Z E M C H 2 0 1 2 I n t e r n a t i o n a l C o n f e r e n c e
528
On the other hand, active effects include the operation of control devices (e.g. windows,
ventilation system) or the use of electrical appliances (e.g. computers, washing
machines) (Mahdavi, 2011). Previously, specific user actions such as lighting control
(Reinhart, 2004) and window opening (Herkel, 2008) (Yun, 2009) were studied in
behavioural models, often defining ‘active’ or ‘passive’ user types to take into account
variations of user behaviour. Others have developed models based on time-use
research to study household occupancy, activities and energy use (Tanimoto, 2008)
(Richardson, 2010) (Widén, 2010) (Wilke, 2011).
In this paper, a model is presented that generates user profiles for residential buildings.
The model uses Belgian time-use data to generate household occupancy and activity
patterns that serve as a precursor for energy use modelling. The disaggregation of the
population into household types will allow us to study user behaviour more precisely.
Since household energy use depends largely on the presence of household members in
the building, respondent- and household types are defined based on the household
structure and the employment type of the adult(s).
Methodology
Time use research
We calibrated the model using a combined Belgian time-use survey (TUS) and
household budget survey (HBS) that was collected in 2005 by the research group TOR
of the University of Brussels and the statistics division of FPS Economy Belgium
(Glorieux, 2008). The combined surveys TUS and HBS include 6218 individuals from
3360 households. All household members over twelve years old were asked complete
the diary for the same two days, one weekday and one weekend day. In these diaries
the respondents described their activities, location and movements from 4:00 AM until
3:50 AM the next day. The Belgian diaries use a continuous registration system with
fixed time slots of 10 minutes. For each time slot, the respondent could enter up to two
activities, a primary activity and a secondary activity, described in their own words.
These self-described activities were recoded into 272 activity codes. In addition to the
time-use diaries, both individual and household questionnaires were completed. The
former included information about the age, position within the household, education,
income and employment, whilst the latter contain details about the family home, the
ownership of (electrical) appliances and vehicles, as well as their expenditures on goods
and services.
Literature
Several researchers have used time-use surveys to build models for user behaviour and
energy consumption. Even though the basic concept of these models is similar, some
important distinctions can be found in the algorithms.
Richardson developed an occupancy model using a two state non-homogenous Markov
Chain Monte Carlo (MCMC) technique, that generates time series by calculation the
transition probability at each time step based on statistical data. He made a distinction
between two states: active and inactive users (Richardson, 2008). Active users are
considered to be at home and awake, which is an important precondition for activities to
occur in the house. At each discrete time step, the probability that active users turn to
inactive in the current time step or vice versa depends on their previous state and the
time. These probabilities are stored in transition probability matrices. Richardson
combines this occupancy model with an activity model, that is derived from time-use
data as well, and a lighting model to predict electricity consumption (Richardson, 2009)
(Richardson, 2010).
