Optical design of a fully LED-based solar simulator

This thesis presents the simulated optical design of a fully LED-based solar simulator. The work focuses on the spectral mismatch, the spatial uniformity acquired with direct light and the spectral uniformity. The proposed LED solar simulator has an illuminated area of 32cm x 32cm and can characterise medium size photovoltaic devices under variable light intensities and variable output spectra. The spectral range covered is between 350nm and 1300nm which offers the capability of characterising various different PV technologies. The spectral match classification is A+ for the 400nm-1100nm spectral range and B for the 350nm-1300nm spectral range. The spatial non-uniformity of irradiance is also A+ across the illuminated area. The temporal stability of LEDs can easily reach class A as proven by previous work in the group and is not examined here. An automated LED selection methodology that optimises the spectral mismatch was developed to replace the trial and error method usually employed. The algorithm created accommodates a more accurate selection of the most appropriate LED wavelengths in order to represent the solar spectrum even more closely than before and improve the uncertainties caused by the spectral mismatch. A genetic algorithm and the chi-squared error criterion were used to create the automated methodology applying a minimisation technique. This technique helps the user choose from a wide variety of LEDs available on the market, determine the wavelengths and the number of LEDs per wavelength needed to accurately represent the AM1.5G solar spectrum and other spectra and provides a cost-effective and straightforward solution. The solution chosen for this project involves 24 different wavelengths. A direct beam approach was followed regarding the collimation of light to account for the measurement errors introduced by the frequent overestimation of the current due to the unpredicted reflections caused by diffuse light. Extended simulations of different optics were performed to determine the best layout that offers good directionality and satisfactory non-uniformity of irradiance and light collection efficiency. Total internal reflectors of 13.5mm diameter proved to be the most appropriate primary optics with the highest collection efficiency. An imaging homogeniser was chosen as secondary optics for its capability to mix the light and achieve low levels of non-uniformity of irradiance. The spatial non-uniformity of irradiance achieved with 612 LEDs is 0.29% across the 32cm x 32cm illuminated area and the irradiance is equal to 1316 W/m2 assuming 1W LEDs. The hexagonal placement set-up was used for the placement of the LEDs since it results in the lowest non-uniformity and it is the best option for keeping the lamp size compact. An optical engineering software called FRED was used for ray-tracing individual optics. Due to the time and computational demands of the simulations a different approach needed to be found for overlaying the irradiance profiles of hundreds or even thousands optical elements. An algorithm was developed in Matlab that takes into consideration the geometry of each case and calculates the final irradiance profile. A placement methodology that accounts for the spectral uniformity on the illuminated target was also developed. It was shown that placing the LEDs randomly does not offer enough spectral mixing and is therefore problematic as it introduces an unexpected source of measurement uncertainty. The influence of spectral non-uniformity varies for different photovoltaic technologies due to their variable spectral responses. Thus, a placement methodology using a genetic algorithm was developed to optimise the positioning of the LEDs. As a result the highest spectral non-uniformity drops from almost 5% to 1.46% and the measurement uncertainty is reduced significantly since an improvement of up to 1.8% is noted in the current density non-uniformity.