Floating PV model reaches 92.3% prediction accuracy

In Malaysia, researchers collected one-minute data from two custom-built floating PV systems. Their 100 W modules sat 250 mm and 800 mm above the water, generating the measurements Curtin University’s team used to build a new model for estimating solar-cell temperature in floating photovoltaic systems.

The model extends the familiar nominal operating cell temperature, or NOCT, method. It adds water temperature to the calculation, alongside the conditions already used by the standard approach. Because it remains compatible with existing PV temperature models, it could be integrated into current solar software without replacing the underlying framework.

The team combined field measurements, computational fluid dynamics and statistical analysis. It developed several candidates, including versions with seven or 10 environmental and design factors, then compared them with the Malaysian measurements. The strongest overall performer was the basic FPV-NOCT model, which was also tested against independent floating-PV datasets from Passaúna Lake in Brazil and Windsor and Oakville in California.

The results varied by dataset. At Passaúna Lake, the basic model predicted observed cell temperatures for 11 of 12 months; with a wind-correction factor, it covered all 12 months. In one Windsor and Oakville validation, the model reached 92.3% prediction accuracy and performed better than the standard NOCT model for solar systems floating on water.

What does it change? Operators and software developers could get a simpler way to estimate how floating arrays heat up, using water temperature as the extra input. That could simplify integration into existing solar software, but the result is still a research model: Ramanan Chidambaram Jayaraj says he is refining its accuracy, and no commercial deployment is reported.