International Journal of Applied Power Engineering (IJAPE)

Vol. 14, No. 2, June 2025, pp. 308~318

ISSN: 2252-8792, DOI: 10.11591/ijape.v14.i2.pp308-318  308

Optimization and dimensioning of stand-alone systems:

enhancing MPPT efficiency through DLGA integration

Moufida Saadi, Dib Djalel, Kadir Erkan

Electrical and Engineering Laboratory (LABGET), Faculty of Electrical Engineering, University of Tebessa, Tebessa, Algeria

Article Info ABSTRACT

Article history: This paper explores optimizing and sizing stand-alone solar power systems
using an intelligent maximum power point tracking (MPPT) method,
Received May 5, 2023 enhanced by artificial neural networks (ANN). The study focuses on both
Revised Nov 4, 2024 system sizing and energy optimization, integrating genetic algorithms (GA)
Accepted Nov 28, 2024 with deep learning (DL) to optimize the architecture of the ANN for
improved performance in predicting solar energy output. The hybrid method,
deep learning genetic algorithms (DLGA), efficiently reduces computational
Keywords: complexity and enhances flexibility through parameter tuning, significantly
improving the performance of multi-layer perceptron networks.
Artificial neural network Additionally, a precise sizing methodology based on solar irradiance data
Battery storage was implemented to ensure the system is neither oversized nor undersized.
Deep learning The system's performance was tested and validated using
Genetic algorithms MATLAB/Simulink simulations, which demonstrated superior predictive
Maximum power point tracking accuracy, faster convergence, and optimized energy capture. This combined
Photovoltaic approach of intelligent MPPT and accurate sizing presents a highly effective
Stand-alone system solution for improving the efficiency and reliability of stand-alone solar
energy systems under varying environmental conditions.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Moufida Saadi
Electrical and Engineering Laboratory (LABGET), Faculty of Electrical Engineering
University of Tebessa
Tebessa 12000, Algeria
Email: [email protected]

1. INTRODUCTION
The shift to sustainable energy solutions increasingly highlights stand-alone photovoltaic (PV)
systems as promising alternatives to traditional power sources. These systems generate electricity from
sunlight, an inexhaustible source that emits no greenhouse gases, making them crucial to renewable energy
transitions, especially in remote areas outside conventional grid reach. For self-sufficiency, efficient battery
storage and accurate sizing of components, like solar panels and batteries, are vital for continuous, cost-
effective power supply [1]–[3].
Sizing optimization determines the best PV configuration to meet energy needs without waste [4].
Various methods are used for this purpose, each with specific advantages and limitations [5]. One common
approach, the 'monthly average solar radiation' method, leverages historical solar data to balance energy
generation and storage effectively in regions with stable weather [6]. Yet, it may be less accurate in areas
with high solar variability [7]. The 'peak sun hours' method simplifies sizing by using peak sunlight hours,
but its simplicity can reduce accuracy [8]. More advanced methods, like 'hybrid simulation-optimization,'
combine simulation with optimization algorithms to adapt to specific conditions, though they require high
computational resources [9], [10]. AI-based approaches, including machine learning and neural networks, are

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

Int J Appl Power Eng ISSN: 2252-8792  309

emerging for PV sizing, yielding accurate predictions when quality data is available [11], [12]. Optimization
strategies further enhance PV system performance, including strategic panel placement, effective battery
management, and maximum power point tracking (MPPT) for optimal energy conversion under changing
environmental conditions [13]–[15]. MPPT significantly boosts energy yields in regions with variable
weather, while advanced techniques like artificial neural networks (ANN) and deep learning genetic
algorithms (DLGA) refine optimization, improving energy management accuracy and adaptability in diverse
environments [16]–[19].
The structure of the paper is methodically organized to facilitate understanding the process of
precise sizing and optimization of stand-alone PV systems. Section 2 discusses the modeling and sizing
methodologies for these systems. Section 3 reviews recent advancements in intelligent MPPT techniques.
Section 4 focuses on the application of ANN and DLGA in optimizing MPPT, while section 5 presents and
analyzes the research findings. Finally, section 6 summarizes the study's key insights and conclusions,
highlighting the potential for future research and development.

2. MODELING AND SIZING STAND-ALONE SYSTEM

The components of a typical isolated system powered by solar energy, supplemented with battery
storage, are modeled mathematically. This system is represented as a stand-alone configuration in Figure 1.
To achieve energy self-sufficiency, the final system configuration consists of the following elements:
- A 1.2 kW solar power unit, comprising 16 PV panels, connected to a DC-DC converter and interfaced
with the direct current (DC) bus.
- Two batteries, each with a capacity of 100 Ah and a voltage of 12 V, integrated into the system via
a bidirectional DC-DC converter. Both batteries share the same connection point and are connected to
the DC bus through both alternating current (AC)/DC and DC-DC converters.

