International Journal of Power Electronics and Drive System (IJPEDS)

Vol. 16, No. 2, June 2025, pp. 1174~1183

ISSN: 2088-8694, DOI: 10.11591/ijpeds.v16.i2.pp1174-1183  1174

Softplus function trained artificial neural network based

maximum power point tracking

Liong Han Wen, Mohd Rezal Mohamed

Electrical, Electronics and Automation Section, Malaysian Spanish Institute, Universiti Kuala Lumpur, Kulim, Malaysia

Article Info ABSTRACT

Article history: To optimize the electrical output of a photovoltaic system, maximum power
point tracking (MPPT) methods are commonly employed. These techniques
Received Aug 15, 2024 work by operating the photovoltaic system at its maximum power point
Revised Feb 4, 2025 (MPP), which varies based on environmental factors like solar irradiance and
Accepted Mar 1, 2025 ambient temperature, thereby ensuring optimal power transfer between the
photovoltaic system and the load. In this paper, an artificial neural network
(ANN) is selected as an MPPT technique. The main contribution of the work
Keywords: is to introduce a softplus function trained artificial neural network-based
maximum point tracking (SP-ANN MPPT). The proposed method is then
Artificial neural network compared with a sigmoid function trained artificial neural network-based
MPPT maximum point tracking (SM-ANN MPPT). The simulation and experimental
Photovoltaic system results show that SP-ANN MPPT is able to track high power than SM-ANN
Sigmoid function MPPT in different conditions.
Softplus function
This is an open access article under the CC BY-SA license.

Corresponding Author:
Liong Han Wen
Electrical, Electronics and Automation Section, Malaysian Spanish Institute, Universiti Kuala Lumpur
Kulim Hi-Tech Park, 09000 Kulim, Kedah, Malaysia
Email: [email protected]

1. INTRODUCTION
The demand for energy is increasing due to the increase in population, industrial, and economic
development. Fossil fuel is one of the important sources of energy. However, the storage of fossil fuel in this
world is limited, and the usage of fossil fuel is polluting our world. Besides, the inconsistency of fossil fuel
prices will affect the economic stability of a country. Therefore, the issue of the development of renewable
energy is gaining attention among developed and developing countries. The main sources of renewable energy
can be found in the form of solar, hydropower, wind, and bioenergy. Solar energy is more area-independent than
other renewable energy resources. Almost everywhere on earth, receiving sunlight can harness solar energy, and
it does not pose an environmental threat. Bulky mechanical generators are not required for solar energy
harvesting, as with wind, hydro, and vibration energy harvesting systems, and these systems may suffer wear
and tear after a long period of time. Therefore, the maintenance costs of parts replacement are relatively high
compared to photovoltaic system. Figure 1 illustrates the general block diagram of a photovoltaic system [1].
The input and output of the DC/DC converter is connected to the photovoltaic array and load, respectively. The
maximum power point tracking (MPPT) controller generates an appropriate duty cycle for the DC/DC converter
so that it can extract maximum power from the photovoltaic (PV) array. The input of the MPPT controller can
be the PV array current and voltage [2], [3] or light irradiance intensity and ambient temperature [4], [5].
Partial shading condition (PSC) is one of the common problems in PV systems. PSC occurs when the
PV array operates in non-uniform irradiance and temperature situations such as shading from trees, passage of
clouds, and internal cell mismatch. The maximum power point (MPP) of a PV system depends on shading
patterns on each individual PV module during PSC. Figures 2(a) and 2(b) illustrate the power vs duty cycle

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

Int J Pow Elec & Dri Syst ISSN: 2088-8694  1175

curves with different shading patterns for a photovoltaic system in which two PV modules are connected in
series. From Figure 2, it is shown that the duty cycle of MPP varies as the shading pattern changes. Hence,
maximum power point tracking (MPPT) is a necessary technique in a photovoltaic system to track the MPP.
There are different techniques of MPPT [6]−[9]. An artificial neural network (ANN) is used in this work as an
MPPT technique. The primary benefits of using the ANN technique with PV systems are its independence
from a deep understanding of the internal system characteristics and its ability to identify nonlinear
relationships between dependent and independent variables.

