The temperature of photovoltaic (PV) cells is a critical factor in evaluating energy yield and predicting system degradation. Although thermo-electrical models allows predicting the evolution of the system over time, precise understanding of the thermal exchanges between the system and its environment is needed as they are implemented in the yield assessment using thermal correlations. These empirical correlations are based on heat transfer magnitudes undergone by similar PV set-ups. The aim of this study is to introduce a non-intrusive experimental methodology for precisely determining the convective heat transfer coefficient (CHTC) at the front of photovoltaic modules using two setups. The method integrates a heat flux sensor glued to the PV surface coupled with environmental data (e.g., irradiance, ambient temperature). This experimental method is applied to PV modules on a roof in an urban area and to a floating photovoltaic (FPV) system. It is demonstrated that the method significantly improves the accuracy of prediction of PV module temperatures in operating conditions compared to the conventional method based on the energy balance of a PV module. By using quantile regression, an empirical forced convection correlation is found based on the average wind speed. Compared to the traditional approach which relies on global transmittance, the CHTC is mainly dependent on the wind, whereas the global transmittance includes the radiative heat transfer which depends on the module temperature. The correlation for CHTC tailored for the floating photovoltaic system shows sensitivity to wind speed that is slightly higher compared to the inland setup in the literature.