Multi-Objective Reactive Voltage Optimization Based on MADRL for Distribution Networks

To address voltage limit violations and insufficient reactive power regulation capacity caused by high-penetration renewable energy integration in distribution networks, this paper proposes a multi-objective reactive power and voltage optimization control method based on Multi-Agent Deep Reinforcement Learning (MADRL). By partitioning regions according to the topological structure of the distribution network, renewable energy sources and reactive power compensation devices are allocated to different agents. An agent-based multi-objective optimization model is established with standardized objectives including minimized voltage limit violation deviation, voltage deviation, active power loss, and maximized reactive power margin. The proposed model is solved using the Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) algorithm. Finally, the effectiveness of the proposed method is validated through case studies on a modified IEEE-33 bus system.