Distributed optimization for structured programs and its application to energy management in a building district

This paper deals with structured multi-agent optimization problems that involve coupled local and global decision variables. We propose an iterative distributed algorithm that explicitly accounts for this structure, and requires the agents to communicate only their tentative solutions for the global variables throughout iterations. Our approach extends to structured multi-agent optimization a proximal-based distributed methodology that has recently appeared in the literature. Privacy of local information is preserved and communication effort is reduced with respect to alternative distributed solutions where local and global optimization variables are grouped together and treated as a single decision vector. Multi-agent optimization problems with the considered structural properties appear in various contexts. In this paper, we apply our approach to energy management in a district where multiple buildings can communicate over a possibly time-varying network and aim at optimizing the use of shared and local resources. We illustrate the efficacy of the resulting distributed energy management algorithm by means of a detailed simulation study on a cooling problem.