Predicting Building Energy Use at a Larger-Scale
In the U.S. the residential buildings and commercial buildings consume 22% and 19% of the total energy use, respectively . More than 50% of the world’s population lives in urban area and the world urban population is expected to rise to 72% by 2050 . Two-thirds of the world’s energy was consumed by urban areas. It hence is extremely important to analyze and understand the pattern and trend of building energy consumption at a city-level. It is necessary and efficient to have a tool that can predict energy consumption for entire building stock without metering or auditing every building in the city. The tool should have sufficient technical details and accuracy that allow the prediction of the implications of new technologies, products, developments, and policies on the current and future energy use of a city.
Most of existing building energy simulation tools can only predict energy load and thermal performance of individual buildings. To estimate the energy use of a city with many buildings of different sizes, types, ages, functions, and operating conditions, simply adding up all energy usages of individual buildings is not correct as buildings operate at different schedules and intensities. Two primary challenges exist in modeling city building energy uses: (a) lack of building details for massive infrastructures (e.g., insulations, glazing); (b) uncertainty of building schedules (e.g., human behaviors, equipment power density). To address these challenges, a probability-based stochastic-deterministic-coupled modeling approach is developed. In this approach, the energy use of a probability-based “representative” building is determined with a deterministic tool (e.g., EnergyPlus) but with stochastic inputs (e.g., building materials, human behaviors). One important criterion to judge the value of this model is that how closely the simulated results can match actual building stock energy use. Several calibration methods (e.g., inverse problem, Bayesian inference) are employed that calibrate the mode using collected field energy data. A city is then composed of a number of such “representative” buildings weighted by land-use functions and densities obtained from urban zoning codes.
We have developed and evaluated the approach and the tool using the campus of University of Michigan Ann Arbor (UM) due to the detailed building and energy information available for 75 buildings on UM campus ,. These 75 buildings were classified into 5 distinct types according to the building function and energy use pattern: campus building, clinic, laboratory, residential building, and service building. The prior distribution and range of each critical building parameter (e.g., insulation of wall and roof) are obtained from literatures for similar structures. Bayesian calibration process was used to calibrate the input variables to match the utility data (Figure 2). Figure 3 shows the calculated EUI distribution using the Monte Carlo simulation method with prior and posterior (calibrated) input parameters, respectively, compared with the measured data. The validated model for each building type can then be used to assess the entire campus building energy use. We are currently improving the method and model and will test an actual eco-city in China to further the development and demonstration.
Figure 1 Energy Use Intensity data in University of Michigan
Figure 2 Process of approach
Figure 3 EUI distributions
 U. EIA, “Annual energy review,” Energy Inf. Adm. US Dep. Energy Washington, DC www. eia. doe. gov/emeu/aer, 2011.
 Department of Economicand Social Affairs, “World Urbanization Prospects The 2011 Revision,” 2012.
 “Campus Information University of Michigan.” [Online]. Available: http://campusinfo.umich.edu/building-search. [Accessed: 06-May-2014].
 “ENERGY MANAGEMENT MICHIGAN.” [Online]. Available: http://energymanagement.umich.edu/. [Accessed: 06-May-2014].