Statistical analysis of building energy load profiles to assess site-specific feasibility of demand charge management

Abstract

Demand charge management (DCM) is a process in which behind-the-meter energy storage devices are discharged during periods of peak energy consumption in order to reduce demand charges established by utilities. Recent efforts focus on developing predictive algorithms to determine when peak demand will occur for individual buildings and creating dispatch logics to dictate when energy storage devices should be charged or discharged. Typically, these methods seek to quantify the performance of DCM in terms of monetary savings, though they have wide error margins and make limiting assumptions regarding utility tariff structures and battery energy storage system (BESS) costs. As such, there is need for an approach to prioritize candidate sites with imprecise knowledge of future loads, demand charges, and battery costs needed to estimate explicit monetary savings. This paper thus develops a statistical model to assess the viability of DCM for candidate buildings based solely on load profile characteristics and rough characterizations of buildings including square footage, construction type, and building use. Instead of outputting explicit DCM profit, the model estimates annual peak demand reduction, in order to evaluate the baseline DCM potential. This estimation can then be used in conjunction with various utility tariff structures and battery costs to determine monetary savings under a variety of scenarios. The model employs multiple linear regression to explore load characteristics from a sample set of one-year historical interval load data for 241 buildings to identify the characteristics that most strongly correlate to annual peak demand reduction. The results indicate whether certain load characteristics are important when assessing a building's base potential to realize significant economic benefits from this service.