中文摘要：

油纸绝缘缺陷问题直接影响换流变压器的运行安全,而实现换流变压器的绝缘诊断对保证直流输电系统正常运行具有重要意义,但由于换流变压器内部同时存在着交流和直流的叠加电场,放电特征与交流电压单独作用时具有明显区别,因此交流变压器缺陷诊断方法不一定适用于换流变压器,基于局部放电的n-q-？谱图信息,构建了灰度图像,提出了基于图像特征的换流变压器缺陷诊断方法,提取属于图像本身的颜色特征、形状特征及纹理特征构成原始特征空间,采用粗糙集理论对特征空间进行优化,通过人工神经网络分类器实现缺陷模式识别。同时搭建了交直流复合电压下5种油纸绝缘典型缺陷放电检测平台,获取不同缺陷放电n-q-？谱图,并构建识别图像,验证了基于图像特征的缺陷诊断方法在换流变压器交直流复合电压下的使用效果,并发现该方法可以有效实现换流变压器缺陷诊断。

英文摘要：

Since the defect in oil-paper insulation directly endangers the operational security of converter transformer, it is of significance to implement the diagnosis on the insulation of converter transformer for ensuring the normal operation of HVDC power transmission system. However, due to the superposition of AC and DC electric fields coexisted inside the converter transformer, the partial discharge （PD） characteristic under such a circumstance obviously differs from the PD characteristic under the circumstance that there is only AC electric field inside the transformer, thus the insulation defect diagnosis method for AC power transformer is not always suitable for converter transformer. Based on the information in n-q-φ spectrogram, a gray level image is constructed and a insulation defect diagnosis method for converter transformer based on image characteristic is proposed to extract the color feature, the shape feature and the texture feature belonging to image itself to constitute an original feature space, and utilizing the rough set theory the original feature space is optimized, and then the defect pattern is recognized by artificial neural network classifier. A PD detection platform for five kinds of typical oil-paper insulation defects under AC-DC compound voltage is constructed to obtain n-q-~o spectrograms of PDunder different insulation defects and constitute the image for the recognition, thus the effect of image feature based defect diagnosis method utilized in the environment of AC-DC compound voltage inside the converter transformer is validated and it is found that the proposed method can be used to diagnose the defect in internal insulation of converter transformer effectively.