Classification of pollution in solar panels by deep learning

Before people can use the energy they need, they must produce it. They primarily preferred fossil fuels to produce energy. The world began to look for new energy sources as a result of environmental pollution. As a result of the search, they started to use renewable energy sources. They started to benefit from energy sources that have been used since the world existed and that are not harmful to the environment. As renewable energy sources, they produced energy from wind, solar, geothermal, biomass and wave energy. Due to the environmental friendliness of renewable energy sources, countries have started to do more R&D studies. They preferred renewable energy sources in energy production due to the decrease in fossil fuels, their damage to the environment and foreign dependency. The sun, which is the largest renewable energy source and the largest energy source, is the most preferred energy production source. Energy is produced from solar energy by means of photovoltaic panels. Photovoltaic panels, which are only costly in terms of installation, produce energy by requiring the least manpower from the sunrise to the sunset. Photovoltaic panels polluted due to weather conditions in the world negatively affect energy production. Energy production in polluted photovoltaic panels decreases. In the literature study I have done, it is seen that there is a direct ratio between clean photovoltaic panels and the amount of energy produced. Considering the researches on solar panels, the electricity production of dirty panels has been examined. Photovoltaic panels must be clean in order to be used efficiently. In this thesis study, photovoltaic panels pollution level, classification analysis was made on solar panel pictures with NasnetLarge and MobilNet algorithms, which are deep learning methods. Classification analysis with NasnetLarge was 97%, and classification analysis with MobileNet was 98%. With this thesis, it is aimed to make the pollution classification of the solar panels installed in large areas quickly and in a shorter time with deep learning and to notify the relevant units for cleaning the panels.