Impact and detection of cyber attacks in wide area control application of cyber-physical power system (CPPS)

Abstract

The most important factor of a comprehensive power grid cybersecurity strategy is the assessment of the effects of cyberattacks. Its insights facilitate the development of resilience, proactive risk management, and effective response plans to emerging cyber threats. To evaluate the potential consequences of a cyberattack on grid infrastructure, it is essential to examine the extensive impact of cyberattacks within the framework of cyber-physical power systems (CPPS). The article investigates the extensive impacts of cyberattacks across three unique scenarios, namely single cyberattack (SCA), coordinated cyber-physical attack (CCPA), and multiple cyberattacks (MCA). These attack scenarios are tested in the wide-area control application of the New England 39-bus test system. Classifiers such as Random Forest (RF), K-Nearest Neighbour (KNN), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) and Support Vector Machine (SVM) identify threats using a learning-based approach. We assess attack detection by multiple performance indicators, including accuracy, precision, and the F-score. Simulation results indicate that MCA is more harmful than a single cyberattack or a coordinated attack. Furthermore, the CNN classifier surpasses other classifiers in attack detection efficacy.