Figure 1. System components and description

In this section, we delve into a detailed exploration of a power system, focusing on the intricacies of
modeling its various components. The equilibrium of power within the DC bus can be formulated as (1).

𝑃𝐿 (𝑡) = 𝜂𝐷𝐴 (𝜂𝐷𝐴 𝑃𝑃𝑉 (𝑡) ± 𝜂𝐷𝐷 𝑃𝑏 (𝑡)) (1)

In this equation, 𝑃𝑃𝑉 (𝑡) and 𝑃𝑏 (𝑡) represent the power outputs from the PV array and the battery bank,
respectively. The constants 𝜂𝐷𝐷 , 𝜂𝐷𝐴 denote the efficiencies of the DC/DC and DC/AC power converters. For
the purpose of this analysis, these efficiencies are assumed to be constant, with 𝜂𝐷𝐷 = 0.95 and 𝜂𝐷𝐴 = 0.9.
The sign convention for Pb(t) designates it as negative when the battery is charging and positive when
discharging. However, it is essential to note that power balance is constrained by certain physical and
operational limitations.
𝑎𝑣
0 ≤ 𝑃𝑃𝑉 (𝑡) ≤ 𝑃𝑃𝑉 (t)

𝑃𝑏𝑚𝑖𝑛 ≤ 𝑃𝑏 (𝑡) ≤ 𝑃𝑏𝑚𝑎𝑥 (2)

𝑎𝑣
Where 𝑃𝑃𝑉 represents the available power generation from the PV array 𝑃𝑏𝑚𝑖𝑛 and 𝑃𝑏𝑚𝑎𝑥 refer to the
minimum and maximum battery bank power, respectively [20].

Optimization and dimensioning of stand-alone systems: enhancing MPPT efficiency … (Moufida Saadi)
310  ISSN: 2252-8792

2.1. Photovoltaic array

The power transmission to the generator shaft in a PV system refers to the conversion of incident
solar radiation into electrical power. This conversion process is accomplished through the operation of the
PV panels [21]. The power transmitted to the generator shaft, represented by (3), is a function of the available
solar radiation, 𝑛𝑝𝑣 the efficiency of the PV panels, 𝐴𝑝𝑣 surface area of the PV panels, 𝐵𝑝𝑣 the temperature
coefficient for the PV panels. This equation quantifies the power output of the PV system, providing valuable
insights into its capacity to generate electrical energy from sunlight.

𝑃𝑝𝑣 (𝑡) = 𝐴𝑝𝑣 . 𝐺(𝑡). 𝑛𝑝𝑣 . (1 + 𝐵𝑝𝑣 . (𝑇(𝑡) − 𝑇𝑟𝑒𝑓 )) (3)

Where 𝑇𝑟𝑒𝑓 reference temperature is temperature as a function of time and 𝐺 is the solar radiation as a
function of time. The average power of 𝑃𝑝𝑣 over the specified time period 𝜏 can be calculated using (4).

𝑎𝑣 1 𝜏
𝑃𝑝𝑣 = ∫0 𝐴𝑝𝑣 . 𝐺(𝑡). 𝜂𝑝𝑣 . (1 + 𝛽𝑝𝑣 . (𝑇(𝑡) − 𝑇𝑟𝑒𝑓 )) . 𝑑𝑡 (4)
𝜏

2.2. Storage of energy

Lead-acid batteries used in PV-wind systems function under defined conditions. In the typical
operational state, it is difficult to anticipate whether energy will be drawn from or supplied to the battery [22].
Each battery within the energy storage system is depicted as an equivalent circuit, comprising a voltage
source (representing open circuit voltage, 𝑉𝑜𝑐 ) in series with an internal resistance (R_int) [23]. As a result,
the terminal voltage of the battery is established by (5).

𝑉𝑏𝑎𝑡𝑡 = 𝑉𝑜𝑐 − 𝑅𝑖𝑛𝑡 𝐼𝑏𝑎𝑡 (5)

In this model, both 𝑉𝑜𝑐 and 𝑅𝑖𝑛𝑡 are dependent on the battery's state of charge (𝑆𝑂𝐶𝑏 ), which
indicates the remaining capacity available for discharge. This correlation is represented as data vectors, with
their values determined through interpolation within the respective vector based on the current 𝑆𝑂𝐶𝑏 . This
accommodates the nonlinear interdependencies between 𝑉𝑜𝑐 and 𝑅𝑖𝑛𝑡 . The state of charge 𝑆𝑂𝐶𝑏 can be
expressed as (6).
∗,𝑚𝑎𝑥 ∗,𝑢
𝐶𝑏 −𝐶𝑏
𝑆𝑂𝐶𝑏 = 100 [%] (6)
𝐶𝑏∗,𝑚𝑎𝑥

