Experimental and Theoretical Study Based on Artificial Neural Networks of Photovoltaic Thermal Collector Using Air Cooling Through a Hybrid Solar Chimney

The growing global demand for sustainable energy solutions has led to increased interest in hybrid systems that combine photovoltaic (PV) and thermal (T) energy recovery. The performance of such systems, however, is often limited by high temperatures, which reduce both electrical and thermal efficiencies. Solar chimneys (SCs), widely recognized for their ability to enhance airflow and cooling, offer a promising solution to mitigate this issue. Recent studies have optimized various parameters of SCs, such as chimney height, tilt angle, and air gap, to improve thermal performance and enhance energy recovery. Despite these advances, there remains a significant gap in research focusing on the integration of SCs with hybrid photovoltaic/thermal (PV/T) systems to enhance their electrical and thermal performance [1, 2]. How to meet the ever-increasing needs and wants in an environmentally sustainable manner has become one of the major issues to be addressed by contemporary societies [3]. Traditional Carbon-Based energy harvesting methods have been instrumental in economic growth, but the upcoming challenges of resource depletion, environmental pollution, and climate change questions on-going use of these resources [4]. The evolution of industrial development in modern economies is profoundly influenced by the adequacy and efficiency of their energy supply systems. Their energy usage influences people's standard of living [5]. There are many purposes for which we need energy, such as transport, agriculture, industries, lighting, and all the home appliances we use. It is estimated that fossil fuels were used to produce electricity before the Industrial Revolution [6]. The unsustainable depletion of conventional fossil energy resources at an accelerated pace has resulted in a serious “energy crisis” and the deterioration of environmental conditions, such as ozone layer degradation, emission of harmful greenhouse gases (GHGs) into the atmosphere, and global warming, which not only pose a hazard to human security but also to the future sustainability of Planet Earth [7]. Therefore, it is increasingly becoming important to seek clean and environmentally friendly energy sources in an effort to reduce reliance on fossil fuels and mitigate the adverse environmental effects imposed [8]. Solar power is being the most accepted and accessible renewable energy generated which are relatively clean, abundant, sustainable and inexhaustible resources of light energy [9, 10]. PV converting systems are a growing trend for harnessing solar energy into electric power and thermal energy [11].

Figure 1 shows the operation Principles of the hybrid PV/T system and configurations where both electrical and thermal conversion are operated in sequence to enhance the energy transformation/efficiency. In the first place, PV panels absorb solar radiation and transform a part of this solar energy into electric power. During the conversion, surplus heat is released as waste and stored in the PV cells, causing an increase in the temperature of operation. To reduce this thermal heating, a cooling system is used to dissipate the excess heat from the PV module. Thus, the recovered thermal energy is further used for beneficial heating purposes such as space heating, drying, or hot water supply. Additionally, by properly utilising and recycling the produced heat, system efficiency is improved. This coupled cycle illustrates that the combined use of electrical and thermal energy simultaneously results in enhanced performance and sustainability of PV/T systems. Magnetocaloric coupling, as opposed to traditional engines, based on a thermodynamic fermion particle model, 18 one can instead exploit magnetocaloric.

Figure 1. Schematic illustration of the energy conversion and heat recovery cycle in a hybrid photovoltaic/thermal (PV/T) system

While several studies have explored the role of Computational Fluid Dynamics (CFD) and experimental methods in optimizing SC designs, Artificial Intelligence (AI) models, particularly Artificial Neural Networks (ANNs), have not been adequately applied to predict and optimize the performance of hybrid PV/T-SC systems. This gap in research highlights the need for more sophisticated, data-driven models that can simulate real-world conditions and optimize system performance under varying operational parameters. By integrating ANN-based performance prediction with experimental validation, this study aims to address the limitations of traditional methods, offering a novel approach to improving energy recovery, efficiency, and economic viability of hybrid systems, particularly in hot and arid climates.

