Fiodar
Kazhamiaka,
PhD
candidate
David
R.
Cheriton
School
of
Computer
Science
I will present two research studies in which we explore the problems of sizing and operating PV-storage systems in a data-limited environment. In our work, we go against the grain of many similar studies by making the realistic assumption that the data which typically guides these decisions is not available at the time of system deployment.
For system design, we compare the effectiveness of three data generation approaches and two direct sampling methods in their ability to generate PV generation and electricity consumption traces, from which we could then make robust system sizing decisions. Extensive experiments based on real data show that it is possible to calculate a robust system size with only one year's worth of hourly trace data. Moreover, we demonstrate how to perform robust sizing with access to only twelve data points of household electricity consumption, one for each month of one year.
For system operation, we design a controller which encodes a model predictive control policy in a neural network, with periodic re-training as data is collected by the system in order to adapt the controller to a dynamic operating environment. We evaluate our controller in the context of a solar PV-storage system deployed in Texas under a time-of-use pricing scheme. We find that our controller gets to within 5-10% of optimal performance, and outperforms a default PV-storage system control strategy within a few months of installation.
Bio: Fiodar is a Ph.D. candidate in computer science at the David Cheriton School of Computer Science. His research interests are in the design and control of systems with energy storage and renewable energy sources. In his spare time, he enjoys playing beach volleyball around the world.