Denoising Convolutional Neural Networks Based Dust Accumulation Status Evaluation of Photovoltaic Panel

Dust on photovoltaic panels can reduce generating efficiency, cause ignition, corrosion and other types of faults. To solve the problem of lack of effective evaluation methods and real-time detection technologies for the dust accumulation status of PV panels, an image acquisition system for dust accumulation status on photovoltaic panels is built, and a novel denoising convolutional neural networks based dust accumulation status evaluation of photovoltaic panel is proposed in this paper. According to the comparison among the different combinations of DnCNN and VGG-16, AlexNet, ResNet models, the serial connection of DnCNN and ResNet-50 model could achieve real-time monitoring and quantitative evaluation tasks of dust accumulation status with a higher accuracy and better time-consuming performance. Our newly proposed method can avoid the traditional calculation of photovoltaic system operating parameters, greatly improving the accuracy and efficiency of the real-time dust accumulation status inspection of photovoltaic panels, which gives us a new idea and perspective for the monitoring and early-warning of other electric power equipment operation status. It is also worth mentioning that, our proposed dust accumulation status evaluation method can further improve its accuracy and efficiency by optimizing the neural network structure, enhancing the real-time image quality, and expanding the image data volume.