Power Prediction via Module Temperature for Solar Modules Under Soiling Conditions

The ability to predict the output power of remote solar modules is key to successful wide-scale adoption of solar power. However, solar power is a direct product of its environment and can vary vastly from one location to another. Predicting generated power for a specific facility requires monitoring the output of the solar modules in the context of ambient variables such as temperature, humidity, solar irradiance, air dust, and wind. This is especially challenging in areas where soiling is a significant environmental variable. Soiling particles such as sand and dust can shade segments of the solar module, thus effectively reducing the amount of solar irradiance absorbed and, consequently, the power produced. Measuring soiling particles requires expensive equipment that can increase the cost of running the facility and therefore lower the total output. However, dust can also serve as a cooling layer that can reduce the temperature of the solar module and to a certain extent, reduce overheating. This observation can be used to correlate the amount of dust accumulated on the surface of the panel with its temperature. In this work, the module temperature and power output of a clean module and a dusty module are observed using an Internet of Things monitoring system. The data is used to train various machine learning and deep learning algorithms to eventually predict the output of a soiled module over time using only its temperature and a reference clean module.