Malicious data injection attacks: A relaxed physics model based strategy for real-time monitoring

Abstract

With the rapid advances in the infrastructure of power networks, modern power systems have become vulnerable to cyber-attacks. An attacker can mislead the operators in power system control centers by introducing malicious data that affect the outputs of the state estimator which turn disrupts in the operation and control functions of many power system applications. Hence, an accurate and fast algorithm for detecting, identifying and correcting malicious data injection attacks is crucial to prevent catastrophic failures in power systems. This paper presents further contributions to power system real-time monitoring in the presence of a malicious data injection attacks. State of the art solutions consider either measurement or parameter is free of error when estimating the state variables, such as complex voltages. However, malicious data in measurements and parameters can be injected simultaneously and such assumption does not provide an accurate solution. In this work, a relaxed model strategy is proposed to handle such simultaneous data attack. The framework of measurement gross error analysis is deployed in processing and analyzing attacks. Chi-square X2 Hypothesis Testing applied to the normalized composed measurement error (CMEN) is considered for detecting cyber-attacks. The property of largest normalized error test is used for identifying malicious data injection. The correction of cyber-attack considers the type of attack and the composed normalized error (CNE) in a relaxed model strategy that takes into account the effect of the measurement in error when correcting the attacked parameter. The proposed model is validated on IEEE 14-bus system.