Solar PV systems produce electricity by directly transforming part of the incident solar radiation into electrical power, and the common conversion efficiencies are between 19% and 21% [12]. The rest of the incoming energy of the sun is reflected either at the surface of the module or the PV cells, which absorb the heat. This heating causes the temperature of the cell to rise, which negatively impacts the electric performance of the PV module. The most common operational issue in PV systems is overheating, especially in areas with large solar irradiance and high ambient temperatures, where excessive heating causes conversion efficiency to decline dramatically [13]. In order to deal with this problem, various cooling and thermal control methods have been suggested and researched over the last few years. One such technology has been PV/T systems, whereby a cooling system is utilized to complement the PV module in order to remove the excess heat. The PV/T system is a setup where a working fluid, liquid or gas, is circulated behind or underneath the PV panel, enabling the system to both produce electrical power and provide useful thermal energy in a single integrated unit. The result of this two-energy production is the improvement of the total efficiency and use of solar energy. PV/T hybrid system is a typical set comprising a PV module and a heat extraction device. Even though the PV module receives almost 80 percent of all the incident solar radiation, not all the energy received is converted into electricity, which is determined by the conversion efficiency of the cells. The rest of this absorbed energy is lost as heat, and this heat can be considered as an output of the PV/T system because it can be removed and used efficiently to reduce the temperature level of the cells and provide an overall better performance of the system [14].

In another CFD analysis is carried out on thermal performances of a flat plate solar collector for various numbers of air inlets [15]. The obtained results reveal that a three-inlet type possesses the maximum value of exit temperature, then two- and on single- inlets, all having higher exit temperatures than those for two and one-inlet designs. In addition, as the heat flux increases and/or the inlet velocity decreases, the exit air temperature increases. Using an underground heat exchanger, the air was precluded in another research work [16]. This study carried out tests to investigate the influence of flow rates and hot ambient air temperatures in the range of 35 ℃, 40 ℃, and 45 ℃ on module efficiency. An effective heat exchanger was demonstrated to maintain the appropriate temperature, and using this cooling method can increase the daily electricity efficiency up to 29.11% [17].

The thermal performance of SCs was presented in a building built with local stones and hollow bricks on the roof. It was found that slope angle, building material selection, and air gap were important to the performance. The optimum chimney parameter was 1 m × 0.65 m(s), with an air gap of 0.25 true and an inclination angle of 45°, giving respective ACH values at solar intensities of approximately 5, 6.5, and 7.8 ACH [18]. Performance enhancement of SCPP by tuning collector tilt angle and chimney designs has been investigated. ANSYS-FLUENT simulation results revealed that the 50% tilt of the collector enhanced the air velocity by 12% and P tapered by 23% higher than that with full tilt. The performance was also better for the semi-divergent chimney coupled with a fully tilted collector, where the average air temperature increase reached 17 K. These results indicate that the correction of collector angle and design of chimney system are vital in improving thermal efficiency of SCPP [19].

Previous studies have shown that SC performance can be significantly improved through the careful optimization of geometric and operational parameters such as chimney height, tilt angle, air gap, thickness, and width. Key factors, including chimney width, air gap, and inclination angle, have been identified as crucial elements influencing airflow and thermal comfort. Building upon these findings, the present study introduces a highly innovative hybrid PV/T system integrated with a SC, combined with ANN modeling for performance prediction and optimization. Unlike traditional approaches that rely solely on CFD simulations or experimental optimization, this study integrates both experimental investigation and ANN-based performance prediction, offering an advanced solution for simulating and optimizing the system under real-world operating conditions. This novel coupling of ANN modeling with physical system enhancement represents a cutting-edge approach to simultaneously improve both electrical power generation and thermal energy recovery, overcoming the operational limitations of conventional solar collectors, particularly in extreme high-temperature environments. The proposed hybrid system significantly expands the potential applications of PV/T-SC technologies, particularly in hot and arid climates, enhancing the economic viability of these systems in solar-assisted cooling, space heating, agricultural drying, and domestic hot water production.

The electrical and thermal efficiencies of a PV/T system are normally measured in terms of both the electrical and thermal efficiency of the system, which combined, represent the capability of the system to efficiently transform incident solar energy into useful electrical power and recoverable thermal energy. Proper evaluation of these performance indicators can be critical in the analysis of the system behavior in various operating and environmental conditions. With the help of the measurements of the current and voltage taken during the operation of the PV panel, the electrical output of the PV panel can be calculated. It is these electrical parameters that allow the calculation of the instant power output and the electrical efficiency of the PV module [21]. Simultaneously, thermal performance is considered by examining the amount of heat that is conducted to the working fluid by the PV module, which is related to the mass flow rate, temperature difference, and specific heat capacity of the fluid [22]. The equations governing the calculation of electrical power, electrical efficiency, and thermal efficiency of the PV/T system are given in the following subsections.

