HetGPS: Scalable Graph Multi-Agent Reinforcement Learning with Physics-Anchored Adaptive Safety for EV Charging

2026-08-01

AI Summary

This work addresses the challenge of voltage violations in distribution networks caused by large-scale electric vehicle (EV) charging, where it is difficult to balance individual agent autonomy with shared safety constraints. The authors propose HetGPS, a novel framework that integrates graph neural networks with power system physics to decouple intervention intensity from correction direction. Specifically, an action-conditional graph residual model dynamically allocates intervention authority, while physical laws dictate safe correction directions. This approach minimizes both safety interventions and policy disruption, and its architecture is decoupled from fleet size. Evaluated across five distribution networks with 200 to 3,218 EVs, the method reduces voltage violation rates to 0.52–3.44% while maintaining departure success rates of 99.06–100%. Remarkably, with only 383,702 parameters, it enables zero-shot transfer to larger networks, demonstrating high efficiency and scalability.

Abstract

Safety interventions for large populations of network-coupled agents must protect shared constraints without unnecessarily overriding task-oriented policy decisions. We present HetGPS, a hybrid graph-control framework synergizing learned graph risk with physics-anchored correction by separating intervention magnitude from corrective direction. An action-conditioned graph residual model schedules state-dependent intervention authority, while a physics model determines its direction. For electric vehicle (EV) charging, we couple this filter with a parameter-shared heterogeneous graph soft actor-critic policy, enabling topology-aware coordination with a learned model size independent of fleet size. Across five nested distribution networks with 200--3,218 EVs and 100 evaluation days, Adaptive Authority reduces bus--step voltage violations from 3.93--7.74\% without filtering to 0.52--3.44\%, while maintaining 99.06--100\% departure success. Relative to the same physics-directed projection with fixed authority, it improves mean reward on all five networks and lowers the mean safety score on four. The deployed policy-and-risk model contains 383,702 learned parameters at every scale; at 3,218 EVs, a matched centralized SAC actor is about $170\times$ larger. A policy trained on the eight-transformer system transfers zero-shot to the 16- and 32-transformer systems, attaining 0.57--0.75\% violation rates and at least 99.99\% departure success. These results show that learned graph risk can allocate intervention authority at scale while feeder physics anchors corrective action.

Problem

• EV charging
• safety constraints
• multi-agent reinforcement learning
• voltage violations
• scalable coordination

Innovation

• graph multi-agent reinforcement learning
• physics-anchored safety
• adaptive intervention authority
• scalable EV charging coordination
• heterogeneous graph policy