Deep reinforcement learning for volt-var control: recent advances and future challenges

The world is shifting towards renewable energy sources, leading to significant changes in power systems worldwide. These changes aim to make the systems more sustainable, distributed, dynamic, intelligent, and open. Incorporating high levels of renewable energy, such as solar and wind power, can cause issues for traditional power grids, such as increased power loss and significant voltage violations. The unpredictability and variability of renewable sources make the situation worse, increasing complexity and volatility in power system dynamics. In this paper, we delve into the potential of Deep Reinforcement Learning (DRL) to tackle the challenges faced by ADNs, with a specific focus on improving volt-var control (VVC). We begin by providing an overview of DRL and proceed to discuss its application in VVC. Ultimately, we examine the opportunities and obstacles that lie ahead in implementing these applications.