基于频域介电响应特征指纹的油纸绝缘受潮及老化状态区间识别

The Range Recognition of Moisture and Aging Status of Oil-Paper Insulation Based on Frequency Domain Dielectric Response Characteristic Fingerprint

频域介电谱技术作为一种无损测试及绝缘诊断手段,近年来受到了广泛关注并成为研究热点,尤其是在变压器固体绝缘受潮的评估方面取得了较为显著的进展。然而,作为影响变压器绝缘性能的两个重要因素,油纸绝缘的受潮和老化对频域介电响应特性影响较为相似,目前还无法通过该技术区分变压器的受潮和老化两种状态,也无法对变压器绝缘的老化状态进行有效的评估。该文提出将油纸绝缘状态按照受潮和老化程度的不同划分为数个状态模式,采用模糊模式识别的方法,以介电响应特征指纹作为分类器的输入,以模糊模式识别结果对应的标准状态模式作为评价油纸绝缘受潮和老化程度的方法。按照油纸绝缘样本聚合度及水分含量划分30个标准状态模式,选用介质损耗频域谱在三个频段的积分作为介电响应特征参数,采用贴近度的方式实现状态模式分类。实验室油纸绝缘样本和现场变压器测试结果的分析初步证实了所提方法的有效性。

Frequency domain spectroscopy(FDS)technology has received widespread attention and become the research hot spot in recent years as a kind of nondestructive test and insulation diagnosis method,especially the assessment of the moisture status of transformer solid insulation has made significant progress.As the two important factors influencing the transformer insulation performance,moisture and aging status present similar effect on the frequency domain dielectric response of oil-paper insulation.In the present study,it is unable to separate the moisture and aging status or effective assess the aging status of transformer with FDS method.The oil-paper insulation state was divided to several state patterns according to different moisture and aging status and a fuzzy pattern recognition method was proposed in this paper,the inputs of the classifier were the dielectric response characteristic fingerprint,and the standard state pattern corresponding to the pattern recognition results was regarded as the assessment moisture and aging status of oil-paper insulation.Thirty standard state patterns were divided according to the degree polarization and moisture content of laboratory oil-paper insulation samples in this paper,the serial integrals of dielectric loss spectroscopy in three frequency bands were selected as the dielectric response characteristic fingerprints and the state pattern was classified based on close-degree.The analysis of laboratory oilpaper samples and field transformer FDS results preliminarily proves the validity of the proposed method in this paper.