LOCAL WEATHER CORRELATIONS AND BIAS IN BUILDING PARAMETER ESTIMATES FROM ENERGY-SIGNATURE MODELS

Synthetic data of building energy consumption are generated by the simulation program ESP with the aim of analyzing the bias in building parameter estimations from energy signature (ES) models. Building parameters are determined by simulation under steady-state weather conditions. The simplest ES model with only outside and inside temperature variables leads to large errors in building parameter estimates for any time-scale of analysis, even when the associated coefficient of determination R2 is high. A short study of the residual of the energy consumption from this model, using the regression of the local solar radiation versus outside temperature, shows the importance of including the solar radiation variable in the model. Such a model leads to excellent estimates of the parameters with weekly data. The bias P(LI) of the loss coefficient parameter, as defined by Hammarsten from one model to the other, is the coefficient of the linear regression between local solar radiation and ambient temperature variables. The second bias P(AI) concerning the constant term is also explained in terms of the parameters of the local weather correlation. These theoretical results, verified with seven different simulated houses, open a new possibility for energy-saving studies in the use of ES models.