Mathematical, empirical, and electrical models have long been implemented and used to predict the energy yield of many photovoltaic (PV) technologies. The purpose of this paper is to compare the annual DC energy yield prediction errors of four models namely the single-point efficiency, single-point efficiency with temperature correction, the Photovoltaic for Utility-Scale Applications (PVUSA), and the one-diode model, against outdoor measurements for different grid-connected PV systems in Cyprus over a 4-year evaluation period. The different models showed a wide variation of prediction errors, demonstrating a strong dependence between model performance and the different technologies. In particular, it was clearly shown that the application of temperature loss correction based on the manufacturer's temperature coefficients of power at maximum power point assisted in improving the energy yield prediction significantly especially for the crystalline silicon (c-Si) technologies. In most cases, the best agreement between the modeled results and outdoor-measured annual DC energy yield for mono-crystalline silicon (mono-c-Si) and multi-crystalline silicon (multi-c-Si) technologies was obtained using the one-diode model. The energy yield for the thin-film technologies was more accurately predicted using the PVUSA model with the exception of the copper-indium-gallium-diselenide (CIGS) technology, which was best predicted using the single-point efficiency with temperature correction and one-diode models, thus demonstrating similar physical properties to c-Si technologies. The paper further quantifies the combined uncertainties associated with the predicted energy yield as a function of the input parameters for the single-point efficiency, single-point efficiency with temperature correction, and the PVUSA models. Copyright © 2011 John Wiley & Sons, Ltd.