Where 𝐶𝑏∗,𝑢 represents the number of ampere-hours already utilized and 𝐶𝑏∗,𝑚𝑎𝑥 signifies the maximum
capacity, measured in ampere-hours. This can be computed as (7).
𝑡 𝐼 𝑏 𝜂𝑐
𝐶𝑏∗,𝑢 = ∫0 𝑑𝑡 [𝐴ℎ] (7)
3600

Where 𝜂𝑐 denotes the charge/discharge battery Coulombic efficiency, which is 0.975 in this case. 𝐼𝑏 signifies
the battery current in amperes, with 𝐼𝑏 > 0 indicating discharge and 𝐼𝑏 < 0 indicating charging. The initial
𝑆𝑂𝐶𝑏 is determined by a nonzero initial value of 𝐶𝑏∗,𝑢 . To ensure optimal performance and battery longevity,
𝑆𝑂𝐶𝑏 must be maintained within specific limits, defined as 𝑆𝑂𝐶𝑏𝑚𝑖𝑛 ≤ 𝑆𝑂𝐶𝑏 ≤ 𝑆𝑂𝐶𝑏𝑚𝑎𝑥 .
The battery current is subject to constraints, and these limits are contingent on 𝑉𝑜𝑐 and 𝑅𝑖𝑛𝑡 , as
described by (8).

(𝑉𝑜𝑐 −𝑉𝑏𝑚𝑎𝑥 )
𝑑𝑢𝑟𝑖𝑛𝑔 𝑜𝑓 𝑐ℎ𝑎𝑟𝑔𝑒
𝑅𝑖𝑛𝑡
𝐼𝑏𝑙𝑖𝑚 = { (8)
(𝑉𝑜𝑐 −𝑉𝑏𝑚𝑖𝑛)
𝑑𝑢𝑟𝑖𝑛𝑔 𝑜𝑓 𝑑𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑒
𝑅𝑖𝑛𝑡

𝑉𝑏𝑚𝑖𝑛 and 𝑉𝑏𝑚𝑎𝑥 represent the minimum and maximum permissible battery bank voltages, respectively.
Furthermore, 𝐼𝑏𝑙𝑖𝑚 is indirectly influenced by 𝑆𝑂𝐶𝑏 through the previously mentioned nonlinear relationships.
Additionally, a mechanism is in place to limit the battery bank current, ensuring zero current when 𝑆𝑂𝐶𝑏
reaches its maximum or minimum value [24].

Int J Appl Power Eng, Vol. 14, No. 2, June 2025: 308-318
Int J Appl Power Eng ISSN: 2252-8792  311

3. SIZING STAND-ALONE SYSTEM BASED ON MONTHLY AVERAGE METHOD

The effective and dependable functioning of a stand-alone photovoltaic/battery system relies heavily
on accurate sizing. Sizing this system utilizing monthly average data entails establishing the suitable
capacities for the PV panels and energy storage elements, which are crucial in proficiently fulfilling the load
demands (refer to Table 1). The general load, PV, and energy produced are given by (9) and (10).

𝐸𝑙𝑜𝑎𝑑 = 𝐸𝑃𝑉 𝑆𝑃𝑉 (9)

𝐸𝑃𝑉 = 𝑆𝑃𝑉 𝐸𝑖𝑟𝑟 𝜂𝑃𝑉 (10)

With:

𝜂𝑃𝑉 = 𝜂𝑃𝑉−𝑆𝑇𝐶 [1 − 𝛽𝑜𝑐 (𝑇𝑗 − 𝑇𝑗−𝑆𝑇𝐶 )] (11)

Through rigorous computations of monthly energy yield for each generator and corresponding load
demand, distinct surface areas for photovoltaic panels are discerned. These quantifications are deduced using
the formulations presented for PV, as elucidated in [25].

𝐸𝐿𝑜𝑎𝑑,𝑚
𝑆𝑃𝑉 = max ( ) (12)
𝐸𝑃𝑉,𝑚

The Montney energies produced by PV are given in (13).

𝐸𝑃𝑉,𝑚 = (∑12
𝑚=1 𝐸𝑃𝑉 )/12
{ (13)
𝐸𝐿𝑜𝑎𝑑,𝑚 = (∑12
𝑚=1 𝐸𝐿𝑜𝑎𝑑 )/12

Here, Elmean represents the energy required to meet the load demand. It is calculated as the average energy
needed to satisfy the system's load demand under various configurations of wind turbines and photovoltaic
panels. 𝐾𝑝𝑒𝑟𝑐 represents the proportion of the load supplied by the PV source. Consequently, we derive the
following result, as expressed in (14).

