Improved Data-Driven Building Daily Energy Consumption Prediction Models Based on Balance Point Temperature

The data-driven models have been widely used in building energy analysis due to their outstanding performance. input variables of the are crucial for predictive Therefore, it is meaningful explore that can improve performance, especially context global crisis. In this study, an algorithm calculating balance point temperature was proposed apartment community Xiamen, China. It found label (BPT label) significantly daily consumption prediction accuracy five (BPNN, SVR, RF, LASSO, and KNN). Feature importance showed BPT accounts 25%. Among all variables, minimum decisive factor affects consumption, while maximum has little impact. addition, study also provides recommendations selecting these model tools under different data conditions: when variable insufficient, KNN best BPNN sufficient.