Modeling energy consumption in residential buildings: A bottom-up analysis based on occupant behavior pattern clustering and stochastic simulation

• Published in: Energy and buildings Vol. 147; pp. 47 - 66
• Main Authors: Diao, Longquan, Sun, Yongjun, Chen, Zejun, Chen, Jiayu
• Format: Journal Article
• Language: English
• Published: Lausanne Elsevier B.V 15.07.2017; Elsevier BV
• Subjects: Air conditioners; Air conditioning; Behavior; Bottom-up model; Building energy simulation; Buildings; Classification; Clustering; Computer simulation; Demographics; Energy consumption; Engineering; Housing; Mathematical models; Neural networks; Occupant behavior; Occupations; Pattern clustering; Refrigeration; Residential areas; Residential buildings; Residential energy; Schedules; Stochasticity; Occupant behavior; Bottom-up model; Building energy simulation; Pattern clustering
• ISSN: 0378-7788; 1872-6178
• DOI: 10.1016/j.enbuild.2017.04.072

•Occupant behaviors are clustered for residential buildings energy consumption estimation.•Clustered occupancy schedules are distinctive to ASHRAE schedule.•Bottom-up and simulation models were utilized to validate the results.•Applied the reliable ATUS and RECS datasets. In residential buildings, occupant behavior and occupancy status have a significant impact on energy consumption variation. Although the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) recommends a uniform occupancy schedule for building energy assessment, occupant behavior patterns and schedules could be different for each building due to occupants’ lifestyles, preferences, occupations, and other differences. Existing occupant behavior models focus on analyzing occupants’ sociodemographic characteristics to predict their energy consumption with statistical approaches. This paper proposes to identify and classify occupants’ behavior with direct energy consumption outcomes and energy time use data through unsupervised clustering. Based on the American Time Use Survey (ATUS), the proposed approach integrates k-modes clustering and demographic-based probability neural networks and identifies 10 distinctive behavior patterns. With the results of the behavior classification and simulation, a bottom-up engineering model reveals that the proposed behavior model offers a more accurate and reliable prediction than the ASHRAE standard schedule. With qualified and sufficient time use data, the model is capable of automatically estimating energy consumption on even larger geographic scales.