Machine Learning Models for Power and Surface Temperature Values of Different Types of Photovoltaic Solar Panels under Antarctic Climate Conditions

In this study, the performances of four different photovoltaic (PV) type solar panels were investigated on the Antarctic Horseshoe Island within the scope of the 8th National Antarctic Science Expedition organized between 27 January-3 March 2024 under the coordination of TÜBİTAK MAM (Marmara Research Center) Polar Research Institute and the expedition coordinator ship of Prof. Dr. Burcu OZSOY. The power values and surface temperatures of each PV panel were measured under the same experimental conditions. The PV panel with the best electrical power generation performance for Antarctica, a disadvantaged region, was determined. In addition, the power and temperature values of PV panels were modeled using six machine-learning methods that considered air temperature, wind speed, humidity, and solar radiation values. The machine learning method best expressed the surface temperature and power values for monocrystalline, polycrystalline, flexible, and transparent PV solar panels. This study is unique because it contains Antarctic climatic data and PV panel types whose performances are analyzed for the first time under these climatic conditions. In addition, using these data in the applied machine learning methods and generating useful models is another study’s originality.