Sensitivity-Based Heterogeneous Ordered Multi-Agent Reinforcement Learning for Distributed Volt-Var Control in Active Distribution Network

Abstract

As power grids expand, maintaining stable voltage and minimizing losses become increasingly crucial. Meanwhile, the widespread use of heterogeneous devices in modern distribution systems necessitates effective multi-device coordination. This issue is exacerbated by the integration of intermittent renewable sources (e.g., solar and wind), which introduce voltage fluctuations. To tackle these challenges, this paper proposes a novel Sensitivity-based Heterogeneous Ordered Multi-agent Reinforcement Learning (SHOM) method for Volt-Var Control (VVC) in Active Distribution Networks (ADNs). By leveraging voltage-reactive sensitivity to explicitly guide sequential policy updates, SHOM ensures a monotonic improvement in control strategies under heterogeneous, networked constraints. Experimental results on IEEE test feeders demonstrate that the proposed approach achieves superior voltage regulation and lower power losses compared to existing methods.