Reinforcement learning for vehicle-to-grid: A review

Abstract

The rapid development of Vehicle-to-Grid technology has played a crucial role in peak shaving and power scheduling within the power grid. However, with the random integration of a large number of electric vehicles into the grid, the uncertainty and complexity of the system have significantly increased, posing substantial challenges to traditional algorithms. Reinforcement learning has shown great potential in addressing these high-dimensional dynamic scheduling optimization problems. However, there is currently a lack of comprehensive analysis and systematic understanding of reinforcement learning applications in Vehicle-to-Grid, which limits the further development of this technology in the Vehicle-to-Grid domain. To this end, this review systematically analyzes the application of reinforcement learning in Vehicle-to-Grid from the perspective of different stakeholders, including the power grid, aggregators, and electric vehicle users, and clarifies the effectiveness and mechanisms of reinforcement learning in addressing the uncertainty in power scheduling. Based on a comprehensive review of the development trajectory of reinforcement learning in Vehicle-to-Grid applications, this paper proposes a structured framework for method classification and application analysis. It also highlights the major challenges currently faced by reinforcement learning in the Vehicle-to-Grid domain and provides targeted directions for future research. Through this systematic review of reinforcement learning applications in Vehicle-to-Grid, the paper aims to provide relevant references for subsequent studies.