Enhancing Power Grid Resilience Against Cyber Threats in the Smart Grid Era Using Bi-LSTM Model

Abstract

Incorporating communication technology into the Smart Grid (SG) is proposed as an optimal approach to address the requirements of the contemporary power system. Numerous vital sectors, including transportation, electric energy grids, and healthcare, are progressively integrating information and communication technology to boost their effectiveness and dependability. These systems, known as Cyber-Physical Systems (CPS), are now confronting a growing risk of cyberattacks. Malicious actors with advanced knowledge of these systems can exploit vulnerabilities, disrupt operations, and access sensitive information without detection. In this research, we present an innovative method aimed at identifying and countering both intelligent and malicious cyber-physical system attacks, thereby bolstering the resilience of these systems. Specifically, we implement this approach within power systems, which can be characterized by linear frequency dynamics in proximity to their standard operational state. Our approach harnesses the power of the Bi- LSTM model and taps into publicly accessible datasets pertaining to cyberattacks on power systems to uncover concealed internal system information, safeguarding against potential attackers. Furthermore, we conducted an extensive exploration of several well-established semisupervised anomaly detection algorithms to determine their effectiveness. Our comparative analysis, when pitted against commonly used supervised algorithms, demonstrates that semi-supervised algorithms surpass their supervised counterparts in terms of detecting and flagging potential attack incidents.