Performance evaluation of forecasting strategies for building occupancy prediction

Abstract

Occupant behavior has been identified as a key factor affecting energy usage in buildings. Integrating occupancy data into HVAC control strategies presents an opportunity for substantial energy savings. The proposed study evaluates different occupancy prediction strategies with a focus on forecasting performance on highly variable signals such as CO2 concentration and noise levels. Our work compares single-step and multiple-steps prediction methods to analyze their impact on accuracy and reliability. The predicted signals can be used to identify future activity to improve occupancy forecasting. In this paper, we highlight the importance of accurate occupancy data and fitting forecasting strategy and propose future research directions to address current limitations in occupancy prediction models.