A Bayesian Approach for Predicting Building Cooling and Heating Consumption

This research proposes a Bayesian approach to include uncertainty that arises from modeling process and input values when predicting cooling and heating consumption in existing buildings. Our approach features Gaussian Process modeling. We present a case study of predicting energy use through a Gaussian Process and compare its accuracy with a Neural Network model. As an initial step of applying Gaussian Processes to uncertainty analysis of system operations, we evaluate the impact of uncertain air-handling unit (AHU) supply air temperature on energy consumption. We also explore the application of Bayesian analysis to building energy diagnosis and fault detection. In concluding remarks, we briefly discuss advantages of the proposed approach.