2021/03/25 by Thanh Long Vu, Sayak Mukherjee, Vu, Thanh Long +5 · 4 citations
Engineering · #Power System Optimization and Stability #Smart Grid Security and Resilience #Smart Grid Energy Management
paper · pdf · doi:10.48550/arxiv.2103.14186
Under voltage load shedding has been considered as a standard and effective\nmeasure to recover the voltage stability of the electric power grid under\nemergency and severe conditions. However, this scheme usually trips a massive\namount of load which can be unnecessary and harmful to customers. Recently,\ndeep reinforcement learning (RL) has been regarded and adopted as a promising\napproach that can significantly reduce the amount of load shedding. However,\nlike most existing machine learning (ML)-based control techniques, RL control\nusually cannot guarantee the safety of the systems under control. In this\npaper, we introduce a novel safe RL method for emergency load shedding of power\nsystems, that can enhance the safe voltage recovery of the electric power grid\nafter experiencing faults. Unlike the standard RL method, the safe RL method\nhas a reward function consisting of a Barrier function that goes to minus\ninfinity when the system state goes to the safety bounds. Consequently, the\noptimal control policy, that maximizes the reward function, can render the\npower system to avoid the safety bounds. This method is general and can be\napplied to other safety-critical control problems. Numerical simulations on the\n39-bus IEEE benchmark is performed to demonstrate the effectiveness of the\nproposed safe RL emergency control, as well as its adaptive capability to\nfaults not seen in the training.\n