Analysis and Modeling of Mechanical Ventilation Operation Behaviors of Occupants in Cold Regions of North China

Abstract

Mechanical ventilation has a great impact on building simulation performance, such as indoor environment quality and building energy consumption. However, there is still a lack of accurate mechanical ventilation models established from long-term field data that can effectively predict building performance. In this study, one-year measurements on mechanical ventilation operation behavior were collected from 85 apartments, which were conducted with a mechanical ventilation system of the same brand in cold regions of North China. This permitted statistical analysis and clustering of the mechanical ventilation operation behavior by using the K-means method, leading to five behavior patterns. The results showed that 24% households operated mechanical ventilation system nearly all day, and there was a large difference in usage behaviors between the split system and the centralized system. Furthermore, two classes of models based on random forest and logistic regression were developed for predicting mechanical ventilation system operation (on/off) behavior. The models based on random forest showed high accuracy as it resulted in a 0.992 average in predictions. These models using field data can guide the selection of accurate input boundary conditions of mechanical ventilation and improve the accuracy of dwelling numerical simulations.

Keywords

• mechanical ventilation operation behavior
• machine learning
• predictive model
• residential building