Verifying reliability of methodology for identifying changes in building energy consumption

Meiru Ou, Purdue University

Abstract

This thesis investigates the robustness of a statistical methodology for identifying significant changes in building energy consumption profiles. The method has its origins in classical change-point analysis techniques, which combines a cumulative sum of errors, bootstrapping, and mean squared error analysis; however, it has been appropriately modified to make it applicable for identifying changes in building energy consumption. The methodology is useful in identifying when a change has occurred in a building that impacts energy consumption, such as the implementation of an energy conservation measure or a change in building usage characteristics. The technique can be used by energy service companies engaged in performance contracting to statistically identify, with a given confidence interval, the timing and magnitude of a change in energy consumption that may be attributable to their efforts. This thesis presents the results of an extensive numerical testing effort to determine the robustness of the change-point analysis technique. The energy consumption datasets in this study are generated using Energyplus and are based on a typical light commercial building. The benefit of using building energy modeling software for this study is that it becomes possible to investigate the sensitivity of the change-point technique to different characteristics such as when the change occurred, the nature of the change, and the magnitude of the change.

Degree

M.S.

Advisors

Koch, Purdue University.

Subject Area

Architectural|Civil engineering|Energy

COinS