Automatic Machine Learning Classification Algorithms for Stability Detection of Smart Grid

The smart grid is a kind of leap in the power grid that emerged to improve electrical service and reduce losses. This work tackles the stability detection problem in smart grids using machine learning algorithms. Four different preprocessing scenarios are used along with five different classifiers (Naive Bayes (NB), Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), and Support Vector Machine (SVM)) to find the best classifier and suitable preprocessing steps to achieve the best accuracy. Automated machine learning (AutoML) is used to develop a proposed model to answer two questions. First, is it feasible to use AutoML to solve problems using a limited dataset size? Do complex algorithms (such as deep learning) constantly improve system accuracy? In this study, feature selection (as a preprocessing step) was sufficient to obtain 100% accuracy with three classifiers (LG, DT, SVM). Compared to other studies that use the same dataset, it was found that it is not always beneficial to choose a complex algorithm to get the best results. Moreover, for researchers with limited professionalism in data analysis, AutoML helps study the dataset and select the appropriate machine learning algorithms before turning to use complex algorithms.