Real-time Autonomous Line Flow Control Using Proximal Policy Optimization

The security of operating modern power grids is often challenged by the increasing penetration of renewable resources and nature disasters due to their intermittent and uncertain nature. At severe operating conditions with major topology changes and/or regional power imbalance, violations of line flow limits may occur in a short time. Consequently, deriving effective control decisions to rapidly mitigate such violations become necessary to avoid power line tripping and potential cascading outages. This paper presents a novel method that explores the full potential of Proximal Policy Optimization (PPO), one promising deep-reinforcement-learning (DRL) algorithm, to provide real-time line flow control decisions. A DRL agent learns its optimal control strategy from scratch through massive interactions with a grid simulator, which can instantaneously respond to rapidly changing operating conditions once properly trained. The training and testing procedures of such DRL agents are conducted on both IEEE 14-bus and Illinois 200-bus systems. Outstanding control performance is observed in autonomously regulating line flows under various load conditions, which validates the effectiveness of the method.