【作者】 魏钢；

【导师】 唐炬；

【作者基本信息】 重庆大学 ， 电气工程， 2013， 博士

【摘要】 电力电缆线路是城市电网中重要的组成部分，其安全可靠稳定运行对于城市电网具有重要意义。据统计，全国已投入运行的110kV及以上的高压电缆线路已经超过8000公里，最高电压等级已达500kV。高压电缆断面是电缆线路的薄弱环节，电缆接头和终端中绝缘屏蔽断口处电应力相对集中容易发生故障。电缆接头采用封闭式绝缘结构且一般在现场完成组装密封，相比本体和户外终端，绝缘安全裕度偏小，另外，电缆接头现场施工质量要求高，如施工工艺不良或密封措施不到位，那么在地下潮湿恶劣运行环境中极易造成绝缘性能劣化，因此电缆接头成为电缆线路最易发生绝缘故障的薄弱点。运行经验和研究均表明：电缆局部放电量与其绝缘状况密切相关，局部放电量的变化预示着绝缘一定存在着可能危及电缆安全运行寿命的缺陷，局部放电测量能很好的反映高压电力电缆及其接头的运行状况，及时发现故障隐患，保障电力电缆线路安全可靠运行，具有重大的经济和社会效益。国内、外专家学者、IEC、IEEE以及CIGRE等国际电力权威机构一致推荐局部放电试验是作为交联聚乙烯电力电缆绝缘状况评价的最佳方法。本文在对国内外交联聚乙烯电力电缆及其接头中局部放电信号检测研究的基础上，详细分析了各种检测和模式识别技术，将局部放电在交联聚乙烯电力电缆及接头中的传播衰变特性、局部放电检测技术和特征提取及识别技术作为本论文研究的重点，取得的主要成果有：①通过对介电弛豫理论的研究，深入分析交联聚乙烯电力电缆半导电层的介电特性，获得了可用于工程实际计算的半导电层复介电常数的修正公式。根据Davidson-Cole方程拟合出半导电层的复介电常数，将该常数代入阻抗分布参数加权比的传输线简化模型中进行理论计算，采用FDTD法对电缆及接头局部放电的传播特性开展仿真研究，为交联聚乙烯高压电力电缆接头局部放电检测试验及故障诊断提供理论和技术支持。同时，根据电缆接头常见绝缘故障形成原因构建了四种典型缺陷模型。②设计了符合高压电缆及接头结构特征的局部放电检测传感器。通过建立电路及天线模型，采用传递函数推导、阻抗特性计算、方波响应等方法系统全面的分析了传感器的频响特性和输出特性，对传感器各项电气和物理参数进行了优化。③主要采用电容传感器构成的电力电缆局部放电检测系统在110kV电缆接头试验平台上对研制的四种典型接头缺陷进行检测，获得大量试验数据并构建出电缆接头局部放电q n谱图及其灰度图像。④提出基于提升小波变换的局部放电一维特征量的提取方法。该方法基于Birge-Massart策略对提升小波变换系数矩阵进行降维，通过奇异值分解运算获取故障识别的特征向量，在保留有效特征的情况下，减少识别维数，降低噪声，缩短识别时间。⑤提出基于(2D)2MMC+LDA框架算法的电缆接头局部放电特征提取方法。该方法既基于图像行或列去判别向量，从而实现挖掘图像的局部特征；又基于整幅图像去找判别向量，考虑了全局特征。不仅可以解决维数危机，消除类内散度矩阵的奇异性，而且能最大限度的保持原有样本模式的结构分布，有效减少样本训练时间，提高了模式识别的精度，并且可以实现参数的自动选择。更多还原

【Abstract】 Power cable line is one vital part of urban power network, and its safety, stabilityand reliability are of great significance in the urban power network. According tostatistics,110kV and above HV cable lines which have been put into operation are morethan8000kilometers in China, with the highest voltage reaching500kV.High voltage cable section is a weak point of cable line. The electric stress ofinsulation shielding fracture section in the joint and terminal can lead to a malfunctionbecause of the electric field relative concentration. With an enclosed insulation structure,joint is always assembled on the spot. Compared with the power cable and terminal, theinsulating safety margin of joint is small. In addition, high quality is required on the siteconstruction of joint. Bad construction technology or inadequate sealing measures willlead to insulation deterioration in the formidable underground operational environment.Therefore, the joint turns into the weak point vulnerable to insulation fault in the cableline.Operational experience and research indicate that: the quantity of partial dischargein cable and its insulating condition are strongly relevant. The change of the quantity ofpartial discharge indicates insulation defects which may endanger the safe operation,while the partial discharge measurement which can represent the operational conditionof high voltage cable and joint leads to the fact that the hidden fault will be discoveredin time, and significant economic and social benefits be gain because of reliableoperation of cable line. Partial discharge experiment is regarded to be the best methodfor the evaluation of XLPE cable insulating condition by experts, scholars home andabroad, and authoritative organization(such as IEC、IEEE and CIGRE).Based on the study home and abroad on partial discharge signal detection in XLPEpower cable and joint, this dissertation analyzes all detection and pattern recognitiontechnology in detail, emphasizes on partial discharge characteristic of transmission inXLPE power cable and joint, partial discharge detection technology and recognitiontechnology. The major achievements are listed as follows:①Based on the analysis of dielectric relaxation theory and dielectric property ofsemiconductive layer, modificative equation for complex dielectric constants ofsemiconductive layer which can be used in the practical calculation is obtained.According to Davidson-Cole equation, complex dielectric constants of semiconductive layer are fit and theoretical calculations while the constant substitutedinto simplified transmission line model of impedance distribution parameters weighedratio are carried out. FDTD method is adopted to carry out simulation research onpropagation characteristics of cable and joint in order to provide reference to partialdischarge detective test and failure diagnosis in joint. Four typical defect models areestablished on the basis of common insulation failure in joint.②Sensors conforming to high voltage cable and joint structure are designed.Through the establishment of circuit and antenna model, methods (such as deduction ofthe transfer function, impedance characteristic calculation, and square-wave response)are adopted to analyze frequency response characteristics and output characteristics ofsensor systematically and thoroughly, and each electrical and physical parameters ofsensor are optimized.③Partial discharge detection system composed by capacitive sensors is chosen todetect the four typical defects in110kV cable joint test platform, mass test data areobtained and partial discharge q nspectrogram and its gray image areestablished.④A method of partial discharge one-dimension feature extraction based on liftingwavelet transform is presented. In the method the dimension of lifting wavelettransform coefficients are reduced through Birge-Massart strategy, and then theeigenvector of pattern recognition are got through singular value decomposition. So lotsof gains can be got in the condition of effective feature retention, such as dimension,noise, and recognition time reduction.⑤A partial discharge feature extraction method is presented based on(2D)2MMC+LDA frame algorithm. By this method, not only the partial feature imagemining can be implemented based on the row or columns to discriminate vector, butalso the general feature can be considered based on general image. It overcomes thecrisis of dimension, eliminates singularity of within-class scatter matrix, furthest keepsstructure distribution of original sample pattern, reduces training time of sampleeffectively, improves the accuracy of pattern recognition, and implements automaticselection of parameter.更多还原