Research On Defect Type Diagnosis Method Of 10 KV Cable Terminal Based On Partial Discharge Characteristics

With the rapid development of economy and the continuous advancement of urbanization,XLPE cables have been widely used in urban power distribution networks due to their excellent insulation properties.As an important power equipment in the urban power supply system,its operation stability is closely related to national life.The cable termination is an important part when the cable is put into operation,and it plays the role of connecting the cable and other electrical equipment.According to the statistics of cable faults for many years,it is found that the cable terminal is a high-incidence component of faults,and it is also a key position to prevent breakdown accidents.The cable terminal is prone to small defects during the production and installation process.The serious electric field distortion at the defect can easily lead to partial discharge.The long-term existence of partial discharge leads to accelerated insulation deterioration,resulting in a gradual decrease in the insulation performance of the cable terminal,which in turn leads to breakdown of the cable terminal..Once the cable terminal fails,it will cause a power outage accident,and cause huge economic losses.Therefore,effectively diagnosing defect types and taking targeted maintenance measures to reduce power outages caused by breakdown of cable terminals is of great significance for improving the operational stability of power systems.In view of this,this topic focuses on the 10 kV XLPE cable terminal,and deeply investigates and analyzes the common defect types and fault causes of the cable terminal.For the cable terminal with typical defects,it mainly focuses on the partial discharge characteristics,texture feature extraction method and defect type diagnosis.The research was carried out in three aspects.First,the XLPE cable terminal samples under four typical defects were made,the accelerated aging experiment was designed,and the cable terminal samples under multiple aging times were prepared.A partial discharge test platform was built to analyze the partial discharge variation characteristics of defective cable terminals under different aging times.The research results show that the partial discharge spectra of the same defect type collected under different aging times have a certain similarity,but with the increase of aging time,the spectral contour and the dense area of discharge show certain differences.Secondly,based on the extraction method of texture features,LBP texture features,GLCM texture features and Tamura texture features of partial discharge spectra are extracted respectively.The change law of texture features under typical defects is analyzed,and monotonic texture features are found,which can be used to represent the change process of partial discharge spectra under different aging times.The feature space dimensionality reduction method based on principal component analysis.The texture feature space before and after dimensionality reduction is established.The research results show that a single texture feature can not effectively represent the change law of partial discharge spectra of four types of defects,so it is necessary to build a multi-dimensional texture feature space.The simple fusion of multiple texture features increases the redundancy of the feature space,and the relevant features are removed by principal component analysis,which provides a data basis for the type diagnosis of cable terminal defects.Finally,an improved SAE defect type diagnosis method based on PCA is proposed,and a variety of cable terminal defect types diagnosis schemes are designed.The confusion matrix is used to analyze the diagnosis results of each scheme.Based on the accuracy,precision,sensitivity and F₁ index,the diagnostic effect of each program was comprehensively evaluated.The research results show that the fusion of texture features after dimension reduction and SAE self-extraction features has the best effect on the diagnosis of cable terminal defect types,and the recognition accuracy rate reaches 95.6%,which verifies the effectiveness of the method proposed in this paper.