Context-Aware Machine Learning for Real-Time Energy Management in EV-Based Microgrid Systems

Abstract

The transformations of traditional electricity distribution into decentralized, dynamic microgrids have been rapid through the proliferation of electric vehicles and distributed renewable power generation technologies. This transformation creates challenges for managing EV-based microgrids with real-time energy management due to the stochastic nature of the load profiles, the variability due to renewable generation sources, and heterogeneous behavior from users. This paper provides a context aware machine learning architecture that combines temporal, spatial and behavioral context data in order to predict short term demand and subsequently provide adaptive charging/discharging schedules for both EVs and local storage resources. The architecture uses lightweight deep learning based predictive models, combined with reinforcement learning based decision agents, to produce schedules in consideration of both power and user centric relationship constraints. Simulation results demonstrate improvements in peak shaving, renewable energy usage, service level adherence while maintaining EV user preferences. Additionally, operational costs have been reduced as well as an increase in grid stability relative to baseline heuristic strategies. The proposed architecture is scalable and can be utilized in edge-enabled controllers to achieve resilient and efficient microgrid operations during periods of uncertainty. The design conserves user privacy through the utilization of privacy-protecting data handling techniques and provides efficient edge deployment capabilities.