A Systematic Literature Review on big data for solar photovoltaic electricity generation forecasting

Abstract

Solar power is expected to play a substantial role globally, due to it being one of the leading renewable electricity sources for future use. Even though the use of solar irradiation to generate electricity is currently at a fast deployment pace and technological evolution, its natural variability still presents an important barrier to overcome. Machine learning and data mining techniques arise as alternatives to aid solar electricity generation forecast reducing the impacts of its natural inconstant power supply. This paper presents a literature review on big data models for solar photovoltaic electricity generation forecasts, aiming to evaluate the most applicable and accurate state-of-art techniques to the problem, including the motivation behind each project proposal, the characteristics and quality of data used to address the problem, among other issues. A Systematic Literature Review (SLR) method was used, in which research questions were defined and translated into search strings. The search returned 38 papers for final evaluation, affirming that the use of these models to predict solar electricity generation is currently an ongoing academic research question. Machine learning is widely used, and neural networks is considered the most accurate algorithm. Extreme learning machine learning has reduced time and raised precision.