FIELD VALIDATION OF VIEW FACTOR AND RAYTRACING IRRADIANCE MODELS FOR BIFACIAL PHOTOVOLTAIC SYSTEMS INSTALLED IN A DESERT CLIMATE

Abstract

Solar photovoltaics (PV) are the fastest growing energy source. As of 2025, PV accounts for 2.26 TW of the ~4.5 TW global renewable capacity and supplies 10.8% of global electricity. Bifacial modules, which convert sunlight on both front and rear surfaces, now represent 64% of global PV manufacturing. This thesis presents a field validation of view factor (VF) and raytracing (RT) bifacial PV irradiance models in a desert environment, comparing front and rear plane-of-array (GPOA) irradiances ({G_{front}\ and\ G}_{rear}) from bifacialvf and bifacial_radiance against measurements from three PV configurations - fixed tilt (FT) south facing, vertical east-west (V-EW) and horizontal single axis tracking (HSAT)- installed at QEERI’s outdoor testing facility (OTF). Using one year of field-measured front and rear GPOA data, VF and RT model accuracies are evaluated via rMBE and rRMSE, while the impact of irradiance source on energy yield modelling is assessed using a single-diode model. For the FT system, both models simulate G_{front} accurately (rRMSE 6.2% for RT and 7.0% for VF), while RT yields lower rear GPOA error (15.5% vs. 25.7%). As rear irradiance contributes only ~25.3% of the total, combined (front + rear) modelled irradiance rRMSE remains low (6.9% for RT and 8.7% for VF). For the V-EW system, the VF and RT models overestimate irradiance on the east plane by 8.3% and 6.66%, respectively, and have similar rRMSEs (15.6% – 16.1%). Whereas RT modelled west irradiance with ~5% lower rRMSE and less bias, for the VF system, opposing east/west biases partially cancel when combined, resulting in similar combined irradiance rRMSEs (~9.5–9.6%). For the HSAT system, RT outperforms VF for total irradiance (rRMSE 6.0% vs. 9.0%). Annual irradiation prediction errors are considerably lower than point-in-time rRMSEs: for FT, annual error is −5.95% (VF) vs. +0.56% (RT); for V-EW, +0.57% (VF) vs. +4.34% (RT); for HSAT, +1.44% (VF) vs. −1.25% (RT) Using measured irradiances and temperatures, the single-diode model predicts yield within -1.20% to +1.60% of measured energy yield for the three PV configurations. Substituting RT and VF modelled irradiance timeseries results in errors of -2.02% to -0.80% and -7.53% to +1.75%, respectively. RT's higher accuracy comes at a computational cost: 7–24 hours per year compared to 2–3 minutes for VF. Replacing ground-measured GHI/DHI/DNI with CAMS satellite data (rRMSE 48.9%/35.4% for DHI/DNI) degrades total GPOA rRMSEs to ~20% and increases yield errors across all systems.

Date of Award	2026
Original language	American English
Awarding Institution	HBKU College of Science and Engineering