CNN-LSTM for Secure Distributed Demand Response in Smart Grid

Integrating cyber and physical elements in smart grids amplifies susceptibility to false data injection attacks (FDIAs), jeopardizing home automation and energy infrastructure. Traditional security strategies often underperform in FDIA detection due to varied data origins. We propose an advanced anomaly detection framework using CNN-LSTM, tailored to detect FDIAs in the grid’s demand response. Our model employs supervised learning for improved precision when enriched with label information. Empirical tests with genuine energy data from Austin, Texas, demonstrate our model’s superiority over existing methods, with metrics like accuracy, precision, recall, F1 score, and false positive rate consistently affirming its robustness and real-world applicability.