基于集成增强KNN的油纸绝缘原始拉曼光谱老化状态判别

Aging State Discrimination of Oil-Paper Insulation Using Raman Spectroscopy and Integrated Enhanced KNN

油纸绝缘老化状态的快速准确检测一直备受关注,本研究对未进行特征提取的油纸绝缘原始拉曼光谱老化状态类别进行判定。首先,根据所测得的绝缘纸聚合度将其老化状态划分为10个类别。同时,对不同老化状态的油纸绝缘样本进行拉曼光谱检测。最后,通过K最近邻(KNN)算法、集成增强KNN算法分别对169组拉曼光谱样本进行老化状态类别判定。结果表明:经过集成增强后的KNN算法对原始拉曼光谱具有更强的识别能力,其判别准确率为98.32%,且具有更好的稳定性。证明了由集成增强KNN算法构建的判别模型能够较为准确地判别油纸绝缘原始拉曼光谱,该模型简化了变压器油纸绝缘拉曼光谱老化状态的诊断步骤,对油纸绝缘拉曼光谱检测方面的研究具有重要意义。

The rapid and accurate detection of the oil-paper insulation aging state has attracted considerable attention.In this study,classification of the original Raman spectral aging state of oil-paper insulation is performed without feature extraction.First,the aging state of the insulation paper is divided into 10 categories according to the measured polymerization degree.Raman spectroscopy is performed on the oil-paper insulation samples in each aging state.Finally,169 groups of Raman spectra are classified by the K-nearest neighbour(KNN)algorithm and integrated enhanced KNN algorithm.The results indicate that the KNN algorithm after integration enhancement has a stronger recognition ability for the original Raman spectrum,its discriminant accuracy is 98.32%,and it has better stability.It is proved that the discriminant model based on the integrated enhanced KNN algorithm accurately discriminates the original Raman spectra of oil-paper insulation.The proposed model simplifies the diagnosis of the aging state of transformer oil-paper insulation using Raman spectra and is of considerable significance for research on this topic.