IDS-Net: Integrated Network for Identifying Dust State of Photovoltaic Panels

Dust accumulation can severely affect the normal balance between different areas of photovoltaic panels, leading to a sharp decline in power generation efficiency and service life. In this paper, a novel identification model for dust state on the surface of PV panels is developed to analyze the dust level. Firstly, a novel identifying dust state of photovoltaic panels network (IDS-Net) is proposed built with residual neural network and convolutional block attention module (CBAM). Secondly, the Q-estimation module is designed on the basis of the improved atmospheric scattering model, which directly outputs clear images of photovoltaic panels through lightweight convolutional neural networks (CNN) and determines the dust state of photovoltaic panels through the weight relationship between the dust and cleaning image. Such a novel end-to-end design makes it easy to embed IDS-Net into other depth models to improve advanced mission performance for dust deposition images. The experimental results on the synthetic image dataset show that the test accuracy of IDS-Net is over 80%, which has better identification effect.