Thermal performance and temperature correction of floating photovoltaic arrays based on high-resolution field measurements

Floating photovoltaic (FPV) deployment is expanding rapidly, but reliable thermal and yield assessment remains limited by scarce long-duration, spatially resolved temperature data. Using the High-Resolution Floating Solar PV Data, multi-year, 1-minute measurements with fifteen module-level sensors per floating array and a co-located land-based reference, this study characterises FPV thermal behaviour. Multi-point sensing reveals persistent intra-array thermal heterogeneity (σ = 1.5–3.5°C), indicating that single-point measurements misrepresent array-average conditions. FPV modules operate marginally cooler than the land reference (mean ΔT≈−0.8°C) under humid subtropical conditions; the differential varies non-monotonically with irradiance, attenuating to near-zero at 600–800 W m−2 and reaching maximum cooling in the highest-irradiance bin. A steady-state Faiman formulation yields site-specific heat-loss coefficients but exhibits substantial minute-scale scatter (FPV R2≈0.13). An empirical four-input correction model (irradiance, wind speed, ambient temperature, shallow water temperature) predicts FPV operating temperature with R2 = 0.69 and RMSE≈6.3°C. The water-temperature term is statistically and physically significant; air-temperature and wind-speed dependencies act largely as proxies for direct water-thermal coupling in humid-subtropical conditions. The cooling corresponds to ≈0.3–0.6% reduction in temperature-related efficiency losses, supporting improved yield modelling and bankable assessment.