Identifying Damage Types in Solar Panels Through Surface Image Analysis with Naive Bayes

Wiliani, Ninuk and T.K.A, Rahman and Ramli, Suzaimah (2024) Identifying Damage Types in Solar Panels Through Surface Image Analysis with Naive Bayes. Journal of Applied Research In Computer Science and Information Systems, 2 (2). pp. 211-217.

Text 200.pdf Download (414kB)

Abstract

The growing utilization of solar panels as a renewable energy source requires efficient maintenance solutions to guarantee their best functioning. Identifying and categorizing faults on solar panel surfaces is essential for maintenance, as these defects considerably affect energy output and system efficiency. This study investigates the utilization of statistical feature extraction methods alongside Bernoulli Naive Bayes (BNB) and Gaussian Naive Bayes (GNB) algorithms to categorize different defect types, such as cracks, scratches, spots, and non-defective surfaces, through digital image analysis. Statistical criteria, including recall, specificity, and area under the curve (AUC), are employed to assess model performance. The findings indicate that the GNB algorithm surpasses BNB, with a mean average precision (mAP) of 39.83% with an 85:15 training-test ratio, whereas BNB reaches a maximum mAP of 29.25% at a 90:10 ratio. Nonetheless, both models demonstrate constraints in precision, as indicated by a total AUC of 0.644. This work illustrates the potential of statistical feature extraction approaches for defect classification, while emphasizing the necessity for future improvements to boost the efficacy of feature extraction and classification techniques in practical applications.

Item Type:	Journal
Uncontrolled Keywords:	Solar Panel Defect Classification, Statistical Feature Extraction, Naive Bayes Algorithms, Digital Image Analysis, Performance Evaluation Metrics
Divisions:	School of Graduate Studies
Depositing User:	Muhamad Aizat Nazmi Mohd Nor Hamin
Date Deposited:	07 May 2025 02:47
Last Modified:	07 May 2025 02:47
URI:	http://ur.aeu.edu.my/id/eprint/1361

Actions (login required)

View Item