(a) (b)

Figure 1. Typical block diagram of a Figure 2. Power vs duty cycle of a PV system for
photovoltaic system (a) shading pattern A and (b) shading pattern B

2. THEORITICAL ANALYSIS
2.1. Artificial neural network
An artificial neural network (ANN) is an information processing model similar to the human brain
[10], [11]. ANN normally starts with the input layer and ends with the output layer. There is/are one or more
hidden layers between the input and output layers, showing an ANN with one hidden layer. Figure 3 illustrates
an ANN structure with one hidden layer.
The input layer received input data which it is preferable that the range is between zero and one. ANN
is composed of a large number of interconnected processing units, which are called neurons, to solve a
particular problem. Each neuron is information connected to other neurons by a connection link. Each
connection link is associated with weights that contain information about the input signal. This information is
utilized by the neural network to address a specific problem. The behavior of an artificial neural network (ANN)
is defined by its capability to learn, recall, and generalize training patterns or data, mimicking the functionality
of the human brain. Therefore, the processing elements in an artificial neural network (ANN) are referred to as
neurons. Each neuron possesses its own internal state, which is known as the neuron's activation level. The
activation signal of a neuron is transmitted to other neurons. Figure 4 shows the fundamental structure of a
neuron in ANN and the linear relationships between the input and output of a neuron. The activation function
is to introduce non-linearity in the network. The output of neurons can be presented as [12], y = f(z), with (1).

𝑧 = ∑𝑀
𝑚=1 𝑤𝑚 𝑥𝑚 + 𝑏 (1)

Where y is the output of the neuron, f is the activation function, m is the number of inputs, wm is the weight
corresponding to the incoming input xm, and b is an offset.

2.2. Activation functions

2.2.1. Sigmoid
The sigmoid activation function is a commonly used activation function in neural networks [13]. It
maps the input value to a range between 0 and 1. Mathematically, it is expressed as (2) [14].
1
𝑓(𝑥) = (2)
1+𝑒 −𝑥

The sigmoid function is characterized by its S-shaped curve, as shown in Figure 5. It takes any real-valued
input and squashes it to the (0, 1) range, making it useful for model probabilities or binary classifications.

2.2.2. Hyperbolic tangent

The hyperbolic tangent activation function, commonly known as Tanh, is another popular choice.
Tanh is an activation function that is similar to the sigmoid function, but it maps its input to a range between -

Soft plus function trained artificial neural network based maximum power … (Liong Han Wen)
1176  ISSN: 2088-8694

1 and 1, providing a wider output range than the sigmoid function, which is 0 to 1. The Tanh activation function
is defined as (3) [15].
𝑒 𝑥 −𝑒 −𝑥
𝑓(𝑥) = (3)
𝑒 𝑥 +𝑒 −𝑥

Tanh improved upon the limitations of the sigmoid function, mainly by being zero-centered, which
helps prevent the vanishing gradient problem to some extent. However, it still suffers from the vanishing
gradient issue outside of the range [-1, 1], where the derivative becomes very close to zero, slowing down
training for deep networks. Despite being better than the sigmoid function in some cases, Tanh still has some
shortcomings that have led to the popularity of the rectified linear unit (ReLU) and its variants. The main
drawback of Tanh is that it can exhibit saturated behavior outside the range [-1, 1], leading to slow convergence
and gradients approaching zero in those regions. Figure 6 illustrates the Tanh function.

Figure 3. ANN structure Figure 4. Structure of neuron in ANN

Figure 5. Sigmoid function Figure 6. Tanh function

2.2.3. Rectified linear unit

The primary constraints of Sigmoid and Tanh activation functions are their tendency to produce
saturated outputs and contribute to heightened complexity. In contrast, the rectified linear unit (ReLU) has
emerged as the leading choice for activation functions, due to its simplicity and enhanced performance. It can
be written as (4) [16].

𝑥 if 𝑥 ≥ 0
𝑓(𝑥) = { (4)
0 otherwise

The waveform of the ReLu function is shown in Figure 7.

