A Deep Reinforcement Learning-Based Real-Time Control for Transfer Limits of Critical Inter-Corridors

Abstract

Controlling transfer power flow below transfer limits of inter-corridors is crucial for power system security. Traditional way that empirically decides pessimistic limits incurs low utilization of inter-corridors. To improve operational flexibility, a real-time intelligent controller that enables precise tracking for transfer limits is carried out. The concerned problem is firstly modelled based on the limit qualification by total transfer capability (TTC). To allow real-time controller, which is reinforcement learning (RL), to learn control law from the model, a physics and data co-driven interactive environment is built, where computational intractable TTC-induced security criterion is substituted by pre-trained supervised learners. Numerical studies on the modified IEEE 39-bus system manifest the reliability and impressive efficiency of the proposed method on transfer limits control.