Photovoltaic Power Analysis and Prediction Using Machine Learning Methods

Due to the rapid increase in population and industri-alization in different parts of the world, new power resources are being developed to meet the rapid increase in power demand. The concern of global warming and environmental impact reinforces shifting from traditional power resources such as fossil fuel to renewable resources such as photovoltaic (PV) sources. The stochastic nature of PV power directly affects the stability of the grid. Therefore, PV power forecasting allows power stations to know before hand how much PV power will be available, which assists in ensuring that the grid remains in stabilized condition. Here, PV power from India is analyzed and predicted using machine learning methods. The main goal of this paper is to analyze power patterns and predict the future 15 minutes of PV power using random forest regression. The analysis shows that irradiance and ambient temperature have the highest correlation with the DC power. It also shows patterns of similarities between power consumption profiles on different days of the week.