2.2.4. Softplus
The softplus function serves as a smooth approximation of the ReLU function, effectively eliminating
the corner point in the ReLU graph and substituting it with a gentle curve. Softplus presents an alternative to

Int J Pow Elec & Dri Syst, Vol. 16, No. 2, June 2025: 1174-1183
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1177

ReLu due to its differentiability and the ease with which its derivative can be demonstrated. The (5) shows the
softplus function formula [17].

𝑓(𝑥) = ln (1 + 𝑒 𝑥 ) (5)

Figure 8 demonstrates the waveform of the softplus function.

Figure 7. ReLU function Figure 8. Softplus function

3. MOTIVATION AND METHODOLOGY

From the review of ANN-based MPPT methods [18]−[26], most of the design mechanisms utilize the
sigmoid function as the activation function for ANN. However, none of the designs consider the use of the
softplus activation function or include hardware validation for ANN-based MPPT. Therefore, the motivation
of this paper is to design a softplus function trained ANN-based MPPT (SP-ANN MPPT) using by
microcontroller. The proposed method is then compared with a sigmoid function trained ANN-based MPPT
(SM-ANN MPPT). Figure 9 demonstrates the proposed block diagram of MPPT. The MPPT controller is
developed by practicing ANN. A multi-layer perceptron (MLP) is elected in this paper. The input layers are
the irradiance of each PV module and the ambient temperature. The hidden layer contains 150 neurons, which
receive data from the input layer and send it to the output layer. The output layer contains a neuron that indicates
the duty cycle of the boost converter.
The proposed method is evaluated through both simulation and experimental testing. For the
simulation evaluation, a Simulink model is developed, as illustrated in Figure 10. The electrical parameters of
the boost converter are set are L = 100 µH, C = 20 µF, and switching frequency (fsw) is 31 kHz. The first step
is to collect data for the training of the ANN. To gather a variety of data values, the model is tested under
various weather conditions, with irradiance ranging from 500 to 1000 W/m 2 and temperature ranging from
30 °C to 40 °C. The maximum power point (MPP) and the corresponding duty cycle for each weather condition
are recorded. The dataset is divided into two groups: training and validation data. The training data is used to
train the neural networks, while the validation data is utilized to evaluate the performance of the trained neural
networks. In this simulation, the dataset comprises 15 training samples and 5 validation samples. Both the SP-
ANN and SM-ANN models are trained using the backpropagation algorithm [27] implemented in MATLAB.
Next, an experimental evaluation was carried out. A main board comprising a boost converter and
modules for sensing voltage, current, temperature, and irradiance was designed and fabricated as a single PCB.
The electrical parameters of the boost converter are identical to those used in the simulation model shown in
Figure 10. An Arduino Mega 2560 was selected as the MPPT controller. Figure 11 illustrates the
microcontroller interfacing with the main board via header pins, functioning as an MPPT charge controller.
Figures 12(a) and 12(b) illustrate the configuration of a PV module and the overall experimental setup. The
photovoltaic modules are shaded with different patterns, and the corresponding photovoltaic modules’
irradiances, temperature, and the maximum power duty cycles are recorded to form the dataset for training and
validation. The data is sent to the laptop via serial communication. In this experimental evaluation, 100 data is
used for training and 30 data is used for validation. The backpropagation algorithm is employed for the training
process. After training, the optimized weights and biases were then implemented in an Arduino Mega 2560,
which serve as an MPPT controller.

Soft plus function trained artificial neural network based maximum power … (Liong Han Wen)
1178  ISSN: 2088-8694

