Improving Bifacial Photovoltaic Models by Quantifying the Impact of Racking, Artificial Reflectors, and Varying Solar Spectrum

Abstract

The deployment of photovoltaic (PV) energy is expected to rise exponentially in the next few decades, driven by decarbonization, electrification, and record-low system costs. Furthermore, bifacial PV modules, which generate electricity from both front and rear surfaces, are projected to make up 70\% of the PV market share within the next ten years due to their increased generation capacity, more efficient land use, and reduced levelized costs. However, irradiance incident on the module rear is more variable than front-incident irradiance, resulting in higher uncertainty in bifacial energy yield models than in monofacial models.

PV energy yield models inform a wide range of stakeholders, including policy makers, lenders, grid operators, system owners, and system designers. Energy yield predictions are subject to uncertainty caused by weather fluctuations, input uncertainty, and necessary approximations. Reductions in these uncertainties and approximations promote PV adoption, reduce financing costs, increase design efficiency, facilitate maintenance, and improve grid integration.

This thesis aims to improve bifacial PV models by quantifying the impact of system design and environmental conditions on energy yield. The work is focused on three factors: (1) structural shading and reflection, (2) artificial ground reflectors, and (3) solar spectral irradiance. I combine field measurements with model results to identify the conditions where these effects are most impactful and when they are negligible.

Firstly, I present a method for quantifying the effect of racking shading and reflection on the rear irradiance, energy yield, mismatch losses, and bifacial gain of a bifacial PV system. I demonstrate that: (1) racking reflection is a significant structural effect on PV performance, and (2) structural effects vary seasonally, by time of day, and with albedo. In an example case with two-in-portrait single-axis-tracked modules in Livermore, California, incorporating racking reflection reduces the average structural rear shading by up to 9.1% per year compared with absorptive racking structures.

Secondly, I quantify the impact of artificial reflectors with 70% reflectivity on a bifacial one-in-portrait single-axis-tracked PV system in Golden, Colorado. Relative to natural albedo, I found a reflector power gain of up to 6.2% before inverter clipping from field experiments and up to 4.5% after inverter clipping in energy yield models. I demonstrate that: (1) the ideal placement of the artificial reflector is directly under the torque tube, (2) clipping effects must be considered when optimizing artificial reflector systems, and (3) artificial reflectors are most viable in locations with low energy yield and high system costs.

Finally, I demonstrate the importance of spectrally-varying solar irradiance in bifacial PV energy yield. The solar spectrum is constantly changing, causing errors in PV models due to deviations from the reference spectrum. I quantify the spectral error in bifacial PV model predictions for a range of North American locations (39.7-69.1°N). I demonstrate: (1) annual spectral irradiance impacts on energy yield from +0.7% to +2.7% and (2) spectral effects are most important for diffuse irradiance and ground-reflected irradiance. I then compare five parameterized spectral correction methods and two measured spectral irradiance correction methods in Roskilde, Denmark. These demonstrate that long-wavelength irradiance cannot be neglected in PV spectral correction.

This work supports bifacial PV model improvement by quantifying the effects of various environmental and system factors with the goal of reducing bifacial PV model uncertainty and increasing PV adoption worldwide.