Analytics for Demand Response Optimization in a Microgrid

In a Smart Grid, both the utility and the consumer can benefit from analytics based on electricity consumption data. In this abstract, we report our on-going research work on performing analytics for the USC campus micro-grid using electricity consumption and curtailment data available for campus buildings. Our goal is to provide the facility managers an insight into the changing electricity demand patterns on campus and enable them to efficiently plan consumption and curtailment activities; and to develop a policy engine for automating Demand Response (DR). The research challenge for us is to design, develop and field analytics solutions for various tasks in DR Optimization. We leverage characteristic information about the campus including academic schedules and building details to derive novel indirect indicators of electricity consumption. We use a combination of learning techniques such as Granger causality, regression trees and time series that together provide a better forecasting accuracy than individual approaches. Eventually, we expect to generalize and extrapolate the successful models that work for the campus to a city scale.