Performance analysis and modeling of bifacial PV system with solar tracking

In this thesis, experimental analyses were carried out on eight different configurations of bifacial PV and monofacial PV panelsinstalled on fixed and single-axis solar tracking systems under the meteorological conditions of Şanlıurfa. A comprehensive dataset was created using PV energy production, irradiation on the panel surface, ambient temperature, module temperature, wind speed, and absolute humidity data, all recorded every 15 minutes over a year. Using this dataset, annual energy production comparisons, performance analyses, the effects of panel pollution, PVsyst simulation validations, feasibility studies for agricultural voltaic applications, carbon emission analyses, and deep learning-based modelling studies were conducted. As a result of the analyses, it was determined that the single-axis solar tracking system provided 18.57% higher annual energy production and a 9.91% higher bifacial gain in bifacial PV panelscompared to the fixed system, while achieving 8.23% higher energy production in monofacial PV panels. The solar-tracking bPV configuration provided 23.07% higher annual energy production than the fixed mPV configuration. The average yearly performance ratio values, calculated according to the IEC 61724-1 standard, are 0.86 for solar-tracking bPV, 0.83 for solar-tracking mPV, 0.86 for fixed bPV, and 0.85 for fixed mPV. The highest energy production per watt was achieved with the solar-tracking bPV configuration, at 2067 Wh/W. PVsyst simulations predicted an average annual energy production 10% higher than the experimental results. CNN, MLP, LSTM, CNN-LSTM, Stacked LSTM and Bidirectional LSTM architectures were used in the modelling phase, and the average R2 value of all models was calculated as 0.989. Especially the Bidirectional LSTM and LSTM architectures stood out, with lower error rates (RMSE, MAE, MSE) and higher R2 values compared to other models in many configurations.