Lyapunov-Regularized Reinforcement Learning for Power System Transient Stability

Transient stability of power systems is becoming increasingly important because the growing integration renewable resources. These resources lead to a reduction in mechanical inertia but also provide increased flexibility frequency responses. Namely, their electronic interfaces can implement almost arbitrary control laws. To design these controllers, reinforcement learning (RL) has emerged as powerful method searching for optimal non-linear policy parameterized by neural networks. A key challenge enforce that learned controller must be stabilizing. This letter proposes Lyapunov regularized RL approach transient lossy Because lack an analytical function, we learn function network. The losses are specially designed with respect physical system. then utilized regularization train network penalizing actions violate conditions. Case study shows introducing enables stabilizing and achieve smaller losses.