Solar power generation is an emerging technology and it has attracted a lot of attention recently. It is divided into two categories: concentrated solar power generation and photovoltaic power generation. Concentrated solar power systems use the thermal energy by concentrating the sunlight. Photovoltaic power systems convert light into electric current using photoelectric effect. Due to high efficiency and convenience of photovoltaics, it is used increasingly in many solar power systems. The performance of photovoltaics is largely dependent on the amount of solar radiation and temperature. Since those factors vary significantly with time of day and season, it is difficult to evaluate the efficiencies of photovoltaic systems. In this paper. using the amount of solar radiation, temperature and generated electric power collected from a solar panel, we propose the probabilistic linear regression models to compute the efficiency of the solar panel. Then, we develop an algorithm to detect failures in the solar panel using the model.