$I s c=I s c_{r e f}+K_I\left(T_c-T_{c, r e f}\right)$   (1)

$V o c=V o c,_{r e f}+K_V\left(T_c-T_{c, r e f}\right)$    (2)

where, $T_c$ is the temperature of the cell Isc represents the short-circuit current, $V O C$ stands for the open-circuit voltage, $\mathrm{K}_I$ and $K_V$ the temperature coefficients for voltage and current, respectively.

$P=I s c \times V o c \times F F$   (3)

where, FF denotes the fill factor, which is assumed to be 0.75. The PV/T module's overall, thermal, and electrical efficiency can be evaluated according to [23]:

$\eta_{\text {ele}}=\frac{P}{G \times A}$    (4)

G denotes the incident solar irradiance on the module in units of watts per square meter (W/m²), while A represents the module's area (m²).

The equation can calculate the module efficiency suggested by [22].

$\eta_m=\eta_c \times P F$    (5)

The maximum power output is obtained based on the approach reported in [23]:

$\eta_c=\frac{P_{\max}}{P_{i n}}=\frac{I_{\max} V_{\max}}{A_c \cdot G}$    (6)

where, $I_{\max }$ is the maximum current, and $V_{\max }$ is the maximum voltage.

The electrical effectiveness of a PV module can be calculated [24].

$\eta_{\text {ele}}=\eta_m\left[1-\beta\left(T_C-T_{C, \text { ref}}\right)\right]$   (7)

where, $\beta$ is the Silicon temperature coefficient, which is 0.0045 1/℃, $T_{C, \text {ref}}$ is the temperature of the cell at 25 ℃.

The air's mass flow rate can be computed according to [25]

$m^{\circ}{ }_{a i r}=\rho_{a i r} V_{a i r} A_c$   (8)

The convective heat transfer to the air is obtained following the approach reported in [26]:

$Q_{\text {conv}}=m_{\text {air}}^{\circ} C p_{\text {air}}\left(T_{o, \text { air}}-T_a\right)$    (9)

The flow rate, specific heat, intake, air exit temperatures, and solar irradiation are all factors that affect thermal efficiency. In a steady-state setting, it is provided according to [27]:

$\eta_{t h}=\frac{Q_{c o n v}}{A_C I_t}$   (10)

This section presents and discusses the experimental and analytical results obtained from the hybrid PV/T system integrated with an SC. The system was operated under forced air-cooling conditions at three different air mass flow rates, namely 0.0635, 0.127, and 0.191 kg/s, in order to evaluate the influence of airflow on thermal regulation and electrical performance.

For each operating condition, the electrical characteristics of the PV module were monitored by measuring the output voltage and current throughout the test period. These measurements were used to calculate the maximum power output and electrical efficiency of the system using Eqs. (6) and (7). The comparative analysis of the three airflow conditions allows assessment of the effectiveness of the SC–assisted cooling mechanism and its impact on both electrical and thermal performance. The following subsections detail the weather conditions, experimental observations, and ANN-based analytical results, providing a comprehensive discussion of system behavior under varying operating and environmental parameters.

5.1 Weather data

The meteorological data obtained in June 2025 were discussed and demonstrated as mean values within the daily working period 7:00 AM to 5:00 PM. The parameters that have been monitored are the solar irradiance and ambient temperature, which are important factors that dictate the performance of the PV/T system. The findings (as shown in Figure 14) reveal that the solar irradiance continued to rise gradually during the morning hours and reached its highest average of about 1100 W/m² at about 2:00 PM. Following this irradiance peak, the irradiance slowly declined towards the late afternoon. Such climatic conditions offered a representative and appropriate data set to test the performance of the system with high levels of solar radiation, which is common during summer seasons in Baghdad.