𝐸𝐿𝑜𝑎𝑑,𝑎𝑣𝑒𝑟
𝑆𝑃𝑉 = 𝐾𝑝𝑒𝑟𝑐 ( ) (14)
𝐸𝑃𝑉,𝑎𝑣𝑒

The subsequent equations establish the quantities of PV panels required, as expressed in (15).

𝑆𝑃𝑉,𝑓𝑖𝑛𝑎𝑙 = 𝑁𝑃𝑉 𝑆𝑃𝑉,𝑢𝑛𝑖𝑡 (15)

The mean energy consumption is expressed as (16).

𝐸𝑙𝑎𝑜𝑑−𝑎𝑣𝑒 = 𝐸𝑃𝑉,𝑎𝑣𝑒 𝑆𝑃𝑉,𝑢𝑛𝑖𝑡 (16)

Table 1. The setup and parameters of the PV and wind energy systems
Month 𝐸𝑖𝑟𝑟 (KWh/m) T (°C) 𝑁𝑃𝑉 𝐸𝑃𝑉 (KWh/m2) 𝐸𝑙𝑜𝑎𝑑 (KWh)
January 85.5 10.1 0.1410 7.78 333.6
February 98.6 11.5 0.1419 9.03 339
March 143.6 16.1 0.1446 13.41 347
April 174.2 19.8 0.1468 16.51 347.04
May 201.5 24.5 0.1497 19.48 336.72
June 207.3 28.7 0.1486 19.89 332.88
July 218.2 32.3 0.1539 21.74 347.04
August 197.1 31.6 0.1512 19.59 345.84
September 156.4 27.1 0.1512 15.27 336.48
October 127.9 22.7 0.1485 12.26 323.52
November 95 15.5 0.1443 8.85 343.44
December 79.6 11.1 0.1416 7.28 329.52
𝐸𝑃𝑉 , ave = 13.42
𝐸𝑙𝑜𝑎𝑑 , ave = 338.5

Optimization and dimensioning of stand-alone systems: enhancing MPPT efficiency … (Moufida Saadi)
312  ISSN: 2252-8792

The sizing parameters for the hybrid system are determined based on the previously outlined
relationships. Table 2 provides a breakdown of the monthly energy production from the solar system. It can
be noted that the average photovoltaic energy output is approximately 13.42 kWh/m 2. Given that the average
load energy demand is 338.5 kWh, and considering that the system in question is a stand-alone PV system,
only the configuration of 40 panels comes closest to meeting the required load energy of 523.47 kWh.
Battery capacity is calculated using the annual monthly average method with the day of autonomy,
as expressed in (17).
𝑑𝑎𝑢𝑡 .𝐸𝑙𝑜𝑎𝑑 ,𝑚
𝐶𝑏𝑎𝑡𝑡 = (17)
𝑈𝑏𝑎𝑡𝑡 .𝑃𝐷𝑃.𝜂𝑏𝑎𝑡𝑡 .𝑁𝑚

Where 𝐸𝑙𝑜𝑎𝑑 ,𝑚 monthly load consumed (kWh/d) and 𝑁𝑚 the number of days of the month that presents the
maximum load (31 days), PDP stands for percentage depth of discharge 𝜂𝑏𝑎𝑡𝑡 . The efficiency of the battery.
The number of batteries used is calculated by (18).
𝐶𝑏𝑎𝑡𝑡
𝑁𝑏𝑎𝑡𝑡 = 𝐸𝑁𝑇[ ] (18)
𝐶𝑏𝑎𝑡𝑡−𝑢

Where 𝐶𝑏𝑎𝑡𝑡−𝑢 represents the selected battery capacity. To summarize, the total maximum power output of
the photovoltaic panels is determined as Ppv = 40 × 80 = 3,600 kW. Moreover, the system utilizes 3 batteries
with specifications of (12 V, 100 Ah).

Table 2. The number of wind turbines and panels was determined through
𝐾𝑝𝑒𝑟𝑐 𝑆𝑃𝑉 (m2) 𝑁𝑃𝑉 𝑆𝑃𝑉 , final (m2) Elmean (Kwh)
0 0 0 0 0
0.1 4.28 7 4.522 60.68
0.2 7.5 12 7.752 104.03
0.3 7.76 12 7.752 104.03
0.4 8.40 13 8.39 112.59
0.5 8.64 13 8.398 112.70
0.6 10.04 16 10.336 198.70
0.7 11.17 17 10.98 147.35
0.8 14.12 23 14.858 199.28
0.9 19.83 31 20.02 268.66
1 26.38 40 26.48 345.36

