Research On Spatiotemporal Characteristics And Intelligent Diagnosis Of Discharge UV Spectrum Of Composite Insulator

Insulators are one of the crucial components in power system,and are susceptible to abnormal condition of breakage,aging and reducing of hydrophobicity under long-term heat,electricity and mechanical stress,the discharge phenomenon in wet and contamination conditions thereafter jeopardizes the safe and stable operation of power system.UV imaging method has been widely used in discharge inspection for its advantages of high detection sensitivity.Presently,UV count and spot area are mainly selected as quantitative parameters of the UV imaging method.Focused on the spot area sequence of UV images or videos,technology interfere is needed in field detection of UV imager deployed on UAV,and its intelligent level needs to be improved.In this dissertation,the discharge emission spectrum of solar-blind UV band and its influencing factors is studied.According to engineering demand,the spatio-temporal quantification parameter characteristics and its intelligent diagnosis method are presented from the perspective of insulator identification and quantitative parameter formulation,extraction,processing and assessment of insulator discharge UV video.The main contents are as follows:The emission spectrum of solar-blind UV band of air discharge is analyzed,and the influence of environmental factors on the spectral characteristics is studied.Based on the needle-plate discharge solar-blind UV emission spectrum research platform,properties of solar-blind UV band emission spectral is discussed,and its characteristics influenced by temperature,humidity and air pressure are experimentally researched.A method for calculating spatiotemporal quantification parameters of insulator discharge video is proposed,and the influence of video duration and so on is studied.The processing of the discharge UV video is realized based on the digital image processing algorithm,and the dicharge frequency characteristics of UV video is extracted.The discharge position characteristics are obtained through the discharge contour extraction and fitting algorithm.The space-time tensor of the UV video is defined and calculated.With homogenization,pseudo-color and image registration processed,the spatio-temporal quantization parameters are obtained by aggregating the spatio-temporal tensor on the dimentsion of spot and frame number.The effects of image processing parameters,video duration and UV gain are analyzed.Characteristics and influencing factors of the discharge images of the contaminate-wet composite insulator are studied based on the UV imaging method,and the discharge characteristics of the water droplets attached to the composite insulator are analyzed theoretically.The relationship between the pollution degree and the conductivity of the polluted aqueous solution under different wetting levels is analysed.A 110kV insulator discharge platform was built,and the effects of volume,conductivity and position of water droplet on discharge are studied by UV imaging method experimentally.The corona onset voltage,discharge position and electric field strength characteristics of the model are studied.Based on the theory of strong electrolyte aqueous solution and the simplified model,the relative permittivity of different contaminated aqueous solutions is theoretically analyzed.Thereafter,an UV video database is established and their spatiotemporal parameters are analyzed.An automatic inspection scheme of the Unmanned aerial vehicle loaded UV imager is proposed based on the insulator identification of the visible channel image and deep learning system optimization strategy.A YOLO based deep learning hardware and software platform is established,training and test database of insulator image of the visible channel of UV imager are constructed.Insulators image recognition performance of the network is optimized based on IoU,data enhancement algorithm and self-built database label data clustering algorithm.An automatic inspection scheme of transmission line insulators based on the expert diagnosis system of the unmanned aerial vehicle ultraviolet imager is proposed.Based on UV video spatiotemporal quantification parameters and convolutional neural network,a diagnosis method of insulator discharge severity is proposed.Based on spatiotemporal quantification parameters and insulator discharge position,the risk coefficient of insulator insulation state is proposed.The preprocessing database construction and labeling of UV image and video are accomplished.Considering the discharge frequency and position characteristics,the diagnostic procedure,system optimization and evaluation reference of UV video based on deep learning are proposed.The influencing factors of UV video statistical parameters and spatiotemporal parameters are studied and compared,and optimization methods are proposed.A comprehensive diagnostic system based on B/S mode is developed and applied.