Figure 14. Hourly profiles of solar irradiance, ambient temperature, and wind speed recorded during the experimental period (07:00–17:00)

5.2 Experimental results

High temperature conditions of PV cells are a significant negative influence on both the electrical capabilities and the overall efficiency of PV panels. As the temperature of PV cells increases, their efficiency decreases, leading to lower power output. This phenomenon is particularly prominent in regions with high solar irradiance and elevated ambient temperatures, where the thermal stress on PV cells becomes more pronounced. To overcome this issue, air was used as the cooling medium in this experiment. This section investigates how varying the rate of air mass flow can be utilized to reduce the temperature of the PV modules and enhance the performance of the system.

Three different air mass flow rates were tested: 0.0635 kg/s, 0.127 kg/s, and 0.191 kg/s. These values were selected to evaluate the effectiveness of air cooling in reducing the temperature of the PV cells and controlling the outlet air temperature. Additionally, the electrical and thermal performances of the PV/T system were compared to identify the most effective cooling approach.

Figure 15 illustrates how the temperature of the PV module changes throughout the day under varying airflow conditions. Without an active cooling system, the maximum temperature of the PV cells reached approximately 68 ℃ during midday. However, when the air-cooling system was activated, a significant drop in module temperature was recorded across all three mass flow rates. The maximum temperatures of the cells were found to be 65 ℃, 58 ℃, and 55 ℃ at the air mass flow rates of 0.0635 kg/s, 0.127 kg/s, and 0.191 kg/s, respectively.

Figure 15. Cell temperature variations with different air mass flow rates

These results demonstrate that increasing the air mass flow rate improves the heat dissipation rate of the PV module, leading to better thermal control and lower operating temperatures. This improved cooling efficiency reduces the thermal stress on the PV cells and prevents performance degradation, allowing the system to operate at its optimal capacity.

The observed reduction in temperature can be attributed to convective heat transfer, where the air flow removes excess heat from the PV module. As the air mass flow rate increases, the convective heat transfer coefficient (h) increases, enhancing the cooling efficiency. This mechanism results in lower PV cell temperatures, which in turn improves the overall electrical performance of the system. The chimney effect further enhances this cooling process, as the vertical air duct generates a pressure difference, which accelerates the airflow, improving heat extraction from the PV cells.

Figures 16 and 17 illustrate the effect of varying air mass flow rates on the electrical characteristics of the PV modules, specifically focusing on output power and electrical efficiency. The results demonstrate a clear correlation between increased airflow and improved electrical performance. As the air mass flow rate increases, the cooling of the PV module surface is enhanced, leading to a significant rise in electrical power output.

Figure 16. Variation of the power maximum

Figure 17. Electrical efficiency of the photovoltaic/thermal (PV/T) system

At the lowest air mass flow rate of 0.0635 kg/s, the maximum measured electrical power output was 83.5 W. As the airflow rate increased to 0.191 kg/s, the power output significantly improved, reaching a higher value due to more efficient cooling. This enhancement in power output can be attributed to the improved convective heat transfer provided by the increased airflow. The higher airflow rates facilitate the dissipation of excess heat from the PV cells, thereby reducing their operating temperature.

Likewise, the electrical efficiency of the system showed a similar upward trend with the increased airflow rate. In the uncooled scenario, the electrical efficiency was approximately 13%. However, as air cooling was applied, the efficiency gradually increased, reaching a maximum value of 15.8% at the highest air mass flow rate of 0.191 kg/s, especially during the early morning hours when the system is subjected to lower ambient temperatures and more direct sunlight.

These findings underline the importance of convective heat transfer in the cooling process. As the airflow rate increases, the rate of convective heat loss from the surface of the PV cells improves. This enhanced heat dissipation leads to a reduction in the thermal load on the PV cells, effectively lowering their operating temperature. Since PV module electrical efficiency is directly proportional to cell temperature, the decrease in operating temperature results in an increase in both electrical power output and efficiency.

The improvements in power output and efficiency underscore the effectiveness of airflow-assisted cooling in optimizing the performance of PV systems, particularly under conditions of high solar irradiance and elevated ambient temperatures. The reduced thermal stress on the PV modules allows for sustained high-performance operation, thereby proving the significant benefits of enhanced convective cooling in PV systems

PV/T systems integrated with SCs contribute significantly to improving system performance by enhancing waste heat recovery and facilitating cooling through air circulation. The temperature differences between the inlet and outlet of the SC under various air mass flow rates are shown in Figure 18. These results reveal the substantial impact that air mass flow rate has on the temperature difference across the SC.

