Paper Conference

ASHRAE & IBPSA-USA SimBuild 2016: Building Performance Modeling Conference

     

A COMPARISON OF GAUSSIAN PROCESS REGRESSION AND CHANGE-POINT REGRESSION FOR THE BASELINE MODEL IN INDUSTRIAL FACILITIES

Joseph Carpenter, Keith Woodbury, Zheng O'neill

Abstract: This study compares the effectiveness of Gaussian Process (GP) models to three-parameter cooling change point (3PC) models for establishing baseline energy usage models in industrial facilities using utility bill data. Several different methods of creating baseline models for commercial and residential buildings have been developed; however, few attempts have been made to create baseline energy models in industrial facilities. Industrial facilities account for 33% of annual energy usage within the United States, so industrial energy usage needs to be analyzed in order to identify energy saving opportunities. Creating a baseline energy model is important to understanding an industrial facility's energy usage. An analysis of the effectiveness of using GP regression to develop a baseline energy usage models in industrial facilities from utility bill data and ambient outdoor dry bulb temperature is presented. Two case studies are presented: using utility bill data and average monthly temperatures to create a GP regression model and 3PC model. In both cases the baseline regression models gave a CV-RMSE of 15% or lower and NMBE of 5% or lower showing that either a GP regression model or 3PC model using utility bill data is capable of producing acceptable baseline energy models by ASHRAE Guidelines. In both cases GP regression models had slightly lower CV-RMSE values than 3PC models.
Pages: 79 - 86
Paper:
simbuild2016_C011