Online defense: Sandy finishes her PhD with great success
05/2021 - online defense: Sandy finishes her PhD with great success
Sandy joined the Hi-ERN team “High Throughput Characterization and Modelling in Photovoltaics” in July 2019. Since then, she has focused on developing a methodology that uses deep learning algorithms to predict photovoltaic (PV) module I-V curves from their respective electroluminescence images. Over 700 PV module electroluminescence images have been used in this work and they present defects such as cracks, fractures, PID effects, and others. The aim of this work is to use the PV module I-V curve predictions and reconstruct a PV string I-V curve to analyze the PV string performance, without the need to depend on sunny weather conditions and without the need for daytime PV system shut down.
In addition to her scientific work at the institute she continued her PhD thesis entitled, "Improving Machine Learning Prediction and Forecasting Models Used in Photovoltaic Monitoring Systems" at her home university “Instituto Superior Técnico, Universidade de Lisboa”, Portugal. In her thesis, Sandy studied various machine learning methodologies to predict and forecast the PV system production to accurately detect faults in real-time and prevent them from taking place in the future. Her work proposed a new hybrid methodology that included the use of five regression machine learning techniques such as the Decision Trees, Artificial Neural Networks, Multi-Gene Genetic Programming, Gaussian Process and Support Vector Machines. The results show that the hybrid methodology successfully detects PV string faults with a machine learning model error value lower than 8%. Due to the CORONA situation, the defense was held via ZOOM meeting. The jury unanimously qualified the PhD work as “Approved with Distinction”. Congratulations, Sandy!