Modelling Household Occupancy Profiles using Data Mining Clustering Techniques on Time Use Data

A strong correlation exists between occupant behaviour and energy demand in residential buildings. The choice of the most suitable occupancy model to be integrated in high temporal resolution energy demand simulations is heavily in uenced by the purpose of the building energy demand model and it is a tradeoff between complexity and accuracy. The current paper introduces a new occupancy model that produces multi-day occupancy profiles and can be adaptable to various occupancy scenarios (e.g., at home all day, mostly absent) and scalable to different population sizes. The methodology exploits data mining clustering techniques with Time Use Survey (TUS) data to produce realistic building occupancy patterns. The overall methodology can be subdivided into two steps: 1. Identification and grouping of households with similar daily occupancy profiles, using data mining clustering techniques; 2. Creation of probabilistic occupancy profiles using 'inverse function method'. The data from the model can be used as input to residential dwelling energy models that use occupancy time-series as inputs.