Power Network Topology Optimization and Power Flow Control Based on Deep Reinforcement Learning

In the pursuit of carbon neutrality, huge changes on the power supply side and the load side have brought forward new requirements and challenges for grid operation and dispatchers. A low-cost and effective measure is real-time power grid network topology optimization and control (NTOC). However, except for the simplest action of line switching, the combinatorial and non-linear nature of the NTOC problem has made existing approaches infeasible for grids of reasonable scales. This paper proposes a novel artificial intelligence (AI) based approach for maximizing available transfer capabilities (ATCs) via network topology control considering various practical constraints and uncertainties. First, imitation learning is utilized to provide a good initial policy for the AI agent. Then, the agent is trained through deep reinforcement learning with a novel guided exploration technique, which significantly improves the training efficiency. Finally, an early warning mechanism is designed to help the agent identify a proper action time, which effectively improves the fault tolerance and robustness of the method. The effectiveness of the proposed approach is tested by using open-sourced data of the IEEE 14-note system.