Optimal energy management in smart buildings

Chaitanya Poolla (ECE '16) addressed the problem of optimal energy management in a stochastic environment under the guidance of Dr. Abe Ishihara and Dr. Rodolfo Milito. Their work was recently accepted for publication in Applied Energy, a prestigious journal in the field. This research was supported by Cisco Systems, Inc., and contributed toward Chaitanya's Ph.D. dissertation.

The researchers investigated the problem of designing optimal Energy Management (EM) policies for smart buildings. The problem was considered in the presence of uncertain solar generation, uncertain building demand, battery dynamics, and utility pricing. Decision problems were posed within a Markov Decision Process (MDP) framework. They postulated rule-based baselines, and proposed a near-optimal policy design using Stochastic Dynamic Programming (SDP). A comprehensive treatment for handling system constraints in the SDP approach was also presented. The near-optimal policy is expected to significantly reduce operating cost compared to the rule-based alternatives. To demonstrate the efficacy of the method, simulations were presented in the context of residential and commercial building environments.

This work is useful in addressing several real-world problems including battery/PV portfolio optimization, optimal utility pricing and demand response programming in the presence of demand-supply uncertainties. 

The preprint of their research is available on arXiv for more details. The final article is expected to appear in Applied Energy during Spring 2019.