Advancing Coherent Power Grid Partitioning: A Review Embracing Machine and Deep Learning

Abstract

With the escalating intricacy and expansion of the interconnected electrical grid, the likelihood of power system (PS) collapse has escalated dramatically. There is an increased emphasis on immunizing renewable-dominated power systems from large-scale cascading failures and cyberattacks through optimal power grid partitioning (PGP). By altering the network’s topology, partitioning aims to create areas within the PS that are not only robust but also have increased flexibility in generation and improved controllability over variable demand. This article provides an updated review of the cutting-edge machine learning and data-driven techniques used for PGP in networked PSs. To this end, an in-depth exploration of the basic principles of PGP and performance quantification is provided. The coherency adequacy and controlled islanding within the power network are comprehensively discussed. Subsequently, state-of-the-art research that envisions the use of clustering-based machine learning and deep learning-based solutions for PGP is presented. Finally, key research gaps and future directions for effective PGP are outlined. This paper provides PS researchers with a bird’s eye view of the current state of mainstream PGP implementations. Additionally, it assists stakeholders in selecting the most appropriate clustering algorithms for PGP applications.