In this study, the outputs from a simple PV performance model were compared to measurements of AC power for three PV systems located across Sydney, Australia. The study aimed to investigate the level of uncertainty and bias of the model when onsite measurements of plane of array (POA) irradiance and module temperature were not available. The results demonstrated that the simple PV performance model estimated the AC performance with a low level of model bias (NBME = +/- 3.2%) and uncertainty (NRMSE < 6%) when onsite measurements of POA irradiance and module temperatures were available. For POA irradiance, the results indicated that modelling uncertainty increased significantly (NRMSE < 13%) when alternative methods to estimate POA irradiance were utilised. For module temperature, the results indicated that the choice of model coefficients had a significant impact on the performance of the module temperature models. In particular, for the three parallel roof mounted PV systems studied, the results suggested that the open rack/free standing or well ventilated module temperature coefficients should be used within the module temperature models investigated. This selection of coefficients was not directly evident given the PV systems investigated were parallel roof mounted PV systems, not free standing rack mounted arrays. (C) 2015 Elsevier Ltd. All rights reserved.