Impact of Diffuse Radiation Modeling on Rear/Front Ratio Estimation for Bifacial PV Systems
Abstract
The accurate modeling of diffuse irradiance is essential for evaluating the rear/front ratio (R/F) in bifacial photovoltaic (PV) systems. This study investigates the performance of different diffuse separation and transposition models for front and rear irradiance estimation using field data collected in Recife, Brazil (8°S, 34°W). Two separation models (Erbs and Starke3) and two transposition models (Isotropic and HDKR) were tested in their four possible combinations. Results show that the Starke3 model consistently outperforms Erbs in estimating both global (GTI) and rear (RTI) tilted irradiances. When combined with the HDKR model, Starke3 achieved the lowest normalized mean absolute error for the rear tilted irradiance (13.8% for RTI) and the best agreement with observed R/F ratios. The HDKR model improved R/F estimations by reducing nRMSE from 24.2% to 19.9%. The study demonstrates that accurate diffuse modeling, especially under medium to high clearness index conditions, is crucial to properly quantify rear irradiance and consequently the bifacial gains. Neglecting anisotropy or horizon brightening can increase the R/F estimation errors, significantly affecting the bifacial gain estimation error in 40%.
Keywords
albedo, rear/front ratio, rear irradiance, bifacial gain, diffuse irradiance, GHI separation, transposition model