Abstract
Under voltage load shedding has been considered as a standard approach torecover the voltage stability of the electric power grid under emergencyconditions, yet this scheme usually trips a massive amount of loadinefficiently. Reinforcement learning (RL) has been adopted as a promisingapproach to circumvent the issues; however, RL approach usually cannotguarantee the safety of the systems under control. In this paper, we discuss acouple of novel safe RL approaches, namely constrained optimization approachand Barrier function-based approach, that can safely recover voltage underemergency events. This method is general and can be applied to othersafety-critical control problems. Numerical simulations on the 39-bus IEEEbenchmark are performed to demonstrate the effectiveness of the proposed safeRL emergency control.