Figure 18. Effect of varying mass flow rates on the temperature differences in the solar chimney (SC)

At the lowest air mass flow rate (0.0635 kg/s), the temperature difference between the inlet and outlet of the chimney increases, reaching a value of 4.8 ℃. This is primarily due to the longer residence time of air within the duct, which allows more heat to be absorbed by the air as it flows past the PV/T module. With more time in contact with the heated surface of the module, the air has a greater opportunity to absorb thermal energy, resulting in a larger temperature gradient between the inlet and outlet of the chimney.

On the other hand, at the highest air mass flow rate (0.191 kg/s), the temperature difference decreases to 2.9 ℃. This reduction is attributed to the higher air velocity that results from increased airflow. As air moves faster, it spends less time in contact with the PV module, reducing the time available for heat absorption. Despite the faster airspeed, the higher flow rate promotes better overall heat transfer, but the rate of temperature change is less pronounced due to the shorter interaction time between the air and the heat source.

In contrast, lower air flow rates tend to lead to a greater temperature difference, as less air is moved through the system, giving it more time to absorb heat. Conversely, higher flow rates do not achieve the same temperature gradient past the heat source but lead to more effective overall heat transfer across the system. This dynamic illustrates that the balance between airflow rate and temperature difference is crucial for optimizing the thermal performance of the PV/T-SC system.

The observed trade-off between flow rate and temperature difference underscores the importance of optimizing air mass flow rates to achieve the best performance. While higher flow rates reduce the temperature difference, they increase the overall heat transfer efficiency, which may be more beneficial for maintaining lower operating temperatures in the PV cells and improving system performance over extended periods. Thus, flow rate optimization is key to maximizing both thermal recovery and cooling efficiency in hybrid PV/T systems.

Figure 19 illustrates the effect of air mass flow rate on the thermal efficiency of the PV/T system. At the lowest air mass flow rate of 0.0635 kg/s, the thermal efficiency was relatively low, ranging between 30% and 50% during the operating period. This lower efficiency can be attributed to the reduced convective heat transfer at lower flow rates, where the air has less time to absorb heat from the PV module, leading to a higher operating temperature for the cells and, consequently, lower thermal performance.

Figure 19. Thermal efficiency of the system

As the air mass flow rate increased to 0.127 kg/s, there was a significant enhancement in the thermal performance of the system, with the thermal efficiency rising to approximately 54%. This increase in efficiency can be explained by the improvement in convective heat transfer. With moderate airflow, the heat extraction from the PV module is more effective, as the air moves faster, reducing the temperature gradient between the module surface and the ambient air, allowing more heat to be transferred out of the system.

The maximum thermal efficiency of approximately 57% was achieved at the highest airflow rate of 0.191 kg/s around 1:00 PM. At this point, the air velocity was sufficient to maximize heat dissipation, and the system's cooling capacity was optimized, leading to the highest recorded thermal efficiency. This improvement can be attributed to the enhanced convective heat transfer between the PV module and the airflow, which accelerates the removal of excess heat, thereby reducing the operating temperature of the PV cells.

In general, the results clearly indicate that an increase in air mass flow rate leads to improved thermal efficiency. The gradual rise in thermal efficiency with increasing airflow supports the notion that airflow optimization is crucial for enhancing the thermal performance of PV/T systems, especially those integrated with SCs. Efficient airflow helps balance the thermal load on the system, optimizing the overall performance by extracting excess heat from the PV cells and ensuring that the system operates at lower temperatures for better efficiency.

5.3 Analytical results of Artificial Neural Network

The present approach involves training an ANN model to optimize network weights based on the experimental data obtained from the system. The model is trained using a dataset collected over a period of one month in Baghdad, where key meteorological conditions—including ambient temperature, solar radiation, and relative humidity—were continuously monitored. This data serves as the foundation for tuning the ANN model, allowing it to learn and adapt to the environmental factors that affect system performance.

In addition to meteorological data, electrical and thermal measurements were also conducted throughout the testing period. These measurements included voltage, current, and PV cell temperature, which were recorded under a wide range of operating conditions. The data collected during the experiment were crucial for understanding the relationship between solar irradiance, ambient temperature, and the operational efficiency of the PV/T system. These measurements were then used to train the ANN model, enabling it to predict the system's performance under various environmental conditions and optimize its operation for maximum efficiency.

