Building integration of photovoltaic systems is in continuous growth due to the high requirements for building energy efficiency. Some of the integration strategies, however, change the boundary conditions in which the PV systems work. Indeed, when embedded in the fa ade or in shading devices elements, the effects of partial shading and actual thermal conditions on power production need to be taken into account. In this work, suitable PV models to estimate the performance of such systems and to analyse the effects of partial shading are introduced and validated against measurements. In order to model the whole system, a detailed optical-thermal electrical model is developed. The electrical model is based on a five parameter PV cells equivalent circuit and it is coupled to a dynamic thermal model and to a specific irradiance calculation method. At the same time, a simplified model is developed to reduce the computational time. A dedicated experimental apparatus, built to reproduce the typical mutual shading conditions that might occur in shading devices, has been used to assess the accuracy of the model predictions - in terms of array voltage, current and cell temperature - and to analyse the effects of mismatching on the shaded cells. The results show that both the detailed and the simplified models predict the electric power production with good accuracies, with daily averaged relative errors between 2% and 10% and with the main deviations occurring during severe mismatching conditions. In such conditions the accuracy of the simulations can be significantly improved using the detailed model and shortening the simulation time-step. Indeed, a model sensitivity analysis has been performed, showing that the detailed model represents adequately the physical response of the PV system to variable environmental conditions, but the results of the simulations are very sensitive to the input data (geometrical tolerances and time-step) during partial shadings. Concerning the thermal model, both stationary and dynamic approaches provide temperature predictions accurate enough for daily power estimations. Nevertheless, the dynamic approach is able to better describe the cell temperature and is necessary for time-steps below the characteristic time of the system. Thus, the detailed optical-thermal-electrical PV model proposed demonstrated to be a valuable tool to investigate the effects of building integration of PV systems, allowing designers to evaluate electrical and thermal effects (e.g. power losses, hot spots). It can also be used for the assessment of simplified models. The simplified model demonstrated to provide comparable results - while requiring less computational effort- respect to the detailed model, thus being suitable to be integrated in long term simulations (i.e. building simulations for early performance evaluation).