation results are shown in Fig. 9.

Fig. 10. PV output power in according to scenario 1.

4.2. Scenario 2: Simulation When Irradiance Changes

Rapidly with Substantial Changes

The irradiance changes slightly from 200 W/m2 to 1000

W/m2 in 1s while the ambient temperature is constant at 25°C.
The received output power of three algorithms is as shown in

2065
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
H. V. P. Nguyen et al., Vol.12, No.4, December 2022

Fig. 11. The obtained results show that the ANN algorithm has ANN 99.60 99.10
a much faster response. Similarly in scenario 1, conventional
MPPT algorithms, i.e. P&O and INC, cannot track the MPP P&O 99.52 34.42
point accordinly when the irradiance changes suddenly with INC 99.55 34.46
large amount from 12.4s to 13.4s. On the contrary, the ANN
always follow the MPP point in this duration.
In addition, a rapid change in radiation will result in a
value of the ratio of the oscillating power to the time of that
power oscillation (∆P/∆t) with larger P&O and INC
algorithms. a lot in ANN-based control algorithms. Therefore,
the ANN control algorithm, if applied, will reduce the
influence on the grid frequency and increase the penetration
of solar energy into the microgrid system [27].

5. Conclusion

Integrating MPPT peak power scoring algorithms into

DC - DC power converters will make it possible to get more
Fig. 11. PV output power in according to scenario 2. energy from the same amount of solar radiation. The
simulation results using MATLAB/Simulink simulation
4.3. Scenario 3: Simulation When Radiation Changes Slowly software in this paper prove that the P&O, INC and ANN
algorithms are all capable of detecting the maximum power
with Substantial Changes
point. The obtained results also show that in each different
weather condition, the algorithms will have different response
The irradiance changes slightly from 200 W/m2 to 1000
times and performance.
W/m2 in 10s at 25°C during the simulation. The output power
of all algorithms is obtained as shown in Fig. 12. The results The obtained research results confirm that the most
show that for a sufficiently slow change time, the tracking effective algorithm is the proposed ANN control algorithm
performance of the algorithms is almost similar. All three with the efficiency ηMPPT always reaching over 99% under
algorthims can track MPP point in the duration of slowly various changing radiation conditions. The current commonly
changed irradiance condition. used algorithms, i.e. P&O and INC, can only respond
effectively to the case when the radiation change is slow.
The efficiency of maximum power point tracking
algorithms is shown in the simulation results of Figs. 10, 11
and 12, corresponding to 3 scenarios of radiation changes. The Acknowledgements
simulated outcomes show that maximum power output line
tracing efficiency of the ANN algorithm is always above 99%. This work was supported by The University of Danang,
Meanwhile, the P&O and INC techniques respond well only University of Science and Technology, code number of
to the case of slow-changing radiation. With the rapidly- Project: T2021-02-34.
changing radiation case, the above algorithms do not respond
in time and the power loss increases greatly as summarised in References
Table 2.
[1] K. Çelik, M. Demirtas, and N. Öztürk, "Analytical
Investigation of PV Panel Operated at Maximum Power
Point on DC Microgrid," 2022 11th International
Conference on Renewable Energy Research and
Application (ICRERA), pp. 324-329, 2022.
[2] R. Z. Caglayan, K. Kayisli, N. Zhakiyev, A. Harrouz, and
I. Colak, "A Case Study: Standalone Hybrid Renewable
Energy Systems," 2022 11th International Conference on
Renewable Energy Research and Application (ICRERA),
pp. 284-292, 2022.
[3] A. Dolara, “Energy comparison of seven MPPT
Fig. 12. PV output power in according to scenario 3. techniques for PV systems”, Journal of Electromagnetic
Analysis and Application, vol. 01, pp. 152–162, Jan. 2009.
Table 2. Efficiency of MPPT algorithms
[4] B. Bendib, H. Belmili, and F. Krim, “A survey of the most
MPP tracking efficiency  MPPT (%) used MPPT methods: Conventional and advanced
algorithms applied for photovoltaic systems”, Renewable
Algorithm
Slowly changed Rapidly changed and Sustainable Energy Reviews, vol. 45, pp. 637-648,
irradiation irradiation 2015.

2066
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
H. V. P. Nguyen et al., Vol.12, No.4, December 2022

