Session: Optimization and Resilience in the Power Grid
Chair: Madeleine Udell
Cluster: Optimization Applications (Communication, Energy, Health, ML, ...)
Talk 1: Leveraging predictions for optimal voltage control: an adaptive approach
Speaker: Wenqi Cui
Abstract: High variability of solar PV and sudden changes in load (e.g., electric vehicles and storage) can lead to large voltage fluctuations. In recent years, a number of advanced controllers have been designed to optimize voltage control. These controllers, however, almost always assume that the net load in the system remains constant over a sufficiently long time. Given the intermittent and uncertain nature of renewable resources, it is becoming important to explicitly consider net load that is time-varying. This talk will describe an adaptive approach to voltage control in power systems with a significant time-varying net load. We leverage the advances in short-term load forecasting, where the net load in the system can be predicted using local measurements. We integrate these predictions into the design of adaptive controllers, and prove that the overall control architecture achieves input-to-state stability in a decentralized manner. We optimize the control policy through a sample-efficient reinforcement learning framework, which update the control policy successively with online date collection . Case studies are conducted using time-varying load data from Caltech's campus grid.
Talk 2: Learning-enhanced Design and Optimization of Microgrids under Uncertainty
Speaker: Harsha Nagarajan
Abstract: To mitigate the vulnerability of distribution grids to severe weather events, some electric utilities use preemptive de-energization as a primary defense, leading to significant power outages. Networked microgrids can enhance resiliency and maximize load delivery, but challenges arise from modeling unbalanced three-phase networks and managing uncertainties in renewables and loads. We present a two-stage mixed-integer robust optimization approach to configure and operate networked microgrids, ensuring robustness to all net-load uncertainties while maximizing load delivery. To solve this problem, we propose an ML-accelerated cutting-plane algorithm with convergence guarantees, which approximates a recourse function with cuts predicted by an ML regressor for worst-case uncertain scenarios. A case study on the IEEE 37-bus system demonstrates the economic benefits of networking microgrids to maximize load delivery.
Talk 3: A Reliability Puzzle for Large Scale Batteries
Speaker: Steven Diamond
Abstract: Large scale batteries are playing an increasingly prominent role in electricity markets. Battery operators earn revenue through two main sources: energy arbitrage and ancillary services. Ancillary services are commitments to support the grid through actions outside of the standard market mechanism. For example, a battery might promise to provide extra power to the grid in an emergency. In this talk we explore the reliability opportunities and challenges posed by batteries selling ancillary services. We discuss the contrasting approaches taken by the California and Texas electricity markets and propose a new mechanism that better aligns the interests of market regulators and battery operators.