Figure 9. Proposed ANN-based MPPT system

Figure 10. Simulink model for simulation validation

The proposed method is evaluated through both simulation and experimental testing. For the
simulation evaluation, a Simulink model is developed, as illustrated in Figure 10. The electrical parameters of
the boost converter are set are L = 100 µH, C = 20 µF, and switching frequency (fsw) is 31 kHz. The first step
is to collect data for the training of ANN. To gather variety of data values, the model is tested under various
weather conditions, with irradiance ranging from 500 to 1000 W/m 2 and temperature ranging from
30 °C to 40 °C. The maximum power point (MPP) and the corresponding duty cycle for each weather condition
are recorded. The dataset is divided into two groups: training and validation data. The training data is used to
train the neural networks, while the validation data is utilized to evaluate the performance of the trained neural
networks. In this simulation, the dataset comprises 15 training samples and 5 validation samples. Both the SP-
ANN and SM-ANN models are trained using the backpropagation algorithm [27] implemented in MATLAB.
Next, an experimental evaluation was carried out. A main board comprising a boost converter and
modules for sensing voltage, current, temperature, and irradiance was designed and fabricated as a single PCB.
The electrical parameters of the boost converter are identical to those used in the simulation model shown in
Figure 10. An Arduino Mega 2560 was selected as the MPPT controller. Figure 11 illustrates the
microcontroller interfacing with the main board via header pins, functioning as an MPPT charge controller.
Figures 12(a) and 12(b) illustrate the configuration of a PV module and the overall experimental setup. The
photovoltaic modules are shaded with different patterns, and the corresponding photovoltaic modules’
irradiances, temperature, and the maximum power duty cycles are recorded to form the dataset for training and
validation. The data is sent to the laptop via serial communication. In this experimental evaluation, 100 data is
used for training and 30 data is used for validation. The backpropagation algorithm is employed for the training
process. After training, the optimized weights and biases were then implemented in an Arduino Mega 2560,
which serves as an MPPT controller.

Int J Pow Elec & Dri Syst, Vol. 16, No. 2, June 2025: 1174-1183
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1179

(a) (b)

Figure 11. MPPT charge controller: (a) top view and (b) side view

(a) (b)

Figure 12. Experimental setup for (a) PV module and (b) overall system

4. RESULT AND DISCUSSION

4.1. Simulation results
Figures 13 and 14 illustrate the training and validation regression plots for SP-ANN and SM-ANN
based on the Simulink model shown in Figure 10. The training regression R-squared values for SP-ANN and
SM-ANN are 0.7922 and 0.5561, respectively, while the validation regression R-squared values are 0.9166
and 0.6887, respectively. The simulation results indicate that SP-ANN outperforms SM-ANN. This finding
provides motivation to further validate the proposed method experimentally.

4.2. Experimental results

Figure 15 depicts the mean squared error (MSE) plot for both softplus function trained ANN (SP-
ANN) and sigmoid function trained ANN (SM-ANN) for the experimental model as shown in Figure 12(b).
Both the training and validation curves are decreasing, showing that the neural networks are trained well.
The smoothness of the curve in Figure 15(a) is higher than that of Figure 15(b). This indicates that SP-ANN
can learn in a stable and consistent manner if compared with SM-ANN. The best performance for both SP-
ANN and SM-ANN trainings was achieved at epoch 100000 and 64310, and the MSE observed is 0.00317
and 0.063. Additionally, the difference between the training and validation curves of SP-ANN is smaller
than that of SM-ANN. This indicates that SP-ANN can predict the duty cycle more accurately than SM-
ANN for both seen and unseen situations.
The experimental comparison of SP-ANN and SM-ANN is further validated by using linear
regression, as shown in Figures 16 and 17. From the statistical results, it is evident that SM-ANN struggles to
accurately predict duty cycles below 0.4. SP-ANN shows an excellent R-squared in both training and validation
if compared with SM-ANN. This suggests that SP-ANN can better capture the variability of the overall data
than SM-ANN.
In the final step, the performance of SP-ANN and SM-ANN-based MPPT methods was evaluated
through field tests conducted outdoors on sunny days. The tests included three validation conditions: unshaded
(pattern 1), shaded at PV module 1 (pattern 2), and shaded at PV module 2 (pattern 3), as illustrated in
Figure 18. The MPP tracked by both methods were recorded from 7 AM to 7 PM, as shown in Figure 19. The

Soft plus function trained artificial neural network based maximum power … (Liong Han Wen)
1180  ISSN: 2088-8694

highest power tracked by SP-ANN for patterns 1, 2, and 3 was 4.35 W, 3.8 W, and 3.73 W, respectively. In
contrast, the highest power tracked by SM-ANN for the same patterns was 3.88W, 2.77 W, and 2.96 W,
respectively. The overall results indicate that the SP-ANN based MPPT is more efficient than the SM-ANN
based MPPT.

