Improved Solar Panel Efficiency through Dust Detection Using the InceptionV3 Transfer Learning Model

Abstract

There is an expanding sector for renewable energy due to raised awareness of carbon footprints, highlighting the necessity of sustainable and eco-friendly energy sources. Although solar power is well-known for producing electricity, it faces notable challenges from the accumulation of dust and debris on its panels, contributing to reduced efficiency and output. An innovative model based on deep learning has been proposed to detect dust on solar panels, desiring to tackle this issue. To address the scarcity of labelled data and enhance accuracy and robustness, the model incorporates Inceptionv3, VGG 16, and a custom Convolutional Neural Network (CNN) and employs data augmentation to improve the dataset. Trained and tested on the Dust Detection on Solar Panel dataset from Kaggle, the presented system, especially the Inceptionv3 model, performed outstandingly with 93.10% accuracy, 95% precision, 94% recall, and a 94% F1 score. This model can significantly improve electricity generation by automating the detection and cleaning process, thereby maintaining solar power as a possible solution for sustainable energy production.

Topics & Concepts