Deep-Learning-Based Denial-of-Service Resilient Framework for Wide Area Situational Awareness of Power Systems

Abstract

Vulnerability of the wide area measurement system (WAMS) to Denial of Service (DoS) attacks can hinder the real-time situational awareness of the grid. In this regard, a DoS cyber-attack resilient WAMS framework is presented, which utilizes deep-learning based overcomplete denoising autoencoder (ODAE) architecture. The proposed framework is able to reconstruct PMU data during DoS attacks on PMU-PDC connections for the entire period of power system contingencies, which helps the operator to be aware of the occurrence of any such disturbance and assist them in decision making related to the stability of the grid. The performance of the proposed framework has been validated on WSCC 9 bus system, which is simulated on the developed cyber-physical WAMS testbed involving real time digital simulator (RTDS), hardware PMUs, Hardware PDC, network switches etc. Subsequently, the response of the suggested framework has also been evaluated on the real field PDC data gathered from the Northern Region of the Indian Power Grid (NRIPG). The experimental results indicate that the reconstructed PMU measurements during DoS attacks help in maintaining the dynamic visualization of system-wide information thereby ensuring the resiliency of wide area monitoring applications against DoS attacks.