Adaptive Multi‐Objective Coordinated Scheduling of Distribution Areas Based on Deep Reinforcement Learning

With the high penetration of renewable energy sources, distribution networks face significant challenges with voltage fluctuations and energy imbalances. Grid‐Forming Distribution Areas (GFDA), with their active voltage regulation capabilities, offer a novel technical solution for the distribution network operations. However, existing research predominantly employs multi‐objective dispatch models with fixed weights, which struggle to adapt to dynamic changes in grid operational states. To address this issue, this paper proposes an adaptive multi‐objective coordination dispatch method for distribution areas based on deep reinforcement learning. First, a power support framework is established for the collaborative operation of Grid‐Forming and Grid‐Following Distribution Areas (GFLDA). Second, a state‐aware reward weight dynamic allocation mechanism is designed to achieve adaptive adjustment of weights between voltage quality and economic objectives. Finally, a stable and efficient reinforcement learning training framework is constructed based on the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. Simulation results on the IEEE‐33 bus system demonstrate that, compared to fixed‐weight schemes, the proposed method achieves significant improvements in both voltage stability and economy, particularly in extreme scenarios with sudden load increases, effectively suppressing voltage limit violations while reducing network losses under normal operating conditions.