Intelligent Solar Power Forecast with Machine Learning

Solar power forecast is a necessity for efficient operation of Smart Grids for good energy management. Before forecasting any data related to solar power, it is important to gather precise data related to several aspects of solar irradiance; which being the main input for generation of solar power. Globally, solar radiation forecasting is performed by several methods, two of them main methods being cloud imagery combined with physical models, and machine learning models. The objective of this paper is to provide an overview of the building of machine learning models for the forecasting of solar power from historical data and how the use of libraries like XGBoost can improve scalability and accuracy in prediction. A few of the recent works on solar irradiance forecasting using other approaches include Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM). Deep Learning can also be utilised for very precise forecasting. In further studies, different approaches can be compared to figure out the ideal approach to build a model for this specific task. This paper also covers certain aspects of artificial neural networks and how they are the building blocks of this particular machine learning model.