In certain applications of building energy simulation, the normally used time-step of one hour (1 h) can be a source of significant error. One such application is the matching of PV production and household energy demand. Short-time peaks in PV generation and demand load produce errors in the matching results. The goal of the investigation is to show the reasons for these errors and to find out how large the errors are. The approach is partly theoretical and partly based on simulation exercises with TRNSYS using a time-resolution ranging from 1 min to 1 h. The amount of matching is expressed using the on-site energy fraction index (OEF) and the on-site energy matching index (OEM). Detailed parametric analyses are conducted with the focused parameters on the size of the generation system and the error of daily matching results with coarser resolutions compared to 1-min resolution. The results show that the shape of the demand curve as well as the production curve, and the interconnection of these curves, is a crucial factor explaining the formation of errors. When the high-resolution generation curve frequently crosses the intermittent long spikes or the continuous saw teeth of the demand curve, noticeable errors (in some cases higher than 60%) are generated with 1-h resolution compared to 1-min resolution. Moreover, for the on-site PV during the summer-time, noticeable errors (in some cases 15% or higher with 1-h resolution) are more likely to happen under scattered cloud conditions compared with clear or overcast skies. (C) 2014 Elsevier Ltd. All rights reserved.