4. APPLICATION OF HYBRID INTELLIGENT MPPT (DLGA)

The application of ANN in maximum power point (MPP) Tracking is particularly essential due to
solar energy's inherently variable nature, which is affected by a range of environmental conditions, including
the intensity of sunlight, temperature, and shadow impacts. ANN functions similarly to the human brain by
learning and retaining information and insights through a network of interconnected links known as weights.
For precise identification of the MPP, these weights associated with the neurons must be meticulously
calculated via an extensive training process. Once this training is complete, the ANN can serve as an
estimator for the MPP, providing the reference value (maximum power voltage (VMP) or maximum power
current (IMP)) to the MPPT controller [26].
The training of an ANN involves a systematic adjustment of weights and biases, often utilizing the
sigmoid activation function. Initially, weights and biases are randomly assigned to set the starting point for
the learning process. During forward propagation, inputs pass through the network, with each neuron
calculating a weighted sum and adding a bias, subsequently passed through an activation function like a
sigmoid. The sigmoid function, mapping values between 0 and 1, is favored for its ability to convert numbers
into probabilities and handle non-linear data relationships. Following this, the backpropagation phase begins,
where the network's output error is calculated and propagated backward, adjusting weights and biases. This
adjustment is based on the error's partial derivatives concerning each weight and bias, guided by a learning
rate parameter. This cycle of forward propagation, backpropagation, and weight and bias adjustments repeats
over multiple iterations, gradually refining the network to minimize prediction errors. The training process
also includes evaluating and adjusting the model with a validation set to prevent overfitting or underfitting,
ensuring the ANN effectively generalizes to new data. In ANNs, an increase in the number of hidden layers
can lead to enhanced tracking efficiency and improved performance in adapting to power fluctuations in the
array, though it may also result in slower tracking speed

Int J Appl Power Eng, Vol. 14, No. 2, June 2025: 308-318
Int J Appl Power Eng ISSN: 2252-8792  313

The integration GA with DL for the optimization of ANN architectures has been a focus of various
researchers. This approach aims to enhance the performance of multi-layer perceptron networks. Given the
computational complexity and extended training duration inherent in DL evolutionary algorithms like GA are
employed to optimize network performance. GA is particularly noted for its robust optimization capabilities.
This method effectively reduces computational complexity and increases overall system flexibility through
parameter tuning, thereby augmenting the performance of DL. In this scheme, DL is utilized to determine the
optimal duty cycle value, ensuring maximum power extraction. The neural network undergoes training with a
dataset, which is then optimized using GA for improved efficiency. The steps involved in implementing the
genetic algorithm are outlined as follows:
- Step I: Assess the fitness function and pinpoint the design parameters.
- Step II: Generate a population, representing potential solutions to the problem.
- Step III: Evaluate this population using an objective function.
- Step IV: From the population, select two parents based on their fitness levels. Higher fitness increases the
likelihood of selection.
- Step V: Create a new population by repeatedly executing selection, crossover, and mutation until the new
population is complete.
- Step VI: Form a new generation and return to step III.
- Step VII: If the end condition (minimization of mean squared error (MSE)) is met, conclude the process
and identify the best solution as the target (see Figure 2).

Figure 2. Block diagram training MPPT using DLGA

Figure 3 illustrates the training dynamics of different ANN architectures: ANN with 10 neurons,
ANN with 100 neurons, deep learning (DL), and DLGA, represented as Figures 3(a)-3(d), respectively.
Among these, DLGA (Figure 3(d)) shows the best performance, with rapid convergence and low mean
squared error (MSE) across training, validation, and test phases, indicating a highly generalizable model.
Figure 3(a) shows initial improvement but reaches a plateau, while Figure 3(b) exhibits overfitting, as seen in
the rise of validation error after initial progress. Figure 3(c), like Figure 3(a), fits the data decently but shows
Optimization and dimensioning of stand-alone systems: enhancing MPPT efficiency … (Moufida Saadi)
314  ISSN: 2252-8792

a slight divergence between training and validation errors, suggesting possible overfitting. Overall, DLGA
proves to be the most robust, making it the optimal choice for real-world applications due to its superior
accuracy and generalization. Table 3 provides a detailed comparison of the architectures based on key
metrics like epoch range (0 to 1000), training time, overall performance, and gradient behavior. These
metrics offer insights into the efficiency and effectiveness of each model, with the gradient target set at
1e-16, reflecting a high precision in the learning process.

