Power systems usually employ bad data detection (BDD) to avoid faulty measurements caused by their anomalies, and hence can ensure the security of the state estimation of power systems. However, recently BDD has been found vulnerable to malicious data deception attacks submerged in big data. Such attacks can purposely craft sparse measurement values (i.e. attack vectors) to mislead power estimates, while not posing any anomalies to the BDD. Some related work has been proposed to emphasize this attack. In this paper, a new malicious data deception attack by considering a practical attacking situation is investigated, where the attacker has limited resources for corrupting measurements. In this case, attackers generate attack vectors with less sparsity to evade conventional BDD, while using a convex optimization method to balance the sparsity and magnitude of attack vectors. Accordingly, the effects of such an attack on operational costs and the risks of power systems are analysed in detail. Moreover, according to security evaluation for individual measurements, such attacks can be detected with high probability by just securing one critical measurement. Numerical simulations illustrate the effectiveness of the proposed new attack case and its detection method.
Bad Data Detection and Identification of Hybrid AC/DC Power Systems with Voltage Source Converters Using Deep Belief Network and K-Means Clustering
