高光谱成像仪对污秽绝缘子进行污秽成分的技术研究

Hyperspectral Imaging Instrument Conducting Technical Research on Pollution Components of Dirty Insulators

摘要: 针对电网绝缘子污秽类型多、污秽成分复杂的问题，研究了基于高光谱成像仪的线路绝缘子污秽程度识别方法。首先，对绝缘体进行高光谱成像，得到400~1000 nm波段的高光谱成像仪，并进行单色校正；此外，对目标区域进行预处理，Savitski-golay平滑，对数导数，一阶导数等。最后，建立基于支持向量机的绝缘子污秽度预测(SVM-ICDP)的绝缘子污秽预测方法和基于偏最小二乘回归的绝缘子污秽度预测(PLSR-ICDP)的绝缘子污秽预测模型。实验结果表明，采用一阶差分变换前提法建立的污秽程度预测模型的计算结果与实测值相差不大，具有较高的可行性。

Abstract: Aiming at the problem of multiple types of pollution and complex pollution components in power grid insulators, a method for identifying the degree of pollution of line insulators based on a hyperspectral imager was studied. First, hyperspectral imaging is performed on the insulator to obtain a hyperspectral imager in the 400~1000 nm band, and monochromatic correction is performed; in addition, the target area is preprocessed, Savitski-golay smoothing, logarithmic derivatives, first derivatives, etc. Finally, an insulator pollution prediction method based on support vector machine (SVS-ICDP) and an insulator pollution prediction model based on partial least squares regression (PLSR-ICDP) were established. The experimental results show that the calculation results of the pollution degree prediction model established using the first-order difference transformation premise method are not much different from the actual measured values, and have high feasibility.