基于绝缘油拉曼光谱的聚合度定量评估方法

Quantitative Evaluation Method of Degree of Polymerization Based on Raman Spectra of Insulating Oil

为实现准确、高效地油纸绝缘老化诊断,该文提出一种油纸绝缘老化拉曼光谱特征提取及聚合度预测方法。通过油纸绝缘加速老化实验,获取了涵盖油纸绝缘老化全寿命周期的老化样本。利用自主设计的拉曼光谱检测平台对其进行了拉曼检测,在老化状态标签的监督下,基于线性判别分析算法提取了能够反映油纸绝缘老化的拉曼光谱特征。通过支持向量回归算法构建了基于绝缘油拉曼光谱的聚合度预测模型,对35个实验室测试样本聚合度预测的均方根误差仅为32.26,相关系数平方达0.987。该文建立了覆盖油纸绝缘老化全过程的油纸绝缘老化拉曼谱图库,揭示了所提取的拉曼光谱特征与油纸绝缘老化的关联关系,为油纸绝缘老化拉曼光谱诊断提供了有力的支撑。

To achieve accurate and efficient aging diagnosis of oil-paper insulation, a method of feature extraction and aging diagnosis of oil-paper insulation by Raman spectroscopy was presentd in this paper. Aging samples covering the whole life cycle of oil-paper insulation aging were obtained by accelerated aging experiment of oil-paper insulation. A Raman spectrum detection platform was designed to detect them. Under the supervision of aging status, the Raman spectral features of oil-paper insulation aging were extracted based on linear discriminant analysis algorithm. A Raman aging diagnostic model of oil-paper insulation aging was established based on the support vector regression algorithm. The root mean square error of the prediction for the degree of polymerization of 35 laboratory test samples is only 32.26, and the square of the correlation coefficient is 0.987. A database of oil-paper insulation aging Raman spectra covering the whole process of oil-paper insulation aging has been established, and the relationship between the extracted Raman spectral features and oil-paper insulation aging has been revealed, which provides a strong support for the oil-paper insulation aging diagnosis based on Raman spectroscopy.