基于拉曼光谱数据处理和谱峰识别的变压器油绝缘老化研究

针对变压器油的拉曼光谱分析通常受到噪声和荧光背景等的干扰以及谱峰位置难以识别的问题，提出了一种改进的数据处理和谱峰识别算法，用于变压器油老化评估时的拉曼光谱分析。提出一种自适应Savitzky-Golay滤波法，引入自适应窗口大小拉曼光谱数据进行去噪处理。采用改进的多项式拟合算法对去噪后的数据进行去除荧光背景处理，减小荧光背景对拟合结果的影响。通过数据点与期望的拉曼信号的接近程度为每个数据点赋予权重，以实现更准确的去荧光背景处理。利用谱峰识别技术判别变压器油的老化程度，采用大小两种尺度高斯窗口判别法来识别谱峰，并结合局部加权信噪比(local weighted signal-to-noise ratio, LW_SNR)来判断疑似拉曼谱峰的真实性。最后通过实验验证了所提算法在变压器油老化评估中的有效性。

Transformer oil insulation aging based on Raman spectral data processing and peak identification

There are problems in that the Raman analysis of transformer oil is usually interfered with by noise and fluorescent background, and it is difficult to identify the position of the spectral peak. Thus this paper proposes an improved data processing and spectral peak recognition algorithm for the Raman analysis of transformer oil aging evaluation. An adaptive Savitzky-Golay filtering method is proposed, and adaptive window-size Raman spectral data is introduced for denoising. An improved polynomial fitting algorithm is used to remove the fluorescence background processing of the de-noised data to reduce its influence on the fitting results. Each data point is weighted according to the distance between the data point and the expected Raman signal, so as to achieve more accurate de-fluorescence background processing. The aging degree of transformer oil is identified by spectral peak recognition technology, and the spectral peak is identified by the Gaussian window discrimination method with two scales, and the authenticity of the suspected Raman spectral peak is judged by the local weighted signal-to-noise ratio (LW_SNR). Finally, the effectiveness of the proposed algorithm in transformer oil aging evaluation is proved by experiment.