Photovoltaic (PV) performance predictions are important to accurately assess the efficiency of any PV technology. In this study, we confront outdoor data with no less than 48 couples obtained by combining eight models of the thermal behavior with six electrical formulas. Calculations are confronted to the power produced by a 2 kWp grid-connected monocrystalline Si photovoltaic plant (GCPV) installed on the rooftop in the Faculty of Science Semlalia Marrakech, Morocco (latitude 31.6497 °N, longitude 8.0169 °W). The measured meteorological parameters (irradiance and air temperature), electrical data (DC power), and modules temperature data from one year have been used. The approach to evaluate the quality of each couple of models is new since this work uses the combination of (i) the best mix of correlation coefficient (R²) and root mean square error (RMSE), and (ii) the number of points validated by the model within a 99% confidence interval. Among the eight thermal behavior models, we propose ourselves a dynamic one which takes into consideration inertia which is usually ignored in stationary models.Keywords Grid-connected PV, PV DC-power measurements, Module temperature measurements, PV DC-power models, Module temperature models, Combinations of DC-power and module temperature models.