Research on Auxiliary Control Strategy for Large-scale Power Grid Based on Deep Reinforcement Learning

Large-scale power grid sections often exceed their limits due to the participation of power generation units in the frequency regulation auxiliary service market. To address this issue, this paper proposes an auxiliary control strategy for large-scale power grid sections based on deep reinforcement learning. The Deep Deterministic Policy Gradient (DDPG) algorithm is applied to construct an intelligent agent for power grid section control, providing real-time control strategies in complex power grid environments, while considering both the safety and economy of power grid operations. On this basis, a two-stage optimization strategy considering sensitivity is proposed. In cases where the section limit cannot be eliminated through real-time control, the optimal market intervention plan is provided for operators to eliminate the section limit. Finally, case studies are designed to verify the effectiveness of real-time control strategies and market intervention plans. Under different types of section limits caused by power generation units participating in the market, the method proposed in this paper reduces the clearing price by an average of 1.2% and the average adjustment amount by 37.6% compared to the current rules, effectively improving the system's economy.