Energy-efficiency oriented occupancy space optimization in buildings: A data-driven approach based on multi-sensor fusion considering behavior-environment integration

Abstract

Buildings contribute significantly to global energy consumption. Optimizing internal building space layout is an essential approach for reducing energy consumption. However, proactively improving energy efficiency by building space design is still challenging requiring comprehensive consideration of complex interactions between indoor environment and occupant behavior, which is less studied previously. Considering behavior-environment integration, this study proposes a data-driven approach based on multi-sensor fusion for energy-efficiency oriented occupancy space optimization in buildings. Firstly, time series data including indoor environment and occupant behavior were collected based on multi-sensor fusion. Then, a data-integrated Convolutional Neural Network (CNN) model was developed for occupancy state classification. Based on obtained occupant schedules, space occupancy patterns of users were extracted using hierarchical clustering, and space optimization was further conducted for energy efficiency improvement. Finally, energy consumption was predicted with random forest regression after space optimization, and the impact of occupancy space optimization on energy efficiency can be evaluated. The proposed method was successfully applied in an academic office building on a campus in Wuhan, China, which helped achieve energy consumption reduction by 23.5 %. This study presents a promising path towards sustainable energy goals in building design, which serves as advanced guidance in the management of building energy performance.