Learning and Control for Energy Systems

The past decade has witnessed success of learning-based control in a broad spectrum of applications, such as game play, robotics, and autonomous driving. As a result, the application of learning-based control in power systems has attracted surging attention recently. Despite the promise, one of the biggest challenges for its deployment in power system control is the lack of stability and performance guarantees. Since power systems are critical infrastructure, failure to maintain stability can lead to catastrophic consequences.

In this presentation, I will present some of our recent advances in the design of stable and efficient learning and control algorithms for power systems. I will start by introducing a stability-constrained reinforcement learning (RL) framework that combines policy learning in RL with Lyapunov stability theory to ensure transient stability of the learned policy. We can further provide steady-state optimality guarantees by synthesizing gradient information of the steady-state optimization. This framework, motivated from solving the voltage control problem in power systems, can be generalized to provide both transient stability and steady-state optimality guarantees for learning-based control in a broader class of networked systems that satisfy equilibrium-independent passivity.