Floating Photovoltaic (PV) system is an emerging and rapidly developing solar PV application that utilizes water surfaces, such as reservoirs and lakes, as the installation grounds for the PV arrays and its balance of systems. In addition to becoming a solution to land scarcity issues of solar PV installation, floating PV systems also benefit from an improved thermal performance due to its proximity to water that results in a cooling effect for the PV modules. However, the extent of this cooling effect is not yet well-understood and established. Conventional thermal modelling tools and approaches, which are mostly validated for land-based PV systems, might not be directly applicable to floating PV systems without proper adaptation. This research aims to improve the accuracy of temperature and energy yield predictions for floating PV systems by evaluating and updating physics-based thermal models based on both analytical and computational methods. This study is driven by several research questions, which includes the analysis on how do existing thermal models perform in predicting floating PV module temperatures, how can Computational Fluid Dynamics (CFD) simulations be used to improve thermal models accuracy for floating PV applications, and how does the improvement in thermal model accuracy impact energy yield estimation in PV system energy yield simulation tools.

Chapter 2 conducts an evaluation of two analytical thermal modelling approaches for predicting floating PV module temperature, which are the Fuentes Model and the Resistive Thermal Model. Results show that while both models reasonably capture the PV module temperature trends, they tend to under-predict water-induced cooling effect that is present in floating PV environments, which results in a overestimation of PV module temperatures in general. Even though the Resistive Thermal Model results in a slightly higher accuracy compared to the Fuentes Model, a notable Root Mean Square Error (RMSE) is still observed with respect to measured values. This discrepancy highlights the need to improve the thermal dynamics representation, especially in capturing the effect of water proximity and the modified convective heat transfer interactions unique to floating PV systems. Chapter 3 implements CFD as a refinement tool for the analytical thermal modelling, with the goal of combining CFD's ability to capture detailed solid-fluid interactions and the computational cost effectiveness of the Resistive Thermal Model. The CFD-Updated Resistive Thermal Model is compared against measured data and shows a substantial RMSE reduction, with an RMSE of 0.72oC, significantly lower than the RMSE of the initial Resistive Thermal Model (1.70oC) and the Fuentes Model (2.30oC). Chapter 4 analyses how the improved thermal models affect energy yield prediction by using two simulation tools PVsyst and the PVMD Toolbox. The energy yield results show that the CFD-updated Resistive Thermal Model achieves the closest match to the measured specific energy yield, with an error of just 0.10%, compared to 1.75% and 1.27% error values for the Fuentes FD Model and the initial Resistive Thermal Model, respectively. These results clearly show that the increase in thermal model accuracy is directly related to a higher accuracy in energy yield simulation, where the energy yield simulation based on the improved thermal model results in a very close agreement with measured values. Chapter 5 focuses analysing how does practical installation scenario, specifically differing PV array row spacing, affect the thermal dynamics of floating PV systems. The results demonstrate that while close-proximity PV arrays with small row-spacing may offer higher power density per unit area, it comes at the cost of increased PV module temperatures and potentially reduced energy yield. On the contrary, large-proximity PV arrays with large row-spacing results in a lower overall PV module temperatures that potentially increase energy yield of the floating PV system.

In conclusion, this study demonstrates the important role of thermal modelling in improving floating PV performance prediction for floating PV systems. The integration of CFD-derived parameters into Resistive Thermal Model leads to significantly improved accuracy in temperature and energy yield predictions. However, the findings are based on limited real-world measurement data and a single floating PV archetype. A more extensive validation across more configurations and climates, as well as the development of generalized Nusselt number correlations for diverse floating PV archetypes are recommended for future works.