Energy performance of residential buildings at district level from data perspective

Abstract

Energy performance becomes more and more important during the urban planning/design process. There are two fundamental methods used to predict and analyze the residential building stock: top-down and bottom-up. However, both methods face the issues from data source and analysis method. While urban planners/designers manage to handle complicated design problems, the huge increase in data from sensors and simulations does not help to reduce the burden of designers. On the contrary, unfamiliar data sets can bring designers into a hopeless tangle. This research proposes a method using data mining techniques (k-means and S3VM) to address such issues and enhance the design process. The problem statement will be discussed both from design and data perspectives. A scenario analysis is proposed to evaluate the new methods to predict the electricity consumption of public housing in Singapore. The low difference proves the feasibility and possibility to apply the familiar and unfamiliar data concept into the design process in the future.