Table 3. Comparison between performances of different architecture of ANN

ANN architecture Number of epochs Elapsed time (s) Performance Gradient
ANN 10N 1000 00 :00 :04 1.44 e-14 9.45 e-14
ANN 100 N 1000 00 :00 :08 1.33 e-14 7.32 e-11
DL 1000 00 :00 :15 4.09 e-12 9.68 e-9
DL GA 14 00:14:37 3.47e-32 5.16 e-17

(a)

(b)

(c)

(d)

Figure 3. The dynamic training of various ANN architectures: (a) ANN with 10 neurons,
(b) ANN with 100 neurons, (c) deep learning (DL), and (d) DLGA

Int J Appl Power Eng, Vol. 14, No. 2, June 2025: 308-318
Int J Appl Power Eng ISSN: 2252-8792  315

5. DISCUSSION OF RESULTS
Set within the specific environmental conditions of Negrine, Wilaya of Tebessa in Algeria, the study
utilizes historical atmospheric data from 2012, including variables like ambient temperature and solar
insolation, to accurately size and optimize the system as shown in Figure 4. The system, comprising PV
panels and battery storage, was simulated in MATLAB/Simulink using localized data and load profiles,
showcasing the effectiveness of ANN-optimized MPPT in improving power generation to meet varying
energy demands. The study provides valuable insights into the deployment of efficient solar energy systems
in arid and semi-arid regions.
Figure 5 presents the power generated by the PV and wind turbine is depicted alongside the load
profile. This figure helps visualize how the combined energy production from these renewable sources aligns
with the demand requirements. By comparing these curves, one can assess whether the generated power
meets, exceeds, or falls short of the load at various points in time.

Figure 4. Historical data ambient temperature and solar insolation in one year

Figure 5. Power load profile chosen

Figure 6 presents a comparative analysis of two MPPT methods: DLGA and perturb and observe
(P&O), applied to a stand-alone PV system, focusing on DC bus voltage. Over 12 hours, the DLGA
consistently maintains a higher and more stable voltage than P&O. While P&O shows a step-like increase
during its initial ramp-up, indicating its iterative approach, DLGA demonstrates a smoother and quicker
convergence to the maximum power point. This is likely due to DLGA's predictive capabilities, which use
historical data for more precise control. The zoomed-in view reveals that DLGA has minimal ripple and
tighter voltage regulation, suggesting better handling of variable environmental conditions, while P&O shows
more pronounced voltage fluctuations, indicating less stability. DLGA's stability reduces power oscillations,
enhancing system efficiency and minimizing wear on components.
Figure 7 compares the performance of four MPPT techniques: DLGA, DL, ANN, and P&O, over 12
hours in a PV system. DLGA, ANN, and DL demonstrate a swift and stable rise to peak power, with DLGA
showing superior stability and minimal fluctuations. As solar irradiance changes, DLGA adapts well,
maintaining near-optimal power around 880 W, while P&O experiences a larger dip to 780 W. During peak
midday irradiance, DLGA sustains around 1550 W, outperforming P&O, which fluctuates near 1500 W.
ANN and DL match DLGA at 1350 W but show a less dynamic response to irradiance changes. As sunlight
wanes, DLGA maintains the highest output (850 W), while P&O declines more erratically, and ANN/DL
drop more sharply.
Optimization and dimensioning of stand-alone systems: enhancing MPPT efficiency … (Moufida Saadi)
316  ISSN: 2252-8792

Figure 8, depicting battery power output, shows that DLGA stabilizes quickly, maintaining
consistent power with minimal fluctuation, indicating efficient battery management. In contrast, P&O
exhibits more pronounced fluctuations, suggesting less efficient battery charge regulation. As the system
transitions to discharging, DLGA handles the shift smoothly, while ANN and DL mirror each other closely in
performance. Overall, DLGA stands out for its robustness and adaptability, ensuring maximum battery
efficiency and system energy availability throughout the day.

Figure 6. Profile of voltage DC bus in 12 hours

Figure 7. Profile of a PV power in the 12 hours

Figure 8. Profile of a battery power in the 12 hours

Int J Appl Power Eng, Vol. 14, No. 2, June 2025: 308-318
Int J Appl Power Eng ISSN: 2252-8792  317

6. CONCLUSION
This study investigated the optimization of a stand-alone solar power system by improving MPPT
algorithms using ANN and genetic algorithms (GA), specifically the DLGA approach. The results, based on
simulations using atmospheric data from Negrine, Algeria, showed that the DLGA method outperforms
traditional techniques like P&O in maintaining higher, more stable voltages, leading to improved energy
capture. The DLGA also demonstrated superior performance in managing battery charging and discharging
cycles, enhancing battery efficiency and lifespan. Additionally, the ANN models showed effective power
management, and mean squared error analysis confirmed excellent generalization capabilities in the ANN
training process. Overall, this research highlights the potential of intelligent MPPT methods to optimize solar
energy systems, offering more reliable and efficient solutions for regions with high solar potential. The
methodologies presented can serve as a benchmark for future renewable energy optimization efforts.

FUNDING INFORMATION
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
Moufida Saadi ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Dib Djalel ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Kadir Erkan ✓ ✓ ✓ ✓ ✓ ✓ ✓

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
Data availability is not applicable to this paper as no new data were created or analyzed in this study.

