Data-driven approaches for analysis of building energy consumption and indoor occupancy behavior

Abstract

A recent surge of interest in building energy consumption has generated a tremendous amount of energy data, which boosts the data-driven algorithms for broad application throughout the building industry. In addition, occupancy behaviour is an important influencer of energy consumption in building. Currently the shallow understanding of occupancy has leaded to considerable performance gap between prediction and measurements of energy use. In this work, data-driven approach, mathematical approach and blind system identification model are developed to investigate building energy consumption and indoor occupancy behavior. As for the occupants’ “active” influence in building energy, we characterize residential appliance usage utilizing the K-means clustering approach through case studies and present the complex residential electricity behaviors in Shanghai. Similarly, the occupant’s “passive” role also has impact on the building performance. Among different passive occupant behaviors, window opening action and occupant profile have been deeply investigated in our research. Furthermore, in order to identify the impact of occupancy behaviors on building energy, prediction models based on the artificial neural network (ANN) are established to predict the electricity consumption of the air-conditioning system at the next time step, the superiority of the ANN model with the supplementary input of estimated current occupancy is verified by comparing the ANN model results without the input occupancy. In summary, the proposed approaches provide a new and detailed way for engineers and building operators to better understand occupant behaviors and their impacts on building performance. Therefore, dedicated energy-prediction models with consideration of occupancy provide an opportunity to couple the electric grid and the building’s control actions, and to be utilised by buildings and utility companies to simultaneously optimise their performance

Date of Award	8 Jul 2019
Original language	English
Awarding Institution	Univerisity of Nottingham
Supervisor	Liang Xia (Supervisor), Yupeng Wu (Supervisor), Fazel Khayatian (Supervisor) & Xingxing Zhang (Supervisor)