Prediksi Daya Hasil Keluaran Solar Cell Menggunakan Yolo V5 dengan Arsitektur Efficientnet

Abstract

This research evaluates the performance of the YOLO V5 model which uses the EfficientNet architecture in predicting solar cell output power based on image data. The research results show that this model achieves an average accuracy of 86.4%. Accuracy for each power class is 90% for the Low Power class, 84.7% for the Medium Power class, and 84% for the High Power class. From a total of 35 test image data, the model succeeded in correctly predicting 33 images, while 2 images were not predicted well. The YOLO V5 model with EfficientNetv2 architecture shows good capabilities in detection and prediction based on power classification. These results indicate that the model has significant potential in solar cell output power prediction applications, although there is still room for improvement in accuracy, especially for various power classes