Comparison of inline hot spot detection and evaluation algorithms for crystalline silicon solar cells
We present a detailed comparison of five inline-feasible hot spot detection and evaluation algorithms that are tested at the hands of inline thermography measurements of 108 industrial monocrystalline silicon solar cells. We evaluate the algorithms concerning their ability to predict the measured equilibrium temperature increases of the reversely-biased solar cells from short-term measurements and compare the requirements of each method. Beyond that, we apply and test for the first time a novel approach using the temporal evolution of the temperature of a hot spot right after turning on a direct voltage, only needing a series of temperature-calibrated thermography images (time series approach). We show that with the most accurate method of iterative convolution of power-calibrated thermography images with temperature response functions, equilibrium temperature increases can be predicted with a mean absolute error of only 3 K and a correlation coefficient of 0.90. Furthermore, relying both on similar physical considerations, we find the new time-series approach and the temperature increase T(100 ms)-T(80 ms) in the shortterm measurement to yield almost as accurate results (correlation coefficients of 0.86 and 0.87 respectively and mean absolute error of 3.5 K for the time series approach), while implementation is less challenging for these.