Model-free based Real-time Authentication Framework for Distributed Synchrophasors

Cyber-attacks have the potential to evade bad synchrophasor data detection in wide-area monitoring systems (WAMS), thus having a detrimental impact on situational awareness and synchrophasor-based applications. To tackle this challenge, this paper proposes a model-free synchrophasor data authentication framework to defend against false data injection. The principal components analysis and gramian angular field are first combined to extract the critical features from the synchrophasor data in the distributed power system. Considering the massive synchrophasor measurement data, a convolutional neural network is further presented to fulfil the data authentication based on the extracted features. The efficacy of the real-time authentication framework is confirmed through various experiments. The experimental results demonstrate an accuracy rate of 93.15% when utilizing field synchrophasor data, without the need for knowledge of system model parameters.