Resilience and Cybersecurity Enhancement of Smart Grids Using Machine Learning

Abstract

The development of modern power infrastructure witnessed a paradigm shift due to the appearance of smart grid (SG) technology. Smart grids integrate various distributed energy resources (DERs), advanced metering, and two-way communication capabilities. The integration results in significant advancements in terms of control and efficiency, but at the same time, new vulnerabilities are introduced into the system. Natural disasters such as floods, earthquakes, and wildfires, as well as cyberattacks on the SG system, proved that cascading failures may occur within the smart grid architecture while being unattainable in a traditional power grid. In light of these challenges, resilience transitioned from a qualitative characteristic to an important design parameter. This article reviews the threat landscape facing today's SGs and discusses the methodologies and metrics used in measuring and achieving resilience. An emphasis is placed on machine learning (ML) algorithms, especially recurrent neural networks, which became widely popular in detecting anomalies and attacks based on state data analysis. The study concludes with an evaluation of existing techniques' shortcomings and potential areas of future research.