Implementation of Deep Learning Based on Convolutional Neural Network for Detecting Images of Solar Panel Damage in Smart Grid Systems

Camelia Putri Lestari (STMIK IKMI Cirebon)
Nining Rahaningsih (STMIK IKMI Cirebon)
Irfan Ali (STMIK IKMI Cirebon)
Dodi Solihudin (STMIK IKMI Cirebon)
Tati Suprapti (STMIK IKMI Cirebon)

Journal of Artificial Intelligence and Engineering Applications (JAIEA)
Vol. 5 No. 3 (2026): June 2026

This study aims to implement Deep Learning based on Convolutional Neural Network (CNN) in detecting solar panel damage using thermal images as part of a Smart Grid system. The main problem addressed is the difficulty of early automatic identification of solar panel cell damage using conventional methods. Through the CNN approach, this study developed a classification model to distinguish between damaged (Defective) and undamaged (Non-Defective) solar panel conditions. The research stages included thermal image dataset collection, pre-processing, model training, and performance evaluation. The results showed that the CNN model was able to achieve an accuracy of over 87% with stable performance on the validation data. Visualization using the Grad-CAM method helps interpret the damaged areas that are the focus of the model's decision.