Research on Behavior Pattern Prediction at Early Stage of Design
- 238 Downloads
Occupant behavior has great impact on building energy consumption. Lots of efforts have been contributed to bridging the gap between simulated energy performance and the reality. Nevertheless, the prediction of potential behavior pattern in a to-be-built building is lack of researches. This study presented an approach to predict the probability of certain behavior pattern basing on available inputs at design stage. Five hundred and forty-six questionnaires about energy-related behavior preference were collected by online survey, from which the control pattern of air conditioner was analyzed and used as data source of this research. A feed forward neural network model with two hidden layers was built and trained to calculate the probability of certain behavior pattern, where the pattern “turning on air conditioner when feel hot” was taken as an example. As result of this research, the trained ANN model was cross-tested by 100 randomly selected “testing group” and reported average correct rate of 61%.
KeywordsBehavior pattern Prediction method ANN model Questionnaire survey Design stage
Project Prediction method of building energy consumption in urban scale based on DEM (Digital Elevation Model) and occupant behavior research was supported by the National Natural Science Foundation of China (Grant No. 51708324).
Permissions Appropriate permissions from participants of above-mentioned questionnaire survey were obtained.