Ed'ler için güneş panelinden toz giderme tekniği temizleme döngüsünün geliştirilmesi

Solar panels play a crucial role in harnessing sunlight to generate electricity, but their efficiency is significantly impacted by many factors. The accumulation of dust on solar panels poses a challenge to energy production efficiency, particularly in regions with high dust rates such as the Middle East. Therefore, implement a PV dust detection system with high accuracy using optimization algorithms like Deep Learning (DL) is essential to enhance module efficiency, reduce maintenance costs, and promote resource conservation. This project introduces a novel approach using DL and image processing to develop a dust detection system for solar panels. Leveraging Convolutional Neural Network (CNN) algorithms, the study aims to fine-tune model parameters for effective dust detection and classification. The overarching goal is to enhance module efficiency, reduce maintenance costs, and optimize resource utilization. The dataset images were cleaned, augmented, and filtered utilizing image processing tools. Then, the dataset was categorized into two classes namely clean, and dusty with training and validation processes of ratio 80% of total images (2049 images), and testing and validating with 20% of total images (513 images) were used for validation and testing processes as (307 images) for validation and (206 images) for testing. Three predefined DL algorithms which are (Inception-v3, VGG-16, and Resnet-50) and the proposed DL that is called Efficient-net were developed and implemented within this work. The proposed Efficient-net outperforms the other three techniques in high accuracy with 87% rather than Inception-v3 with 84%, VGG-16 with 72, and Resnet-50 with 80%.