REFERENCES
[1] S. Ahmad et al., “A Review of Hybrid Renewable and Sustainable Power Supply System: Unit Sizing, Optimization, Control, and
Management,” Energies (19961073), vol. 17, no. 23, p. 6027, 2024, doi: 10.3390/en17236027.
[2] O. Hafez and K. Bhattacharya, “Optimal planning and design of a renewable energy based supply system for microgrids,”
Renewable Energy, vol. 45, pp. 7–15, Sep. 2012, doi: 10.1016/j.renene.2012.01.087.
[3] H. Rezzouk and A. Mellit, “Feasibility study and sensitivity analysis of a stand-alone photovoltaic–diesel–battery hybrid energy
system in the north of Algeria,” Renewable and Sustainable Energy Reviews, vol. 43, pp. 1134–1150, Mar. 2015, doi:
10.1016/j.rser.2014.11.103.
[4] G. Zubi, R. Dufo-López, M. Carvalho, and G. Pasaoglu, “The lithium-ion battery: State of the art and future perspectives,”
Renewable and Sustainable Energy Reviews, vol. 89, pp. 292–308, Jun. 2018, doi: 10.1016/j.rser.2018.03.002.
[5] R. Luna-Rubio, M. Trejo-Perea, D. Vargas-Vázquez, and G. J. Ríos-Moreno, “Optimal sizing of renewable hybrids energy
systems: a review of methodologies,” Solar Energy, vol. 86, no. 4, pp. 1077–1088, Apr. 2012, doi: 10.1016/j.solener.2011.10.016.
[6] G. Merei, C. Berger, and D. U. Sauer, “Optimization of an off-grid hybrid PV–wind–diesel system with different battery
technologies using genetic algorithm,” Solar Energy, vol. 97, pp. 460–473, Nov. 2013, doi: 10.1016/j.solener.2013.08.016.
[7] R. Sen and S. C. Bhattacharyya, “Off-grid electricity generation with renewable energy technologies in India: an application of
HOMER,” Renewable Energy, vol. 62, pp. 388–398, Feb. 2014, doi: 10.1016/j.renene.2013.07.028.
[8] A. Mellit and S. A. Kalogirou, “Artificial intelligence techniques for photovoltaic applications: a review,” Progress in Energy and
Combustion Science, vol. 34, no. 5, pp. 574–632, Oct. 2008, doi: 10.1016/j.pecs.2008.01.001.
[9] H. X. Yang, L. Lu, and J. Burnett, “Weather data and probability analysis of hybrid photovoltaic–wind power generation systems
in Hong Kong,” Renewable Energy, vol. 28, no. 11, pp. 1813–1824, Sep. 2003, doi: 10.1016/S0960-1481(03)00015-6.
[10] A. A. Abou El Ela, M. A. Abido, and S. R. Spea, “Optimal power flow using differential evolution algorithm,” Electric Power
Systems Research, vol. 80, no. 7, pp. 878–885, Jul. 2010, doi: 10.1016/j.epsr.2009.12.018.
[11] D. Yue, F. You, and S. B. Darling, “Domestic and overseas manufacturing scenarios of silicon-based photovoltaics: Life cycle
energy and environmental comparative analysis,” Solar Energy, vol. 105, pp. 669–678, 2014, doi: 10.1016/j.solener.2014.04.008.

Optimization and dimensioning of stand-alone systems: enhancing MPPT efficiency … (Moufida Saadi)
318  ISSN: 2252-8792

