Photovoltaic Panels Fault Detection with Convolutional Neural Network and Bitterling Fish Optimization (BFO) Algorithm

This paper presents a robust framework for detecting faults in PV panels using Convolutional Neural Networks (CNNs) for feature extraction and Bitterling Fish Optimization (BFO) algorithm for feature selection. The system integrates five pre-trained CNN architectures—GoogleNet, SqueezeNet, ResNet-50, VGGNet-16, and AlexNet—with BFO to optimize feature selection and improve classification performance. The VGGNet-16 + BFO combination achieved the highest accuracy of 98.75%, with excellent sensitivity of 98.72%, specificity of 98.78%, precision of 98.76%, and F1 score of 98.74%, demonstrating highly balanced and reliable performance. The present research shows that the BFO algorithm plays a crucial role in selecting the most significant features and improving classification accuracy compared to conventional techniques. The measures like accuracy, sensitivity, specificity, precision, and F1 score validate the success of the model in correctly predicting faulty PV panels. From the overall performance perspective, BFO performs better than alternative optimization methods such as Gray Wolf Optimization (GWO) and Ant Colony Optimization (ACO). The success of this approach with various CNN models highlights the effectiveness of BFO for accurate PV panel defect identification. The combination of BFO and CNN provides high accuracy and balanced performance in PV panel defect detection, making it a viable solution for improving the efficiency and reliability of solar systems.