Probabilistic Forecasting of Market Operations via Online Learning

A key challenge to operate the modern grid is the increasing level of uncertainty arising from both supply and demand sides. How to characterize future operating conditions in the presence of renewable generation and stochastic load with behind-the-meter distributed resources is crucial. In this talk, we will present an online learning approach to probabilistic forecasting market operations. The proposed technique provides estimates of joint probability distributions of power flows, locational marginal prices, and network congestion. Case studies will be shown to demonstrate its efficiency and scalability.