Enhanced thermal models of photovoltaic modules by electrical operating conditions dependency

The increasing penetration of photovoltaic (PV) systems poses challenges to the reliability and adequacy of power systems. To support grid stability, PV systems must evolve to be capable of providing frequency regulation and reserve services—including not only down frequency reserve but also up reserve. This latter service requires PV modules to operate away from their maximum power point (MPP), a condition that requires an enhancement in PV module thermal behavior assessment. Consequently, there is a growing need for advanced thermal models that account for electrical operating conditions to ensure accurate temperature prediction under all operating scenarios. While traditional thermal models primarily depend on meteorological inputs, they typically neglect the Electrical Operating Status (EOS). Overlooking this issue can lead to significant prediction errors—up to 5–7 °C—especially during operation away from MPP. The proposed investigation developed an enhanced thermal model incorporating EOS dependency by including the ratio of measured current to the calculated current at MPP as an additional input. Two cases of the Faiman and Sandia models were optimized using Genetic Algorithm, Particle Swarm Optimization, non-linear least squares, and polynomial regression. Optimization is performed using three identical PV systems operating under reference EOS conditions: open circuit, short circuit, and MPP. Results demonstrate that EOS-integrated models significantly improve temperature prediction accuracy. The EOS sensitive models achieved prediction errors as low as 0.1–1.13 % and R 2 values above 0.91, outperforming traditional models that exhibited errors from 2 to 29 %. These findings support the need for EOS-aware thermal modelling in modern PV system design and operation. • Electrical operating status sensitivity improved the accuracy of the thermal model predictions. • Modified Faiman and Sandia models performed similarly across optimization methods. • Integrating the electrical operating condition into the thermal model with NLLS yielded the highest accuracy. • Modified Sandia model predicted module temperature with less than 1 % error.