Predictive analytics of the generation and distribution of photovoltaic energy: modeling and applications via computational intelligence algorithms

Abstract

Understanding the energy exportation model as promoted by photovoltaic systems that are connected to the electricity grid is of paramount importance in the field of power generation. The development of predictive methods in this context can be useful in making important decisions, specially for government agencies, local consumers, and the whole energy industry. Moreover, getting such a forecasting regarding the use of this kind of energy leads to a greater energy security to the country. This research aims at studying the impact caused by the energy conveyed to the power grid by residential and commercial photovoltaic systems, considering for that the use of Machine Learning (ML) and Exploratory Data Analysis (EDA) tools. In order to enable the proposed methodology and its validation in a real scenario, our study focuses on the Queensland state, Australia, which bears a vast repository of public data for research purposes. In addition, Australia has made important advances in the field of photovoltaic energy, a fact that supports our research in the sense of having a rich technical apparatus as well as a practical archetype of how distributed generation of solar energy behaves in a fully integrated grid. In order to conduct this research, three ML models have been considered: Random Forest, Support Vector Machine, and Gradient Boosting. The models will be implemented and validated by using real databases of generation and distribution of photovoltaic energy so that the computational framework to be developed will be useful towards aiding energy efficiency plans for both government agencies and Brazilian electric power industry.