Photovoltaic Module Defects Classification Analysis using DenseNet Architecture

A Photovoltaic (PV) module is a module that absorbs solar energy and converts it into electrical energy, which will be used for electrical equipment. In recent years, the PV capacity of PV technologies has increased. However, the efficiency of solar modules has been reduced because of their defections on it, making the classification method required. There are many techniques that have been used in PV defect classification. One of the techniques is ResNet architecture which had been chosen as a suitable technique for the classifiers however there is a flaw in the architecture that affect the result of accuracy in classification. The flaw in the ResNet architecture is the identity shortcut that has the disadvantage of causing the collapsing domain problem, which decreases the network’s learning capacity. Therefore, the DenseNet-201 architecture is applied for classifying the types of PV defect modules. For comparison purposes, other CNN architectures namely, Deep CNN (DCNN) and VGG-16 Architecture are used in analysis performance. The result shows that DenseNet-201 gives the highest performance with achieved 94.70% of accuracy. In addition, DenseNet-201 also outperform in term of F1-Score (94.65%), Sensitivity (92.76%), Specificity (98.43%) and Precision (96.82%) when compared to other CNN architectures. Thus, the analysis will help in improving the PV module defects classification.