This paper presents an approach for power generation analysis in photovoltaic systems. The objective is to identify the qualitative impact on the energy produced according to fault detection and prediction based on the electrical and weather measurement dataset. Currently, machine and deep learning have been used effectively to recognize patterns in solar panel measurements. Dirt, breakage and shading can greatly reduce the designed efficiency of solar power plants. Thus, increasing the quantity and quality of the failure measurement dataset can significantly improve reliability and its performance.