Comparison of temperature models for bifacial photovoltaic modules in Germany and Brazil

Abstract

Accurate temperature modeling of bifacial photovoltaic (PV) modules is essential for precise energy yield prediction and system design optimization. This study compares the performance of three widely used steady-state thermal models — Ross, Faiman, and PVsyst — under two contrasting climatic conditions: Cologne, Germany (temperate oceanic), and Florianópolis, Brazil (humid subtropical). Identical bifacial glass–glass minimodules equipped with integrated thin-film thermocouples enabled direct cell temperature measurements, reducing uncertainties associated with conventional external surface sensors. Each model was evaluated using both literature-based and locally calibrated coefficients, with prediction accuracy assessed by root mean square error (RMSE) and mean error (ME). Results show that locally calibrated coefficients consistently outperformed literature defaults across all models and both locations, with RMSE values ranging from 2.06 K to 2.47 K after calibration. In Brazil, the PVsyst model with Uc=40 W/m²K achieved the lowest error (RMSE=2.30 K), while in Germany the Faiman model with locally derived coefficients yielded the best performance (RMSE=2.06 K). Despite the contrasting climates, calibrated model performance was comparable across both sites, suggesting that coefficient calibration is more influential than climatic region on prediction accuracy. This study further demonstrates that temperature sensor location systematically affects coefficient calibration: models calibrated on rear-glass surface measurements underestimated true cell temperature, increasing RMSE by up to 0.6 K compared to cell-surface calibration. These findings highlight the importance of direct cell temperature measurement and site-specific coefficient calibration for reliable thermal modeling of bifacial PV systems.