Identification And Analysis Of Dust On The Surface Of Photovoltaic Panels In Desertification Areas

In recent years,the solar photovoltaic power generation industry has developed rapidly.Open areas such as deserts and wastelands are important areas for the construction of large-scale centralized photovoltaic power plants.The surrounding areas are usually relatively empty,and the surface form is generally sandy.When severe weather such as strong winds and sandstorms occur,floating dust and flying dust will appear in the photovoltaic power station,causing sand particles in the sky to settle on the surface of the photovoltaic panels,which will have many important negative effects on the photovoltaic power generation system.Due to factors such as the arrangement of photovol taic panels,topography and other factors,the amount of dust on the surface of the photovoltaic panel is not uniform,and the dust on the surface of the photovoltaic panel in some locations is relatively more serious,which affects the performance and life of the photovoltaic cell in the corresponding range to a certain extent.It brings a lot of uncertainty to the forecast of the power generation of photovoltaic power plants.Therefore,the realization of non-contact identification of dust on the surface of the photovoltaic panel and direct extraction of sand parameters is of great significance to the maintenance and operation of photovoltaic power plants.The rapid development of deep learning target detection technology and image processing technology provides new technical ideas for the automatic detection of dust accumulation in photovoltaic power stations.In view of this,the work content of this article is as follows:First of all,in response to the demand for detection of dust targets on the surface of photovoltaic panels in the operation of photovoltaic power plants,an SSD model with Resnet50 as the feature extraction backbone network is proposed,and the channel and spatial attention mechanism modules and feature pyramid networks are added to detect the surface dust of photovoltaic panels.Training on 2400 photovoltaic panel surface dust image data sets,the average accuracy of the improved SSD algorithm is 92.65%,and the detection speed is 26.4FPS.And through YOLOV3,Faster-RCNN,SSD and improved SSD algorithm to compare the surface dust detection of photovoltaic panels and the detection results of photovoltaic panel surface dust in complex environments.The experimental results show that the improved SSD algorithm can effectively improve the detection ability of objects,and quickly and accurately detect the dust on the surface of the photovoltaic panel,and effectively reduce the missed detection rate of the surface dust of the photovoltaic panel.Secondly,the dust on the surface of the photovoltaic panel is magnified by an optical microscope,and an improved watershed algorithm is proposed to segment the image of the adhered sand on the surface of the photovoltaic panel to solve the problem of the adhesion of sand on the surface of the photovoltaic panel.The image is mainly used for distance transformation,morphological reconstruction and minimum calibration technology..Compared with the traditional segmentation algorithm of adhesion particles,they will produce over-segmentation phenomenon when segmenting the image of adhesion particles.Experiments show that the improved watershed segmentation algorithm in this paper can accurately segment the conglutinated sand and suppress the phenomenon of over-segmentation.Finally,perform sand parameter extraction on the processed sand image,calculate the conversion coefficient,measure the actual sand parameters for analysis,and calculate the sand particle size by the equivalent circle method.A MATLAB-GUI-based detection platform for sand particles on the surface of photovoltaic panels is designed,and the platform can display the process of sand detection,and can calculate the proportion of dust accumulation,which more intuitively reflects the cleanliness of the photovoltaic panel surface.