Application of BIM-DesignBuilder and K-means clustering-random forest for prediction of building energy consumption

The rapid development of the construction industry has significantly increased the proportion of building energy consumption in total energy use, resulting in a large amount of resource waste and environmental pollution. In order to be consistent with the goal of sustainable social development, it is crucial to accurately predict the energy consumption of buildings. This paper first establishes a building model and conducts energy consumption simulation analysis based on DesignBuilder (DB). Then, a building energy consumption (BEC) classification method is proposed based on K-means clustering. Finally, random forest (RF) is used to predict the three groups of data. We take a teaching building as an example for research. The data is divided into three groups of low energy consumption, medium energy consumption and high energy consumption by K-means clustering, and the RF model is applied to each group for prediction. The results show that: (1) The relative error of DB simulation based on BIM data is within the range of [-5%, 5%] per month, which is consistent with the actual value. (2) K-means clustering can accurately classify the samples while retaining the monthly differences in energy characteristics. (3) Compared with machine learning models such as multilayer perceptron (MLP), decision tree, XGBoost, etc., the RF prediction model based on clustering results performs better. These methods can effectively capture the energy consumption patterns under different energy consumption levels and achieve precise predictions. This case study has demonstrated the outstanding performance of this method and provided new strategies for improving building energy efficiency and sustainability.