5.3.1 Experimental and Artificial Neural Network model results for photovoltaic module temperature

A comparative analysis of the converged ANN prediction models is presented in Table 2, which includes a detailed evaluation of the convergence time per training epoch and the fitting behavior of the models to both the training and testing datasets. This comparison was conducted to assess the precision and robustness of different ANN structures in predicting the temperature of the PV module. To determine the appropriateness and validity of each model, the measured and predicted PV module temperatures were compared using several common performance indices, including:

• Mean Bias Error (MBE)
• RMSE
• Mean Percentage Error (MPE)
• Correlation Coefficient (R)

Table 2. Comparative performance of Artificial Neural Network (ANN) models

The results indicate significant variations in the accuracy of predictions across the tested network architectures. Among the different ANN configurations, the 6-4-3-1 ANN architecture exhibited the best predictive performance, achieving the highest correlation coefficient (R = 0.966) and the lowest MPE (MPE = 4.3003%). These results validate the superior ability of this architecture to model the nonlinear relationship between the input variables and the temperature of the PV module accurately.

Additionally, this chosen 6-4-3-1 ANN model provides an optimal balance between network complexity and computational efficiency. The model successfully captures the complex nonlinear trends within the data without overfitting, ensuring that it generalizes well to unseen data. This capability makes it highly applicable for reliable performance predictions under various operational conditions, demonstrating that it can be effectively deployed for real-time performance forecasting and optimization of PV/T systems.

The validity and reliability of the simulation model were verified by comparing the model's outputs to the corresponding experimental data. This comparison is crucial for evaluating the model’s potential to accurately represent the physical behavior of the hybrid PV/T system under various operating conditions. Figure 20 presents a comparison of simultaneous daytime measurements of PHM-CT sensor data and ANN model predictions of the cell temperature over a typical daily period, from 8:00 AM to 6:00 PM. Both the experimental findings and the predicted values (represented by the fitting curve) are shown in the Figure 20, demonstrating that the two sets of data are in excellent agreement.

Figure 20. Comparison of actual and predicted temperature values

The temperature of the PV module starts at 30 ℃ in the early morning and gradually increases as solar radiation intensifies. The maximum temperature of 52 ℃ is reached between 1:00 PM and 2:00 PM, coinciding with peak solar irradiance. At midday, the maximum difference between the predicted and observed temperatures is less than 1 ℃, which indicates exceptional model accuracy under high solar radiation conditions. This suggests that the ANN model is highly effective at capturing the system's thermal behavior and accurately predicting the cell temperature even during rapid fluctuations in solar irradiance.

Furthermore, the ANN model accurately responds to sudden changes in cell temperature, demonstrating its capability to adapt to dynamic operating conditions. This strong agreement between the experimental data and the model predictions confirms the reliability of the model in predicting thermal performance under varying environmental factors, making it a robust tool for optimizing PV/T system performance.

Figure 21 shows the predicted and test outlet air temperatures of the SC over the amount of 20 test samples. The results show a good consistency of the ANN-predicted values with each of the actual measurements and that deviations within the full range of the sample are small, confirming good precision for the proposed model. The outlet air temperature shows a gradual increase from around 42 ℃ at the first sample up to its maximum of about 51 ℃ at the 15th sample. Below this level, there is relatively little cooling observed in succeeding samples. This trend reflects the solar insolation and ambient condition difference during measurements, and proves again the accuracy of ANN model to trace the dynamic thermal performance of SC.

Figure 21. Actual and predicted outlet temperature of the solar chimney (SC)

5.3.2 Experimental and Artificial Neural Network model results for efficiency modeling

A systematic assessment of the predictive capabilities of the ANNs developed for predicting the thermal and electrical efficiencies of the PV/T system is presented in Table 3. The comparison highlights the performance of different network architectures and their ability to predict efficiency profiles under varying operating conditions. The analysis shows that different ANN structures perform differently depending on the operating modes, with certain designs offering better prediction accuracy for specific system behaviors.

