Application of Nature-inspired Algorithms for Optimising Photovoltaic System Energy Production

Researchers have explored methods to maximize energy output from PV systems, with tilt and azimuth optimization being a significant area of focus. While some studies have proposed standard guidelines for tilt and azimuth based on regional differences, this project utilizes intelligent optimization algorithms to achieve optimal setting of these variables based on known mathematical model, to maximize energy production. Specifically, the study explores two nature-inspired algorithms, the Genetic Algorithm (GA) and Simulated Annealing (SA). Real-world data was obtained from South African University Radiometric Network (SAURAN) and residential PV system from Pretoria, Gauteng Province for the four seasons of the years 2023 to 2024. Performance was measured using metrics such as prediction accuracy, Standard Deviation (SD), percentage difference, Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). Although both algorithms obtain tilt and azimuth settings that yield better energy production than actual production obtained from the recommended settings used in practice, the optimal settings produced by SA are more realistic (13-40°) compared to GA (30-80°). In terms of standard deviation, both algorithms exhibit high precision (low variability) across all seasons. GA exhibited a lower MAPE, indicating performance that is closer to what is already obtained in the system being studied. MAE values for both algorithms were relatively similar. Finally, sensitivity heat maps demonstrated that irradiance is more stable to variations in tilt and azimuth with SA compared to GA. Overall, this study provides valuable insights for designers and system owners, enabling them to determine the optimal tilt and azimuth angles for effective solar tracking, thereby maximizing energy output