[12] S. Sinha and S. S. Chandel, “Review of recent trends in optimization techniques for solar photovoltaic–wind based hybrid energy
systems,” Renewable and Sustainable Energy Reviews, vol. 50, pp. 755–769, Oct. 2015, doi: 10.1016/j.rser.2015.05.040.
[13] P. Paliwal, N. P. Patidar, and R. K. Nema, “Determination of reliability constrained optimal resource mix for an autonomous
hybrid power system using particle swarm optimization,” Renewable Energy, vol. 63, pp. 194–204, Mar. 2014, doi:
10.1016/j.renene.2013.09.003.
[14] H. Patel and V. Agarwal, “Maximum power point tracking scheme for PV systems operating under partially shaded conditions,”
IEEE Transactions on Industrial Electronics, vol. 55, no. 4, pp. 1689–1698, Apr. 2008, doi: 10.1109/TIE.2008.917118.
[15] K. Karabacak and N. Cettin, “Artificial neural networks for controlling wind–PV power systems: A review,” Renewable and
Sustainable Energy Reviews, vol. 29, pp. 804–827, Jan. 2014, doi: 10.1016/j.rser.2013.08.070.
[16] A. Y. Taha, M. Aljanabi, A. N. Al-Shamani, and Z. H. Hadi, “Intelligent maximum power point tracking for photovoltaic system
using meta-heuristic optimization algorithms: A holistic review,” AIP Conference Proceedings, 2023, vol. 2776, no. 1, p. 050003,
doi: 10.1063/5.0136227.
[17] E. Koutroulis, D. Kolokotsa, A. Potirakis, and K. Kalaitzakis, “Methodology for optimal sizing of stand-alone photovoltaic/wind-
generator systems using genetic algorithms,” Solar Energy, vol. 80, no. 9, pp. 1072–1088, 2006, doi: 10.1016/j.solener.2005.11.002.
[18] M. L. T. Zulu, R. P. Carpanen, and R. Tiako, “A Comprehensive Review: Study of Artificial Intelligence Optimization Technique
Applications in a Hybrid Microgrid at Times of Fault Outbreaks,” Energies, vol. 16, no. 4, p. 1786, 2023, doi: 10.3390/en16041786.
[19] A. G. Abo-Khalil and D. -C. Lee, “MPPT control of wind generation systems based on estimated wind speed using SVR,” IEEE
Transactions on Industrial Electronics, vol. 55, no. 3, pp. 1489–1490, Mar. 2008, doi: 10.1109/TIE.2007.907672.
[20] D. Feroldi and D. Zumoffen, “Sizing methodology for hybrid systems based on multiple renewable power sources integrated to
the energy management strategy,” International Journal of Hydrogen Energy, vol. 39, no. 16, pp. 8609–8620, May 2014, doi:
10.1016/j.ijhydene.2014.01.003.
[21] A. Z. Sahin, K. G. Ismaila, B. S. Yilbas, and A. Al‐Sharafi, “A review on the performance of photovoltaic/thermoelectric hybrid
generators,” International Journal of Energy Research, vol. 44, no. 5, pp. 3365–3394, Apr. 2020, doi: 10.1002/er.5139.
[22] A. A. Kebede et al., “Techno-economic analysis of lithium-ion and lead-acid batteries in stationary energy storage application,”
Journal of Energy Storage, vol. 40, p. 102748, Aug. 2021, doi: 10.1016/j.est.2021.102748.
[23] D. P. Jenkins, J. Fletcher, and D. Kane, “Lifetime prediction and sizing of lead–acid batteries for microgeneration storage
applications,” IET Renewable Power Generation, vol. 2, no. 3, pp. 191–200, 2008, doi: 10.1049/iet-rpg:20080021.
[24] V. H. Johnson, “Battery performance models in ADVISOR,” Journal of Power Sources, vol. 110, no. 2, pp. 321–329, Aug. 2002,
doi: 10.1016/S0378-7753(02)00194-5.
[25] D. Rekioua, T. Rekioua, A. Elsanabary, and S. Mekhilef, “Power management control of an autonomous photovoltaic/wind
turbine/battery system,” Energies, vol. 16, no. 5, p. 2286, Feb. 2023, doi: 10.3390/en16052286.
[26] 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, vol. 21, no. 4, p. 1244, Feb. 2021, doi: 10.3390/s21041244.

BIOGRAPHIES OF AUTHORS

Moufida Saadi is a Ph.D. student at the University of Tebessa in Algeria. She

specializes in renewable energy and the application of artificial intelligence. Her work primarily
involves integrating AI technologies with renewable energy systems to enhance efficiency and
sustainability. Her research is focused on developing intelligent algorithms for managing solar
and wind energy, contributing significantly to the advancement of green technology. She has
made her mark by presenting my findings at various international conferences and publishing
papers in emerging energy journals. Her academic journey includes a master's degree in
electrical engineering with a specialization in renewable energy. She can be contacted at email:
[email protected].

Dib Djalel is a full professor and currently works as a research director in the
Department of Electrical Engineering at the University of Tebessa in Algeria. He received his
Ph.D. in power systems in 2007 in Algeria, and he was awarded the professor rank in December
2015. He holds the position of research director and head of the electrical engineering team at
LABGET and LM laboratory at the University of Tebessa. He participated in several
conferences and published several papers, now he is a reviewer for several journals. His research
interests are power grids, power quality, and renewable energy. He can be contacted at email:
[email protected].

Kadir Erkan is an associate professor at Yildiz Technical University in the

Department of Mechatronics Engineering. He received his Ph.D. in electrical engineering and
specializes in the design and control of electromagnetic systems, micro-robotics, and magnetic
levitation. Erkan has participated in numerous international conferences and has published
extensively on topics such as hybrid electromagnetic systems and precise control techniques in
mechatronics. His research contributions have been instrumental in advancing the fields of
robotics, automation, and control systems. He can be contacted at email:
[email protected].

Int J Appl Power Eng, Vol. 14, No. 2, June 2025: 308-318