THE Solar Module Surface Condition Detection using Convolution Neural Network

Abstract

Solar energy is an efficient energy source for generating electricity, enabling widespread application of solar energy as a clean and renewable substitute for conventional fuels. Indonesia has great potential in developing Solar Energy, with an average solar radiation intensity of 4.8 kWh/m²/day. One of the components in a Solar Power Plant (PLTS) is a solar module that converts solar energy into electrical energy. errors in production, delivery, installation and use over a certain period of time can reduce the performance and effectiveness of solar modules, therefore periodic monitoring and maintenance are needed to maintain the performance and effectiveness of the solar modules. This study aims to detect disturbances and damage to the surface of solar modules using . The surface condition of the solar module is divided into 5 classes, namely bird drop, clean, damage and dusty. The application of Convolutional Neural Network (CNN) is used because of its ability to recognize patterns in images, ideal for categorizing the surface condition of solar modules. By using transfer learning techniques, pre-trained models such as MobileNetV2 and Densenet201 are used to improve the accuracy of solar module surface detection. The results showed that DenseNet201 has an accuracy rate of 91%, and MobileNetV2 achieves the highest accuracy of 88%. This model is expected to help in the maintenance of solar modules so that it can maintain the performance of the device.