Akademik Veri Yönetim Sistemi
JOURNAL OF POLYTECHNIC-POLITEKNIK DERGISI, 2023 (ESCI)
Solar trackers maximize solar radiation collection but are less commonly used due to their high cost, maintenance requirements, and the additional expenses associated with monthly angle adjustments. The main goal of this effort is proposing the optimization of solar energy absorption by determining the optimal tilt for fixed-site solar panels in Turkey. It introduces a mathematical equation that uses artificial neural networks to predict the ideal angle based on five specific features of the selected locations (latitude, longitude, Julian day, hour, and altitude) and cos 0, which are used for training and testing without requiring complicated calculations. Input variables, training procedures, and network design significantly impact the accuracy of Neural Network models' predictions. Using MATLAB, three distinct multilayer ANN models for this investigation were created, each employing unique training setups and procedures, MATLAB graphs are used to select algorithms and models based on the minimum MAE and RMSE, while the linear correlation coefficient (R) should be maximum. The RMSE value obtained according to the calculations of selected model which employs the feed forward Lunberg-Marquardt training algorithm, was 3.5881e-6, and the R value was 0.99998. The estimated data of the network were compared to the cos0 data, which were used for training and testing, yielding an RMSE error of 0.43% and an R2 value of 0.99978, indicating high accuracy. The average annual optimum inclination angles for the studied cities are as follows: Ankara (35.18 degrees), Antalya (34.29 degrees), Agri (34.91 degrees), Istanbul (34.50 degrees), Sivas (34.96 degrees), Izmir (35.19 degrees), Sinop (35.06 degrees), and Gaziantep (34.97 degrees).