| Résumé | This paper presents a proof-of-concept of an inverse model-based method for the characterization of envelope thermal transmittance and air permeance with historical space heating load and building automation system trend data. The method is demonstrated with hourly heating, indoor temperature and CO2, and outdoor temperature and global horizontal solar radiation data from two buildings in Ottawa, Canada. Of them, one is a highly instrumented recently constructed academic office building, and the other one is a 40-year-old government office building. The method first identifies the parameters of a steady-periodic weekly ventilation schedule. Then, the air permeance is estimated by modelling the decrease rate of the average indoor CO2 concentration once the air handling units (AHUs) switch to unoccupied mode. Lastly, a steady-state, a single-node transient, and a multi-node transient change point model are trained and compared to estimate the thermal transmittance. Cluster analysis is employed on indoor temperature data for dimensionality reduction prior to training the multi-node transient change point model. The results indicate that transient models offer a modest improvement in the CV(RMSE). All three model forms generate consistent thermal transmittance estimates. The thermal transmittance estimates of the recently constructed building are in agreement with those calculated by a physics-based model based on as-built drawings. |
|---|
