Meiwasandi, Putu Niar and Cahyaningtyas, Rizqia and Luqman, Luqman (2025) OBJECT DETECTION BERBASIS SSD VGG16 UNTUK DETEKSI KERUSAKAN PADA PERMUKAAN MODUL SURYA. Diploma thesis, ITPLN.
202131018_Putu Niar Meiwasandi_Revisi_Skripsi_Putu Niar Meiwasandi 1.pdf
Restricted to Registered users only
Download (6MB)
Abstract
Modul surya merupakan salah satu komponen utama dalam sistem pembangkit listrik tenaga surya yang berfungsi mengubah energi matahari menjadi energi listrik. Namun, efektivitas modul surya dapat menurun akibat kerusakan seperti physical damage dan electrical damage yang mengganggu kinerja modul. Penelitian ini bertujuan untuk membangun model deteksi terhadap kerusakan tersebut menggunakan metode Object Detection berbasis arsitektur Single Shot Multibox Detector (SSD) dengan backbone VGG16. Dataset yang digunakan berasal dari platform Kaggle dan telah dianotasi secara manual dalam format PASCAL VOC, kemudian diperluas melalui proses augmentasi hingga menjadi 679 gambar dengan 966 label bounding box. Model dilatih menggunakan konfigurasi image 300x300 piksel dengan variasi batch size dan skenario split data. Evaluasi kinerja dilakukan menggunakan metrik mean Average Precision (mAP), confusion matrix, precision, recall, dan F1-score. Hasil terbaik diperoleh pada konfigurasi batch size 16 Skenario A, dengan akurasi sebesar 94% dan nilai mAP mencapai 0,92. Penelitian ini menunjukkan bahwa metode SSD VGG16 mampu mendeteksi kerusakan pada permukaan modul surya secara efektif dan akurat.
Solar modules are one of the main components in solar power generation systems that convert solar energy into electrical energy. However, the effectiveness of solar modules can decrease due to damage such as physical damage and electrical damage that interfere with module performance. This study aims to develop a damage detection model using the Object Detection method based on the Single Shot Multibox Detector (SSD) architecture with the VGG16 backbone. The dataset used was sourced from the Kaggle platform and manually annotated in PASCAL VOC format, then expanded through augmentation to 679 images with 966 bounding box labels. The model was trained using a 300x300 pixel image configuration with variations in batch size and split data scenario. Performance evaluation was conducted using the mean Average Precision (mAP) metric, confusion matrix, precision, recall, and F1-score. The best results were obtained with a batch size of 16 and 50 epochs, achieving an accuracy of 94% and an mAP value of 0.92. This study demonstrates that the SSD VGG16 method can effectively and accurately detect damage on the surface of solar modules.
| Item Type: | Thesis (Diploma) |
|---|---|
| Uncontrolled Keywords: | SSD VGG16, Object Detection, Modul Surya, Physical Damage, Electrical Damage, Deep Learning. SSD VGG16, Object Detection, Solar Module, Physical Damage, Electrical Damage, Deep Learning. |
| Subjects: | Skripsi Bidang Keilmuan > Teknik Informatika |
| Divisions: | Fakultas Telematika Energi > S1 Teknik Informatika |
| Depositing User: | Sudarman |
| Date Deposited: | 13 Oct 2025 08:29 |
| Last Modified: | 13 Oct 2025 08:29 |
| URI: | https://repository.itpln.ac.id/id/eprint/2169 |