(a) (b)

Figure 13. Training regression plot of (a) SP-ANN and (b) SM-ANN for the simulation model

(a) (b)

Figure 14. Validation regression plot of (a) SP-ANN and (b) SM-ANN for the simulation model

(a) (b)

Figure 15. MSE plot of (a) SP-ANN training and (b) SM-ANN training

(a) (b)

Figure 16. Training regression plot of (a) SP-ANN and (b) SM-ANN for the hardware model

Int J Pow Elec & Dri Syst, Vol. 16, No. 2, June 2025: 1174-1183
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1181

(a) (b)

Figure 17. Validation regression plot of (a) SP-ANN and (b) SM-ANN for the hardware model

(a) (b) (c)

Figure 18. Test setup to simulate (a) pattern 1, (b) pattern 2, and (c) pattern 3 shading conditions

(a) (b)

(c)

Figure 19. SP-ANN and SM-ANN based MPPT for (a) pattern 1, (b) pattern 2, and (c) pattern 3

5. CONCLUSION
In this paper, a soft plus function trained ANN-based (SP-ANN) MPPT is proposed. The proposed
method is compared with sigmoid function trained ANN-based (SM-ANN) MPPT by using simulation and
experimental validation. The proposed photovoltaic system consists of two series connected PV modules, a
MPPT controller, boost DC/DC converter, and a resistive load. Arduino Mega 2560 is used as MPPT controller.
Based on the performance results, it has been identified that SP-ANN MPPT has better mean square error
(MSE) and linear regression if compared with SM-ANN MPPT. Besides, SP-ANN MPPT demonstrates better
performance than SM-ANN MPPT in field test. The future work will focus on validate the proposed method
with different kinds of load.

Soft plus function trained artificial neural network based maximum power … (Liong Han Wen)
1182  ISSN: 2088-8694

ACKNOWLEDGEMENTS
This work was executed with the help of financial and resources available at Electrical, Electronics
and Automation Section, Malaysian Spanish Institute, Universiti Kuala Lumpur.

FUNDING INFORMATION
Authors state no funding involved.

AUTHOR CONTRIBUTIONS STATEMENT

This journal uses the Contributor Roles Taxonomy (CRediT) to recognize individual author
contributions, reduce authorship disputes, and facilitate collaboration.

Name of Author C M So Va Fo I R D O E Vi Su P Fu
Liong Han Wen ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Mohd Rezal Mohamed ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

C : Conceptualization I : Investigation Vi : Visualization

M : Methodology R : Resources Su : Supervision
So : Software D : Data Curation P : Project administration
Va : Validation O : Writing - Original Draft Fu : Funding acquisition
Fo : Formal analysis E : Writing - Review & Editing

CONFLICT OF INTEREST STATEMENT

Authors state no conflict of interest.

DATA AVAILABILITY
The data generated and analyzed during the current study are not publicly available due to institutional
restrictions but are available from the corresponding author upon reasonable request.

