Smart Grid Seminar

Increase in supply side variability due to increases in renewable generation requires integrating new resources utilizing improved models of the power system to reduce electricity delivery costs. Data from consumers, markets and devices in power networks has become broadly available and represent an opportunity to learn better models of the behavior of consumers, power networks and markets. In this talk we introduce two problems in statistical learning from data power networks: reconstructing distribution power networks and learning virtual bidding dynamics in markets. In the first part of the talk, we introduce VADER (Visualization and Analytics for Distributed Energy Resources), a system that learns models of distribution networks by fusing traditional utility SCADA data with novel sources of information such as measurements from inverters and smart meters. We demonstrate how appropriately constructed maximum likelihood inference and machine learning based models can significantly outperform traditional approaches in estimating network topology and parameters. In the second part of the talk we focus on models of wholesale electricity markets developed to understand virtual bidding.  We demonstrate that data can be utilized to test market efficiency and to characterize competitive equilibrium conditions in the presence of virtual bidding utilizing market and agent models learnt from data.