Selection of Building Components Using Sequential Design via Statistical Surrogate Models

• Activity Number: 92
• Type: Invited
• Date/Time: Sunday, August 4, 2013 : 8:30 PM to 10:30 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Fei Liu*+ and Rui Zhang and Angela Schoergendorfer and Youngdoek Hwang and Young Lee and Jane Snowdon
• Companies: IBM Watson Research Center and IBM Watson Research Center and IBM T.J. Watson Research Center and IBM Watson Research Center and IBM Watson Research Center and IBM Watson Research Center
• Keywords: Computer Experiment ; optimization ; sequential design

Abstract

Choosing the optimal combination of building components that minimize the investment and operational cost is a topic of great importance in the building simulation community. Energy simulation models, such as EnergyPlus, are widely accepted as powerful tools to assess the building energy demand. Traditional approaches for building design rely on exhaustive search of the input parameter space of the energy simulation models. The computational cost is the major challenge for such approaches. We propose a novel approach to address this problem. The key idea of the proposed approach is to first build a statistical surrogate model for the energy simulation model and then to update the surrogate model based on the concept of sequential design of experiments. We demonstrate the proposed approach using a case study of a live retrofit project for Building 661 at the Navy Yard of Philadelphia, USA, and show that the statistical surrogate model allows for fast evaluation of the building's energy consumption, and the sequential design reduces the computational cost by requiring a smaller number of runs of the energy simulation model.