A framework for the design of representative neighborhoods for energy flexibility assessment in CityLearn

Abstract

The electricity grid in the U.S. continues to undergo system-wide changes due to electrification of end-uses, as well as the adoption of intermittent on-site renewable energy systems. When effectively managed, distributed energy resource in buildings can provide energy flexibility to alleviate grid loads at critical periods. However, it is important to understand the role of geographic, climatic and occupant behavioral differences on the effectiveness of their flexibility. Thus, we provide a framework that uses an open-source U.S. building-stock database with clustering techniques to design representative neighborhoods for distributed energy resource control algorithm benchmarking. We demonstrate an application in three neighborhoods that use reinforcement learning control for energy storage system management and simulation results show up to 42% reduction in peak demand, amongst other key performance indicators.