Estimating Building Electricity Performance Gaps with Internet of Things Data Using Bayesian Multilevel Additive Modeling

Abstract

Energy models should be simplified to handle data limitations and should predict reliable energy use. Currently, it remains challenging to ensure an appropriate level of detail for simplifying building energy models and to avoid performance gaps when predicting electricity consumption. In this respect, this research proposes to identify an appropriate level of simplifying a building energy model, predict electricity demands and performance gaps using the simplified energy model, and expand the model usability through the operational stage. Building electricity demands predicted through EnergyPlus (version 8.7.0) simulation are compared with actual electricity data collected through Internet of Things (IoT) sensors. Consideration of performance gaps increases the predictability of electricity consumption of a simplified energy model. Also, the Bayesian multilevel additive model updates the performance gaps along with the collection of new IoT data. The findings of this study contribute to forecasting electricity demands with a simplified energy model by predicting performance gaps that can be applied to predicting the electricity needs of similar buildings in the design stage and controlling operational electricity use in the operational stage by comparing sensor measurement with reference data provided by the energy model.