Abstract
A set of grey box models for a passive single family house is evaluated to understand the influence of inputs (solar radiation, internal gains and ventilation) knowledge. The models are evaluated with both a log likelihood ratio test and evaluation of the long-term prediction RMSE. It is shown that solar radiation is a necessary input, while internal gains and ventilation gain show more ambiguous results. It is also shown that a strict log likelihood evaluation based on 1-step predictions is not suited for evaluating models for applications where longer predictions are necessary.
