Prediction of Building Stock Energy Demand

Abstract

Recent years witnessed an increasing interest in the prediction of building energy demand at large-scale. The prediction of large-scale building (or building stock) energy use is essential for energy policy development, energy management, urban development decision, and distributed power generation. However, it is not a simple task to estimate large-scale building energy consumption because of significant uncertainties in building information for a variety of critical characteristics. Modeling every single building within a building stock is impractical, and proper methods are hence inevitable to reduce modeling efforts and simulation time. This study presents a stochastic building stock energy modeling approach using archetypes and Bayesian calibration. The paper introduces the procedure of the proposed method and then demonstrates and validates the method with a campus-scale application with 80 buildings. The predicted campus-scale building energy demand matches the measured energy data and provides much comprehensive knowledge on the building performance with estimated stochastic distributions of building energy usages.