SYSTEM-LEVEL PERFORMANCE AND RELIABILITY OF SOLAR PHOTOVOLTAIC FARMS: LOOKING AHEAD AND BACK

In a world of ever-increasing demand for energy while preventing adverse effects of climate

change, renewable energy has been sought after as a sustainable solution. To this end,

the last couple of decades have seen an advancement in research and development of solar

photovoltaic (PV) technology by leaps and bounds. This has led to a steady improvement

in the cost-effectiveness of solar PV as compared to the traditional sources of energy, e.g.,

fossil fuels as well as contemporary renewable energy sources such as wind and hydropower.

To further decrease the levelized cost of energy (LCOE) of solar PV, new materials and

technologies are being investigated and subsequently deployed as residential, commercial, and

utility-scale systems. One such innovation is called bifacial PV, which allows collection of

light from the front as well as rear surfaces of a flat PV panel.

In this thesis, we present a detailed investigation of bifacial solar PV farms analyzed across

the globe. We define the problem, explore the challenges, and collaborate with researchers

from academia and the PV industry to find a novel solution.

First, we begin by developing a multi-module computational framework to numerically

model a utility-scale bifacial solar PV farm. This requires integrating optical, electrical,

thermal, and economic models in order to estimate the energy yield and LCOE of a bifacial

PV system. The first hurdle is to re-formulate the LCOE so that the economist and the

technologist can collaborate seamlessly. Thus, we re-parameterize the LCOE expression

and validate our economic model with economists at the National Renewable Energy Lab

(NREL).

Second, we extend the existing optical and electrical models created for stand-alone

bifacial PV panels to models that can simulate a large-scale bifacial solar PV farm. This

brings the challenge of mathematically modeling solar farms and light collection on the rows

of PV panels elevated from the ground by taking into account the mutual shading between

the rows, reflections from the ground, and elevation-dependent light absorption on the rear

surface of the PV panels from several neighboring rows. Next, we integrate temperaturedependent

efficiency models to take into account the effects of location-dependent ambient

temperature, wind speed, and technology-varying temperature coefficients of the solar PV

system in consideration.

Third, we complete the comprehensive modeling of bifacial solar PV farms by including

two types of single-axis tracking algorithms viz. sun-tracking and power tracking. Using these

algorithms, we explore the best tracking orientation of solar farms i.e., East-West tracking

vs. North-South tracking for locations around the world. We further find the best land type

suitable for installation of these E/W or N/S tracking bifacial solar PV farms.

Fourth, we reduce the computation time of numerical modeling by utilizing the advantages

of machine learning algorithms. We train neural networks using data from the alreadybuilt

models to emulate the numerical modeling of a solar farm. Amazingly, we find the

computation time reduces by orders of magnitude while accurately estimating the energy

yield and LCOE of PV farms.

Fifth, we derive, compare, and experimentally validate the thermodynamic efficiency

limits of photovoltaic-to-electrochemical energy conversion for the purpose of storing solar

energy for future needs.

Finally, we present some new ideas and guidelines for future extensions of this thesis as

well as new challenges and problems that need further exploration.