Inference and Control of Electric Loads Given Sparse Measurements and Communications Delay
Johanna L. Mathieu, University of Michigan
April 20, 2016, 11:00am - 12:00pm, EBUII 479
Abstract: Actively engaging distributed energy resources, such as flexible electric loads and energy storage, in electric power system operation can reduce the inefficiency, cost, and environmental impact of the system. However, the sensing and communication requirements of many existing algorithms are significant, and the required infrastructure would be costly to implement. We are developing inference and control methods that rely on existing measurements and communication systems to coordinate distributed loads and storage. In the first part of the talk, I will describe how we apply dynamic mirror descent, an online learning algorithm, to determine the real-time demand of air conditioners served by a distribution feeder using feeder-level active power demand measurements and a collection of load models. These real-time estimates could be used within real-time energy efficiency and demand response strategies. In the second part of the talk, I will describe control and estimation approaches we have developed to shape the aggregate power consumption of thermostatic loads when both the control inputs and output/state measurements are delayed. We compare the performance of a model predictive control approach to a linear quadratic regulator approach.
Bio: Johanna Mathieu is an assistant professor in the Department of Electrical Engineering and Computer Science at the University of Michigan at Ann Arbor. She received her PhD in mechanical engineering from the University of California at Berkeley in 2012 and was a postdoc at ETH Zurich, Switzerland from July 2012 to December 2013. Her research focuses on ways to reduce the environmental impact, cost, and inefficiency of electric power systems via new operational and control strategies, with an emphasis on demand response and energy storage.