REFERENCES
[1] M. M. A. Awan, M. Y. Javed, A. B. Asghar, and K. Ejsmont, “Performance optimization of a ten check MPPT algorithm for an off-
grid solar photovoltaic system,” Energies, vol. 15, no. 6, 2022, doi: 10.3390/en15062104.
[2] M. Sarvi and A. Azadian, “A comprehensive review and classified comparison of MPPT algorithms in PV systems,” Energy
Systems, vol. 13, no. 2, pp. 281–320, 2022, doi: 10.1007/s12667-021-00427-x.
[3] G. A. Raiker, U. Loganathan, and S. B. Reddy, “Current control of boost converter for PV interface with momentum-based perturb
and observe MPPT,” IEEE Transactions on Industry Applications, vol. 57, no. 4, pp. 4071–4079, 2021, doi:
10.1109/TIA.2021.3081519.
[4] C. Gonzalez-Castano, C. Restrepo, S. Kouro, and J. Rodriguez, “MPPT algorithm based on artificial bee colony for PV System,”
IEEE Access, vol. 9, pp. 43121–43133, 2021, doi: 10.1109/ACCESS.2021.3066281.
[5] M. N. I. Jamaludin et al., “An effective salp swarm based MPPT for photovoltaic systems under dynamic and partial shading
conditions,” IEEE Access, vol. 9, pp. 34570–34589, 2021, doi: 10.1109/ACCESS.2021.3060431.
[6] M. F. Jalil, S. Khatoon, I. Nasiruddin, and R. C. Bansal, “Review of PV array modelling, configuration and MPPT techniques,”
International Journal of Modelling and Simulation, vol. 42, no. 4, pp. 533–550, 2022, doi: 10.1080/02286203.2021.1938810.
[7] H. Abidi, L. Sidhom, and I. Chihi, “Systematic literature review and benchmarking for photovoltaic MPPT Techniques,” Energies,
vol. 16, no. 8, 2023, doi: 10.3390/en16083509.
[8] E. S. Wirateruna and A. F. A. Millenia, “Design of MPPT PV using Particle swarm optimization algorithm under partial shading
condition,” International Journal of Artificial Intelligence & Robotics (IJAIR), vol. 4, no. 1, pp. 24–30, 2022, doi:
10.25139/ijair.v4i1.4327.
[9] M. Fathi and J. A. Parian, “Intelligent MPPT for photovoltaic panels using a novel fuzzy logic and artificial neural networks based
on evolutionary algorithms,” Energy Reports, vol. 7, pp. 1338–1348, 2021, doi: 10.1016/j.egyr.2021.02.051.
[10] M. F. El-Khatib and E. A. Aner, “Efficient MPPT control for a photovoltaic system using artificial neural networks,” ERU Research
Journal, vol. 2, no. 3, pp. 385–398, May 2023, doi: 10.21608/erurj.2023.301364.
[11] M. T. Hussain, A. Sarwar, M. Tariq, S. Urooj, A. BaQais, and M. A. Hossain, “An evaluation of ANN algorithm performance for
MPPT energy harvesting in solar PV Systems,” Sustainability, vol. 15, no. 14, p. 11144, Jul. 2023, doi: 10.3390/su151411144.
[12] O. I. Abiodun et al., “Comprehensive review of artificial neural network applications to pattern recognition,” IEEE Access, vol. 7,
pp. 158820–158846, 2019, doi: 10.1109/ACCESS.2019.2945545.
[13] T. Szandała, “Review and comparison of commonly used activation functions for deep neural networks,” Studies in Computational
Intelligence, vol. 903, pp. 203–224, 2021, doi: 10.1007/978-981-15-5495-7_11.
[14] Y. Koçak and G. Üstündağ Şiray, “New activation functions for single layer feedforward neural network,” Expert Systems with
Applications, vol. 164, p. 113977, Feb. 2021, doi: 10.1016/j.eswa.2020.113977.
[15] K. Et-Torabi and A. Mesbahi, “MPPT based artificial neural network versus pertub & observe for photovoltaic energy conversion
system,” E3S Web of Conferences, vol. 336, 2022, doi: 10.1051/e3sconf/202233600044.

Int J Pow Elec & Dri Syst, Vol. 16, No. 2, June 2025: 1174-1183
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1183

