Power grid stability prediction using stacked machine learning based classification and regression models
Uma Rani V1, Jebamalar Tamilselvi J2, Karthika RN3, Deepa J4 and Hemavathy P5
Associate Professor, Department of Cybersecurity, Faculty of Science and Humanities,SRM Institute of Science and Technology, Ramapuram, Chennai,Tamil Nadu,India2
Assistant Professor, Department of Information Technology,Saveetha Engineering College, Chennai,Tamil Nadu,India3
Assistant Professor, Department of Information Technology,Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology,Tamil Nadu,India4
Assistant Professor, Department of Computer Science and Engineering,Saveetha Engineering College, Chennai,Tamil Nadu,India5
Corresponding Author : Uma Rani V
Recieved : 29-October-2024; Revised : 25-January-2026; Accepted : 27-January-2026
Abstract
Electrical power grid stability (EPGS) is critical for the reliable transmission and distribution of electricity and for preventing power outages. Consequently, the timely identification of potential grid disturbances, optimal resource allocation, and the implementation of proactive maintenance and stability control strategies are essential. This study aims to predict power grid stability using a hybrid stacked machine learning (ML) framework. The proposed approach develops a stacked ensemble classification model comprising ten different ML classifiers, along with a separate stacked regression model built using ten regression algorithms. The framework effectively captures complex relationships among multiple influencing factors, including load demand, frequency variations, and voltage stability. Advanced data preprocessing techniques—such as correlation analysis, multicollinearity assessment, and principal component analysis (PCA)—are employed to enhance data quality and improve feature interpretability. Experimental results demonstrate that the proposed stacked model achieves high predictive performance, with a classification accuracy of 0.95, an F1-score of 0.94, and strong regression performance, reflected by an R² score of 0.9412 across all base ML models. Furthermore, local interpretable model-agnostic explanations (LIME) and Shapley Additive exPlanations (SHAP) are integrated to provide both local and global interpretability, enabling transparent decision-making for EPGS prediction and identifying the key features influencing grid stability.
Keywords
Power grid stability, Stacked ensemble learning, Machine learning, Explainable artificial intelligence, LIME and SHAP, Feature engineering.
Cite this article
V UR, J JT, RN K, J D, P H. Power grid stability prediction using stacked machine learning based classification and regression models. International Journal of Advanced Technology and Engineering Exploration. 2026;13(134):144-159. DOI : 10.19101/IJATEE.2024.111101957