Paper Conference

Proceedings of SimBuild Conference 2012: 5th conference of IBPSA USA


Creating Zoning Approximations to Building Energy Models Using the Koopman Operator

Michael Georgescu, Bryan Eisenhower, Igor Mezić
Department of Mechanical Engineering, University of California, Santa Barbara

Abstract: As the scope of building construction increases and designs become more integrated, building energy models have found widespread use in evaluating building performance. Despite the growing sophistication of building modelling tools, errors can arise from the approximations that are made during model creation. This paper addresses model zoning, i.e., how the volume of a building is divided into regions where properties are assumed to be uniform. Zoning is important during the creation of a model because the accuracy of prediction from simulating a model reduces when dissimilar zones are lumped together. In this paper, a systematic approach to creating zoning approximations is introduced to investigate the effect of zoning on simulation accuracy. Applying the Koopman operator, an infinite-dimensional, linear operator that captures nonlinear, finite-dimensional dynamics without linearization, a detailed building model is studied. Using the Koopman operator, the temperature history of rooms produced by a building simulation can be decomposed into Koopman modes. These modes identify dynamically significant behavior which will form a basis for the creation of zoning approximations. An implementation of this technique is illustrated in a building model of an actual building designed with both mechanical and natural conditioning.
Pages: 40 - 47