Table 3. ANN structure performance in predicting PV/T efficiencies

Among the tested models, the 6-10-1 ANN architecture delivered the most accurate predictions of thermal efficiency, with an impressive coefficient of determination (R² = 0.9951) and an RMSE = 0.3747. These performance metrics indicate that the 6-10-1 ANN model is highly reliable in predicting thermal efficiency, with only minimal deviations from the experimental data. The R² value close to 1 suggests that the model explains a very high proportion of the variability in thermal efficiency, while the low RMSE indicates that the predictions are highly consistent with the measured values.

Figure 22. Comparison of the electrical efficiency

Figure 23. Comparison of the thermal efficiency

Figures 22 and 23 present a comparative analysis between the experimental and ANN-estimated thermal and electrical efficiencies of the PV/T system. The results indicate an excellent match between the actual and predicted values across all tested operating points, confirming that the proposed ANN model is highly accurate. The model's predictions closely follow the temporal patterns and degradations of both thermal and electrical efficiencies, capturing the system's behavior under both peak and off-peak performance conditions.

The small residuals observed between the experimental data and the predicted values further emphasize the high accuracy of the ANN model. These minor deviations highlight that the model is able to predict system performance with a high degree of reliability, demonstrating its ability to effectively model the complex and nonlinear relationships inherent in the behavior of the hybrid PV/T–SC system.

These findings validate the ANN model as a trustworthy tool for modeling and predicting the performance of PV/T systems. Given its ability to handle the nonlinear characteristics of the system and its strong agreement with experimental data, the ANN model can be employed as an efficient computational tool for performance prediction and analysis under varied climatic and operational conditions. The results further support the potential of using ANN-based models for optimizing system performance and guiding future PV/T system designs.

5.3.3 Performance evaluation using key metrics

To evaluate the precision and strength of the ANN model, key performance indices were used, including the MBE, RMSE, and the correlation coefficient (R). These metrics, when applied to both thermal and electrical efficiency predictions, indicate that the model exhibits exceptional performance:

• Thermal Efficiency: The model achieved a correlation coefficient (R² = 0.9951), indicating that it explains over 99% of the variance in thermal efficiency. The RMSE (0.3747) further validates the model's accuracy by indicating minimal error in the predictions.
• Electrical Efficiency: The model’s predictions were similarly accurate, with residuals showing a high degree of correlation between experimental and predicted values, demonstrating that the model consistently captures electrical performance under real conditions.

Table 4 shows the comparison between the ANN model predictions, traditional regression models, and theoretical models.

Table 4. Comparison of Artificial Neural Network (ANN) model predictions with traditional regression and theoretical models

5.4 Comparison with previous studies

As the demand for sustainable energy solutions continues to grow, optimizing the performance of PV/T systems and SCs has become a key area of research. Various methods, including cooling techniques for PV panels and adjustments to SC configurations, have been explored to enhance system performance. Additionally, the integration of ANNs has shown promise in improving the electrical and thermal efficiency of these hybrid systems.

Table 5 provides a comparison of significant studies in this field, focusing on different approaches to optimizing PV/T systems and SCs. The table summarizes key findings from studies that examined hydraulic cooling methods, the impact of cooling techniques on conversion efficiency, and the optimization of SC design parameters. The table also highlights the results of the present study, which integrates an ANN-based optimization approach to improve both electrical and thermal performance of the hybrid PV/T system with a SC

Table 5. Comparison of studies on photovoltaic/thermal (PV/T) systems and solar chimney (SC) optimization with Artificial Neural Network (ANN) integration

Even though the results of the latest study are valuable in determining the performance of the hybrid PV/T–SC system, there are some limitations that should be noted. Here are these restrictions:

• The current experiments were performed under summer-specific climatic conditions of Baghdad, which may restrict the generalization of results to other climates or seasons.
• The system was tested only within a limited air mass flow rate range, so the optimum operating conditions may not have been accessed.
• It relied on a single monocrystalline PV module and did not explore the impact of different PV technologies or module regimes.
• The ANN model was developed based on a small amount of data, and that might cause some errors in predicting for an untested operation range.
• The ANN model is data-based and doesn't consider the detailed operation mechanism of heat transfer and airflow in the system.
• Effects of long-term operations, including aging and dust accumulation, were not included in the analysis.