Representative weather information is essential for a reliable building energy performance evaluation. Even if detailed energy analyses can be carried out considering the multi-year weather data, generally a single reference year is adopted. Thus, this artificial year has to correctly approximate the typical multi-year conditions. In this work, we investigate the representativeness of the method described in the technical standard EN ISO 15927-4:2005 for the development of reference years. Energy performance of a set of different simplified buildings is simulated for 5 north Italy locations using TRNSYS. The energy needs computed using the reference year are compared to those of a multi-year simulation. The annual variability of energy results for the studied thermal zones is investigated, paying attention to its effects on the building envelope energy ratings according to a proposed classification. Also, those configurations more influenced by the annual weather changes are identified by means of statistical indexes. The analyses demonstrate that the representativeness of the reference year results can vary significantly in the considered locations and, consequently, the accuracy in building energy assessment and classification can be reduced, especially for some building envelope configurations. (C) 2013 Elsevier B.V. All rights reserved.