False Data Injection Detection in Power System Based on LASSO and Ensemble Mashine learning

The digitization of modern power systems through smart grids has introduced new vulnerabilities, especially False Data Injection Attacks (FDIAs), which can compromise the integrity and security of real-time data. This project, titled "False Data Injection Detection in Power System Based on LASSO and Ensemble Machine Learning", presents a comprehensive detection framework that integrates LASSO regression for feature selection and a combination of machine learning and deep learning models including Decision Tree, Random Forest, AdaBoost, CNN-LSTM, and Graph Neural Networks (GNN). To enhance trust and interpretability, LIME (Local Interpretable Model-Agnostic Explanations) is incorporated, providing human-understandable insights into model decisions. A full-stack Flask web application has been developed, enabling functionalities like user registration, login, data upload, result visualization, and real-time prediction with explanations. Evaluated on a public FDIA dataset from Kaggle, the system demonstrates high accuracy, low false alarm rates, and operational transparency. The solution is designed to assist power grid operators in detecting and mitigating cyber threats effectively.

Keywords:
False Data Injection Attack, Smart Grid, LASSO Regression, Ensemble Machine Learning, Decision Tree, Random Forest, AdaBoost, CNN-LSTM, Graph Neural Network, Feature Selection, Lime,Cybersecurity in Power Systems, Power Grid Monitoring.