Key points are not available for this paper at this time.
Distribution systems are integrating a growing number of distributed energy resources and converter-interfaced generators to form active distribution networks (ADNs). Numerous studies have been undertaken to mitigate various challenges in ADNs. However, voltage deviation and reactive power control still requires more attention from researchers and power system engineers. The Volt/VAr control (VVC) concept has been developed to improve the voltage quality, minimize active power losses, and maintain the voltage profile in ADNs. The deployed utility-owned legacy voltage control mechanisms such as on-load tap changers, capacitor banks, and automatic voltage regulators operate in discrete, slow timescales and unidirectionally, rendering them insufficient for optimal voltage regulation in ADNs. Owing to the increasing use of smart meters, smart inverters (SIs), smart sensors, data analytics tools, and improved communication networks, data has become an important resource. Data-driven control approaches, particularly reinforcement learning (RL)-based, have therefore gained more attention in recent years in effectively solving the VVC decision-making problem. This comprehensive review presents a detailed analysis of advanced approaches used to address the VVC problem. It includes a general overview of the problem formulation, control frameworks, and basic notations, as well as detailed comparisons of the existing and recently proposed methods. This study focuses on data-driven approaches, especially RL-based algorithms. Some of the open research challenges experienced in the application of these algorithms such as safety, data, scalability, communication problems, interpretability and cybersecurity threats are presented alongside the future research perspectives such as Internet of Things (IoT), Transfer Learning (TL), hybrid and human-in-the-loop AI approaches.
Abdelkader et al. (Mon,) studied this question.