Unscented Kalman Filter based interval state estimation of cyber physical energy system for detection of dynamic attack

Information and communication technologies (ICTs) introduce many Internet-based entry points that pose potential risks to cyber physical energy system (CPES). Therefore, it is significant to enhance the cybersecurity of CPES from cyber-attacks. In this paper, a new dynamic attack model that takes into account the dynamic characteristics of energy systems is developed based on traditional false data injection attack. The proposed attack model can be used to describe the attack behaviors of a malicious attacker over time. Then, we propose a new generalized interval state estimator to quantify the normal fluctuations of all CPES state variables. In this state estimator, the Unscented Kalman Filter (UKF) is used to predict the real-time operating level of the state variables. Copula theory is introduced to model the prediction uncertainty of sustainable energy and load as a set of conditional quantiles. We then model the normal fluctuation range of each CPES state as a bilevel nonlinear programming problem based on the worst case analysis. Consequently, an anomaly detection method is developed to detect whether there is an attack or not in the CPES. In this method, any state variable that falls outside its estimated interval is considered an abnormal point. Finally, the feasibility of the dynamic attack model and the effectiveness of the anomaly detection method have been extensively validated on test systems in power and energy society of the Institute of Electrical and Electronics Engineers (IEEE).