[5] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, Power Point Tracking Algorithm for the Photovoltaic
"Optimization of perturb and observe maximum power Power System”, Proceedings of 2019 International
point tracking method", IEEE Transactions on Power Conference on System Science and Engineering, pp. 159 –
Electronics, vol. 20, no. 4, pp. 963-973, July 2005. 163, 2019.
[6] R. Reisi, M. Hassan Moradi, and S. Jamasb, [17] F. N. Shaker, A. A. Obed, and A. J. Abid, "Comprehensive
“Classification and comparison of maximum power point Design for a Neuro-Fuzzy Controller for a Safe Hydrogen
tracking techniques for photovoltaic system: A review”, Energy Storage," 2022 10th International Conference on
Renew. Sustain. Energy Rev., vol. 19, pp. 433–443, Mar. Smart Grid (icSmartGrid), pp. 124-130, 2022.
2013.
[18] M. Q. Duong, V. T. Nguyen, G. N. Sava, M. Scripcariu,
[7] C. Hua, and C. Shen, “Comparative study of peak power and M. Mussetta, "Design and simulation of PI-type
tracking techniques for solar storage system”, APEC ’98 control for the Buck Boost converter", 2017 International
Thirteenth Annual Applied Power Electronics Conference Conference on ENERGY and ENVIRONMENT (CIEM),
and Exposition, vol. 2, pp. 679–685, 1998. pp. 79-82, 2017.
[8] C. Aoughlis, A. Belkaid, M. A. Kacimi, I. Colak, and O. [19] A. Raj, and R.P. Praveen, “Highly efficient DC-DC boost
Guenounou, "New Dynamic and Self-Adaptive converter implemented with improved MPPT algorithm
Incremental Conductance Algorithm for Standalone PV for utility level photovoltaic applications”, Ain Shams
System," 2022 11th International Conference on Engineering Journal, vol. 13(3), May 2022.
Renewable Energy Research and Application (ICRERA),
[20] N. Obeidi, M. Kermadi, B. Belmadani, A. Allag, L.
pp. 268-273, 2022.
Achour, and S. Mekhilef, “A Current Sensorless Control
[9] B. Bendib, F. Krim, H. Belmili, A. Fayçal, and B. Sabri, of Buck-Boost Converter for Maximum Power Point
“An intelligent MPPT approach based on neural-network Tracking in Photovoltaic Applications”, Energies, vol. 15,
voltage estimator and fuzzy controller, applied to a stand- no. 20, Oct. 2022.
alone PV system”, IEEE International Symposium on
[21] S. Motahhir, A. El Hammoumi, and A. El Ghzizal, “The
Industrial Electronics, pp. 404–409, 2014.
most used MPPT algorithms: Review and the suitable low-
[10] A. K. Pandey, V. Singh, and S. Jain, Study and cost embedded board for each algorithm”, Journal of
comparative analysis of perturb and observe (P&O) and Cleaner Production, vol. 246, 2020.
fuzzy logic based PV-MPPT algorithms, Applications of
[22] Y. Singh, and A. S. Chauhan, “Neural networks in data
AI and IOT in Renewable Energy, Academic Press, 2022,
mining”, Journal of Theoretical and Applied Information
ch. 11.
Technology, Vol. 5(6), pp. 36-42, 2009.
[11] T. Hai, J. Zhou, and K. Muranaka, “An efficient fuzzy-
[23] A. Mellit, and S. A. Kalogirou, “Artificial intelligence
logic based MPPT controller for grid-connected PV
techniques for photovoltaic applications: a review”,
systems by farmland fertility optimization algorithm”,
Progress in Energy and Combustion Science, Vol. 34(5),
Optik, vol. 267, 2022.
pp. 574 - 632, 2008.
[12] M. Laurino, M. Piliougine, and G. Spagnuolo, “Artificial
[24] M. Fathi, and J. A. Parian, “Intelligent MPPT for
neural network based photovoltaic module diagnosis by
photovoltaic panels using a novel fuzzy logic and artificial
current–voltage curve classification”, Solar Energy, vol.
neural networks based on evolutionary algorithms”,
236, pp. 383-392, 2022.
Energy Reports, vol. 7, pp. 1338-1348, 2021.
[13] A. Pamain, P.V. Kanaka Rao, and F.N. Tilya, “Prediction
[25] C. G. Villegas-Mier, J. Rodriguez-Resendiz, J. M.
of photovoltaic power output based on different non-linear
Álvarez-Alvarado, H. Rodriguez-Resendiz, A. M.
autoregressive artificial neural network algorithms”,
Herrera-Navarro, and O. Rodríguez-Abreo, “Artificial
Global Energy Interconnection, vol. 5(2), pp. 226-235,
Neural Networks in MPPT Algorithms for Optimization of
2022.
Photovoltaic Power Systems: A Review”, Micromachines,
[14] N. A. Kamarzaman and C. W. Tan, “A comprehensive vol. 12, no. 10, Oct. 2021.
review of maximum power point tracking algorithms for
[26] K. Jain, J. Mao, and K. M. Mohiuddin, “Artificial neural
photovoltaic systems”, Renewable and Sustainable Energy
networks: a tutorial”, Computer, vol. 29(3), pp. 31 – 44,
Reviews, vol. 37, pp. 585-598, 2014.
March 1996.
[15] N. Nguyen, V. T. Nguyen, M. Q. Duong, K. H. Le, H. H.
[27] V. T. Nguyen, D. H. Hoang, H. H. Nguyen, K. H. Le, T.
Nguyen, and A. T. Doan, “Propose a MPPT Algorithm
K. Truong, and Q. C. Le, "Analysis of Uncertainties for the
Based on Thevenin Equivalent Circuit for Improving
Operation and Stability of an Islanded Microgrid," 2019
Photovoltaic System Operation”, Front. Energy Res., vol.
International Conference on System Science and
8, p. 14, 2020.
Engineering (ICSSE), pp. 178-183, 2019.
[16] N. Nguyen, V. K. Pham, V. T. Nguyen, D. H. Hoang, T.
B. T. Truong, and H. V. P. Nguyen, “A New Maximum

2067