Improving transient thermal simulations of single dwellings using interpolated weather data

The lack of studies on the interpolation of weather data to calibrate building thermal simulations means that the skills of different interpolation techniques are unknown. This study evaluates the performance of five interpolation techniques in reproducing on-site weather data for building thermal simulations. To this end, 18 weather stations of a government weather agency spread over 4,495 km(2) of a northwest Spanish province were used. A representative building was chosen for running transient thermal simulations with the software TRNSYS. The interpolation technique effectiveness was tested and evaluated for one year on an hourly basis. The results of the method comparison show that not all weather variables have the same influence on the results of the thermal simulation: global radiation had the strongest influence on the simulation results. While thin plate splines (TPS) might be the best choice for generating weather data files, Universal Kriging (UK) is better for the simulation results. The TRNSYS models using weather data reduce the Coefficient of Variation of the Mean-Squared Error (CVRMSE) value from more than 18% (using the Nearest Weather Station) to below 3% (using Universal Kriging). Moreover, based on the Mean Bias Error (MBE) values, the thermal simulation results indicate that the interpolation techniques tend to overestimate heating demands. (C) 2016 Elsevier B.V. All rights reserved.