Projects

With a variety of factors for building energy consumption, the prediction accuracy largely depends on the data integrity of building basic information and measured energy consumption. In view of the fact that practical building energy performance related data are general lacking, how to sufficiently use limited available data to achieve reasonable and accurate building energy prediction? To answer this urgent question, it is crucial to fundamentally explore the indispensable variables for building energy prediction models. This study creatively combined the space filling design and feature dimension reduction method with current building energy prediction models. By means of feature analysis of high-dimensional data space formed by a designed building performance database, we deeply explore the complex mapping relation between building factors and building energy consumption and the further feature selection for building energy prediction. Using office building as the research object, this study will achieve a construction method of both building performance database and minimum variables set for building energy prediction models. The findings will hopefully solve the problem of using limited data condition to quickly achieve reasonable and accurate building energy prediction. They can be utilized as the theoretical support and data basis for building energy prediction, optimization design and benchmark evaluation of office buildings.