中文摘要：

分析了主成分分析用于变压器励磁涌流识别的理论依据。将每组采样数据看作采样空间中的一点，使用主成分分析对采样空间进行变换，找到变压器不同状态下采样数据分布差异最大的一个方向，在该方向上识别励磁涌流和故障电流。分析了常规主成分分析过程用于励磁涌流识别的不足，提出了新的数据预处理方法和主成分分离度的概念，给出了依据分离度大小选择主成分构成励磁涌流识别判据的原则。大量仿真数据计算结果表明基于主成分的判据具有足够的可靠性和灵敏性，性能上优于二次谐波制动判据和间断角判据。

英文摘要：

The theoretical basis of using the method of principal component analysis （PCA） to identify transformer magnetic inrush currents is analyzed. Each set of sample data can be viewed as a point in sample space, the sample space is transformed to a new space using PCA, and a direction in which the distribution difference of sample data is the maximum under different transformer states is determined, and the projection values in the direction of the sample data are used to identify inrush currents and fault current. Based on the analysis of the shortcomings of conventional PCA, the new data perprocessing method and the concept of separating degree are presented, based on separating degree, principal components are selected to establish inrush currents identification criterion. Digital simulations results show that this method has sufficient sensitivity and reliability. Comparing with the secondary harmonic restraint Ddncinle and the dead angle principle, the presented method has better performance.