Data preparation to address heterogeneity in time use data based activity modelling

Abstract

Stochastic person-based activity models play an important role in the prediction of the realistic time-series energy demand for residential buildings. These models generally use input parameters developed based on time-use data. This paper evaluates how the adopted data preparation approach alters the variability in simulated activities among households. Four simulation cases were developed in the study, representing 1) clustering, 2) regression and 3) integrated approaches combining the first two. A comparison of the results indicates the integrated approach to be the most advantageous. Finally, a strategy to enhance heterogeneity in the simulated household activity for residential energy demand modelling is discussed.