The main objective of this research is to define a method to assess reliability in energy simulations for underground cellar modelling. Main energy programs do not guarantee thermal precision in heat exchange involving building and the surrounding ground. A thermal survey carried out in wine growing and producing farms allowed to collect inside temperatures in sampled buildings including an underground cellar. Data have been collected throughout one year and used as reference for the modelling assessment. The method is based on the comparison between temperature recorded in the case study and those provided by its model simulation. The model is considered reliable if both fits the recorded thermal behaviour of the case study, and provides acceptable results according to the target precision of the research. The results obtained in this paper show the method can improve building temperature trend previsions, therefore can be considered as a reliable tool for energy-efficiency-oriented building and system design. (C) 2014 Elsevier B.V. All rights reserved.