Predicting smart grid stability using machine learning algorithms

Abstract

Machine learning (ML) allows applications to become more accurate at predicting outcome using data as inputs to predict output values, there are multiple models providing different results and each of them represents a complement of the other models with a principal goal that is the improvement of the prediction performances, ML is applied in various fields to give good predictions and allow to take the right actions depending on the subject. The current paper represents a case study of ML models use in order to predict the smart grid stability to forecast problems that can occur in the grid when renewable energy is used and electricity's production is not sufficient to meet the consumer's needs. This study is based on the "Electrical Grid Stability Simulated Dataset" provided by the Karlsruher Institut für Technologie, that have been preprocessed and used by various ML models. Extreme Gradient Boosting (XGB) model allowed to have better results and reached an accuracy of 95.23% using f_classif features selection technique.