High-Fidelity Self-Learning Tool for Residential Load and Load Flexibility Forecasting

Sponsor: Department of Energy (DOE) via Fraunhofer USA

Award Number: DE-EE0009696

The project will research, develop, and demonstrate technology that enables the modulation of controllable household loads, to provide multiple grid services, including peak capacity management, ramp support, and frequency regulation.  We will show how the fusion of data from multiple sources, including communicating thermostats, smart appliances, weather forecasts, utility bills, solar production data, and interval electric meter data can be used as inputs to a grey-box model of household net load to increase the accuracy of generation and load forecasts.  Model predictive control (MPC) will be used to plan the electricity use by flexible resources at individual dwellings while accounting for factors such as occupant comfort and resource availability.  Responsive loads will be aggregated and used to offer demand response in a globally optimal fashion according to pre-defined policy objectives, using pricing as a feedback signal to coordinate the behavior of individual nodes.

The work in Paschalidis’ Network Optimization and Control (NOC) laboratory will focus on designing a methodology for coordinating decisions at individual homes through the use of a coordination agent that receives information about predicted behavior for individual nodes and determines optimal aggregated response based on the specific combination of services on offer.  For this component, we will use a pricing feedback signal to induce coordinated, yet distributed, load balancing among individual nodes, while taking into account inferred comfort preferences using an inverse optimization methodology we have developed in earlier work.  The proposed approach will decentralize these decisions; designing a mechanism with which detailed load-scheduling decisions are made locally and the information exchanged between the individual homes and the coordination agent reduces to a pricing signal communicated to individual homes and a response communicated to the coordination agent.

PI: Dr. Michael Zeifman, (Fraunhofer Center for Manufacturing Innovation, U.S.A.)

Other Senior Personnel: Dr. Kurt Roth (Fraunhofer), Matt Kromer (Fraunhofer), Professor Yannis Paschalidis (Boston University), Dr. George Zavaliagkos  (Sense Labs), Mr. Chris Micali (Sense Labs), Mr. Michael Phillips (Sense Labs).

Learn more here.