Early detection of cyber-physical attacks in electric vehicles fast charging stations using machine learning

Abstract

In smart grids, the concept of “vehicle-to-grid” allows the electric vehicles to export power to the grid to support the electric utilities in the power distribution system’s operation. The implementation of such a concept dictates the integration of a set of communication networks, which leads to numerous cyber vulnerability issues. The work in this thesis investigates the development of a novel approach that uses machine learning to early detect such denial-of-service attacks to the fast-charging stations. The study investigated the effectiveness of the proposed approach when considering different time resolutions of the advanced metering infrastructure data including hourly, half-hourly and quarter hourly. The proposed approach has been tested through MATLAB simulation environment on a microgrid equipped with renewable energy resources as well as electric vehicles in vehicle-to-grid-mode. The results have shown that the proposed approach was successful in early detecting cyberattacks at an average accuracy of nearly 98%.