Professional Awards Received
Place	Name	Description
2	American Statistical Association (ASA), Michigan Chapters	$50 Award of Merit
2	Society of Women Engineers - Detroit	Certificate of recognition, gift basket, invitation to SWE Detroit banquet
1	Environmental Protection Agency	Congratulatory Letter of encourgement
1	Association of Women GeoScientists	Awards Certificate
1	Environmental Management Association	First place scholarship award on behalf of the Environmental Management Association, certificate and an award of $1500. Invitation to the Environmental Achievement Awards event at the Detroit Yacht Club on May 18 to receive their award and discuss + displ
5	MSEF	Selected for competition into the Michigan Science and Engineering Fair on March 15, 2023.

 

Research Plan

In an age of booming technology and heightening climate change, solar energy plays a critical role in reducing greenhouse gasses and bringing clean energy to the world [1]. Energy obtained from the sun is limitless and clean unlike other sources of energy such as fossil fuels [2]. Solar panels are one of the most valuable ways to harness the Sun’s energy but over time they can degrade, reducing their efficiency [3]. To prevent the use of degraded solar cells, I propose a user-friendly web application that can be used to detect the likelihood of damage in solar cells. To create this project, a photovoltaic elpv-dataset will be used; it is composed of over 2000 images and labels and will be split to be used as the training and testing dataset. I will train various machine learning classifiers, including Random Forest, SVM, Naive Bayes, Decision Trees, MLP, and CNN on the dataset and determine the highest accuracy. After this, I will input the highest accuracy models into a Django web application. The web application will serve as a tool to identify whether solar panel cells are likely defective or not. Overall, a user-friendly solar cell application would make identifying damaged solar cells more accessible, leading to higher efficiency solar resources.

Abstract

As environmental concerns relating to climate change and dwindling energy resources heighten, the need for renewable, clean energy becomes more pressing. Solar panels are one of the most valuable ways to harness the Sun’s energy. Over time, though, they can degrade, reducing their efficiency. To prevent the use of degraded solar cells, a user-friendly web application is created to detect the likelihood of damage in solar cells. To create this project, a dataset composed of 2624 electroluminescent solar cell modules and labels was used. These images and labels were trained on various machine learning classifiers, including Random Forest (RF), SVM, Naive Bayes, Decision Trees, MLP, and CNN. The highest accuracies were found in RF (78.9%), SVM (78.1%), and CNN (77.0%). Various aspects of these three models were manipulated (including image preprocessing techniques, number of training/testing data, feature extraction techniques, and more) to raise the classifier accuracies. RF obtained the highest accuracy of 84.4%. The RF classifier was integrated into a novel Django web application with an easy-to-use user interface. Users can input thousands of images at once into the application and obtain rapid predictions on defect probability. At almost 85% accuracy, the web application is a valuable tool to aid solar panel users in determining whether their solar cells are defective, leading to higher efficiency solar resources.