Ensembling techniques in solar panel quality classification

Trong Hieu Luu (Can Tho University)
Phan Nguyen Ky Phuc (International University-Vietnam National University)
Tran Lam (Hanoi University of Science and Technology)
Zhi-qiu Yu (National Taiwan University of Science and Technology)
Van Tinh Lam (Bac Lieu Vocational College)

International Journal of Electrical and Computer Engineering
Vol 13, No 5: October 2023

Solar panel quality inspection is a time consuming and costly task. This study tries to develop as reliable method for evaluating the panels quality by using ensemble technique based on three machine learning models namely logistic regression, support vector machine and artificial neural network. The data in this study came from infrared camera which were captured in dark room. The panels are supplied with direct current (DC) power while the infrared camera is located perpendicular with panel surface. Dataset is divided into four classes where each class represent for a level of damage percentage. The approach is suitable for systems which has limited resources as well as number of training images which is very popular in reality. Result shows that the proposed method performs with the accuracy is higher than 90%.