Photovoltaic system performance enhancement: Validated modeling methodologies for the improvement of PV system design

Abstract

Photovoltaic (PV) energy generation is rapidly expanding, driven by decreases in capital and operational costs. Modeling of expected energy output is a major factor affecting the operation and construction of these systems, and the objective of this thesis is to present methodologies and tools which improve PV system modeling. Initially, the modeling of short circuit current in PV modules is investigated and demonstrates techniques for filtering of data from fielded PV modules, allowing performance metrics traditionally derived in laboratory tests to be derived from data collected from stationary outdoors systems.

The spectral component of irradiance is important for PV system modeling, and dataset of hourly spectral irradiance from 250nm-2500nm was collected from November 2012 to April 2014. This data set can be used to validate the derived equations from the previous chapter, and is aimed to enable improved modeling of spectra using an iterative methodology.

Albedo, or reflected, irradiance is then considered, where effects of spectral mismatch can be substantial. It was found that spectral mismatch error can lead to modeling errors ranging from 0.04 % to 10.5 % as PV module tilt increases from 25 degrees to 90 degrees from the horizontal, and methods of predicting and modeling this albedo spectral mismatch were presented.

Detailed methods for identifying snowfall losses were investigated, and provided results from two typical winters in 2010 and 2011. It was found that there is not a simple correlation between meteorological factors and snowfall losses, which warrant further investigation in the the modeling an prediction of this phenomenon. Consequently, an empirical lag 1 time-series model is proposed, based on the stochastic variation in snow depth that is able to predict daily snow fall losses within 3%.

Finally, an investigation is undertaken into the use of planar diffuse reflectors for albedo augmentation. It was found experimentally that such a system can increase system output by 18%. A physically based non-empirical model was developed that predicted system output with a bias error of 1%, and was used to perform a sensitivity analysis that demonstrated a yearly energy increase of 30% is achievable at the latitude of Kingston, Ontario.