Abstract :
This paper proposes a method used to process and group raw data from a solar installation in order to form a lookup table of maximum power points (MPPs). This lookup table returns the current and voltage values of the MPP for inputs of temperature, and solar irradiance. A main challenge addressed in this paper is how to identify data points collected under uniform solar irradiance. Two approaches are tested: The first compares the power output of separate solar arrays to deduce if there may be partial shading. The second approach uses a Gaussian Mixture Model to form 3 density based groups of data sharing similar characteristics. Although slightly less accurate, the second approach requires fewer data, and less sensing equipment. The group corresponding to uniform irradiance is then divided into pixels of temperature and irradiance. One voltage and one current corresponding to each pixel are determined by a maximum likelihood estimate (MLE). The lookup table outperforms a simple linear regression in reproducing the electrical characteristics of a solar array, especially at irradiance levels lower than 400 W/m2. To test the functionality of the lookup tables and linear regression, simulations using an emulator are performed in Simulink. The emulator, implemented in software, is otherwise as described in Di Piazza, and Vitale’s paper “Photovoltaic field emulation including dynamic and partial shadow conditions.”
|