Distributed Unsupervised Detection for Robust Power System False Data Attacks via Flexible Dynamic Time Warping Strategy

Abstract

This article studies the modeling method of false data injection attacks (FDIAs) considering topological changes and relevant countermeasures. A novel robust FDIA model is built, which incorporates network and measurement uncertainties into the attack subnet and can be applied to attack scenarios during topological changes. To tackle such FDIAs, a flexible dynamic time warping strategy-based distributed unsupervised detection mechanism is developed. Furthermore, an enhanced recognition model via hierarchical agglomerative clustering and local outlier factor techniques is proposed to facilitate operators to distinguish FDIAs. Compared with related works, the proposed model is more applicable to topological change scenarios and the detection framework can effectively discern such FDIAs in a distributed fashion. Simulation results demonstrate the stealthiness of the proposed FDIA model during and related to topological changes and the effectiveness of the distributed unsupervised detection method in tackling such attacks.