Abstract: In view of the characteristics of multi-correlation of infrared thermography of polluted insulators, a method based on partial least squares regression analysis is proposed to determine the pollution class of high voltage insulators. On the premise of preserving the original data to the maximum extent, the partial least squares regression equation between the characteristic parameters of high-voltage insulator contamination and contamination grades is built, and through analyzing the importance of  indicators of variable projection of the regression model equation, the effect degree of various characteristic parameters on contamination grades is obtained. The method solves the multi-correlation problem of the independent variables effectively, and quantifies the relationship between the characteristic parameters and contamination grades. The test result shows that, judging the high-voltage insulator contamination grades by applying partial least squares regression analysis is scientific and reliable, with high accuracy and strong practicability.

Key words: insulators pollution class, characteristic parameters, Partial Least Squares（PLS） regression, model equation, variable importance in projection

摘要： 针对污秽绝缘子红外热像特征数据具有多重相关性的特点，提出基于PLS(Partial Least Squares, PLS)回归分析的高压绝缘子污秽等级判定方法。在最大限度保留原有数据信息的前提下，建立起高压绝缘子污秽特征量与污秽等级之间的PLS回归模型方程，通过对回归模型方程进行变量投影重要性指标分析，可以得到各个特征量对污秽等级判定结果的影响程度。此方法有效解决了自变量之间的多重相关性问题，量化了污秽特征量与污秽等级之间的关系。测试结果表明，将PLS回归分析应用于高压绝缘子污秽等级的判定，科学可靠，准确率高，具有较强的实用性。

关键词: 绝缘子污秽等级, 特征量, 偏最小二乘（PLS）, 模型方程, 变量投影重要性指标