[16] I. Jahan, M. F. Ahmed, M. O. Ali, and Y. M. Jang, “Self-gated rectified linear unit for performance improvement of deep neural
networks,” ICT Express, vol. 9, no. 3, pp. 320–325, 2023, doi: 10.1016/j.icte.2021.12.012.
[17] X. Liao, T. Zhou, L. Zhang, X. Hu, and Y. Peng, “A method for calculating the derivative of activation functions based on piecewise
linear approximation,” Electronics (Switzerland), vol. 12, no. 2, 2023, doi: 10.3390/electronics12020267.
[18] F. Saeed, M. H. Yousuf, H. A. Tauqeer, M. R. Akhtar, Z. A. Abbas, and M. H. Khan, “Performance benchmark of multi-layer neural
network based solar MPPT for PV applications,” in 2021 International Conference on Emerging Power Technologies, ICEPT 2021,
2021. doi: 10.1109/ICEPT51706.2021.9435583.
[19] A. G. Abdullah, M. S. Aziz, and B. A. Hamad, “Comparison between neural network and P&O method in optimizing MPPT control
for photovoltaic cell,” International Journal of Electrical and Computer Engineering, vol. 10, no. 5, pp. 5083–5092, 2020, doi:
10.11591/IJECE.V10I5.PP5083-5092.
[20] M. N. Ali, K. Mahmoud, M. Lehtonen, and M. M. F. Darwish, “Promising mppt methods combining metaheuristic, fuzzy-logic and
ann techniques for grid-connected photovoltaic,” Sensors (Switzerland), vol. 21, no. 4, pp. 1–18, 2021, doi: 10.3390/s21041244.
[21] S. A. Rizzo, N. Salerno, G. Scelba, and A. Sciacca, “Enhanced hybrid global MPPT algorithm for PV systems operating under fast-
changing partial shading conditions,” International Journal of Renewable Energy Research, vol. 8, no. 1, pp. 221–229, 2018, doi:
10.20508/ijrer.v8i1.6704.g7295.
[22] Y. E. A. Idrissi, K. Assalaou, L. Elmahni, and E. Aitiaz, “New improved MPPT based on artificial neural network and PI controller
for photovoltaic applications,” International Journal of Power Electronics and Drive Systems, vol. 13, no. 3, pp. 1791–1801, 2022,
doi: 10.11591/ijpeds.v13.i3.pp1791-1801.
[23] S. R. Kiran, C. H. H. Basha, V. P. Singh, C. Dhanamjayulu, B. R. Prusty, and B. Khan, “Reduced Simulative performance analysis
of variable step size ANN based MPPT techniques for partially shaded solar PV Systems,” IEEE Access, vol. 10, pp. 48875–48889,
2022, doi: 10.1109/ACCESS.2022.3172322.
[24] R. Divyasharon and R. Narmatha Banu, “Performance assessment of Boost, Cuk and Zeta converter with neural network MPPT in
PV systems for varying weather patterns,” International Journal of Engineering Research and Applications, vol. 12, no. 4, pp. 9–
20, 2022.
[25] K. Krishnaram, T. S. Padmanabhan, F. Alsaif, and S. Senthilkumar, “Performance optimization of interleaved boost converter with
ANN supported adaptable stepped-scaled P&O based MPPT for solar powered applications,” Scientific Reports, vol. 14, no. 1,
2024, doi: 10.1038/s41598-024-58852-8.
[26] M. Yılmaz and M. Corapsiz, “PSO training neural network MPPT with CUK converter topology for stand-alone PV systems under
varying load and climatic conditions,” Türk Doğa ve Fen Dergisi, vol. 13, no. 1, pp. 88–97, 2024, doi: 10.46810/tdfd.1423852.
[27] A. Singh, S. Kushwaha, M. Alarfaj, and M. Singh, “Comprehensive overview of backpropagation algorithm for digital image
denoising,” Electronics (Switzerland), vol. 11, no. 10, 2022, doi: 10.3390/electronics11101590.

BIOGRAPHIES OF AUTHORS

Liong Han Wen received a B.Eng. (Hons) degree in Electronics and Control
System Engineering from Liverpool John Moores University in 2005 and an M.Sc. degree in
Electrical and Electronic from the University of Bradford in 2013. He is currently working
toward a Ph.D. degree in Electrical and Electronic Engineering at Universiti Kuala Lumpur. He
can be contacted at email: [email protected].

Mohd Rezal Mohamed is a senior lecturer in Electrical, Electronics and

Automation Section at the Malaysian Spanish Institute, Universiti Kuala Lumpur, Kulim,
Kedah, Malaysia. He received his B.Eng., M.Sc., and Ph.D. degrees in Electrical and Electronics
Engineering from Universiti Putra Malaysia and Universiti Sains Malaysia in 2002, 2008, and
2019, respectively. His research interests include the field of electric machines, optimization,
electromagnetic applications, electric vehicles, power electronics, and motor drives. He can be
contacted at email: [email protected].

Soft plus function trained artificial neural network based maximum power … (Liong Han Wen)