2018/06/02 by Malini Ghosal, Ghosal, Malini
Engineering · #FOS: Mathematics #Fault Detection and Control Systems #Optimization and Control (math.OC) #Power System Optimization and Stability #Smart Grid Security and Resilience
paper · pdf · doi:10.48550/arxiv.1806.00705
openalex publication_date 2018/06/02 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
In a state estimator, the presence of malicious or simply corrupt sensor data\nor bad data is detected by the high value of normalized measurement residuals\nthat exceeds the threshold value, determined by the \χ2 distribution.\nHowever, high normalized residuals can also be caused by another type of\nanomaly, namely gross modeling or topology error. In this paper we propose a\nmethod to distinguish between these two sources of anomalies - 1) malicious\nsensor data and 2) modeling error. The anomaly detector will start with\nassuming a case of malicious data and suspect some of the individual\nmeasurements corresponding to the highest normalized residuals to be\n`malicious', unless proved otherwise. Then, choosing a change of basis, the\nstate space is transformed and decomposed into `observable' and `unobservable'\nparts with respect to these `suspicious' measurements. We argue that, while the\nanomaly due to malicious data can only affect the `observable' part of the\nstates, there exists no such restriction for anomalies due to modeling error.\nNumerical results illustrate how the proposed anomaly diagnosis based on Kalman\ndecomposition can successfully distinguish between the two types of anomalies.\n