采用实测数据和主成分分析的直流输电线路故障识别方法

由平波电抗器和直流滤波器构成的直流输电线路两端实体电气边界具有高频阻塞作用,使得线路外部故障下,其故障电压起始变化平缓、幅值小;在线路内部故障下,其故障电压起始变化陡峭、幅值大、长时窗时域波形有振荡。利用主成分分析(PCA)方法提取线路内部、外部故障下的极线电压曲线簇样本数据蕴含的此种时域特征信息,并将其投影到主元空间,形成由cPC1和cPC2坐标构成的PCA空间(元件),其线路内、外部故障呈现为具有显著区别的两个不同聚类点簇团,借此可实现直流线路内部故障和外部故障的表征和甄别。故障发生后,利用故障数据于PCA故障识别元件的投影点与PCA识别元件本身多个聚类中心之间的欧氏距离来自适应地判别线路内、外部故障。经大量实测数据试验表明,该方法改善现行以du/dt为核心的直流线路行波保护的性能,若将直流系统历史故障数据复用来增加PCA的聚类点簇,则可继续完善PCA故障识别元件。

Fault Identification Method for DC Transmission Lines Using Measured Data and Principal Component Analysis

As the electrical entity boundary installed at both ends of the high voltage direct current(HVDC)transmission line has the ability to block high-frequency components, the start voltage caused by the external fault changes gently and amplitude is small; the start voltage caused by the line fault changes steeply when amplitude is large and the time-domain waveform is shown shaking violently. Principal component analysis(PCA)is used to extract time domain features of pole fault voltage and project the time domain features on to PCA space and form cluster point clusterings of external fault and line fault constituting cPC1 and cPC2 coordinates. Hence the characterization and distinction of external fault and internal line fault. After fault, the Euclidean distance between the projection values of fault data mapped to PCA space and the cluster centers of PCA identification element is used to distinguish internal and external faults adaptively. The large number of measured data tests show that this method has the ability to interfere without lightning disturbance, samples dithering, harmonic interference and other factors, while improving the protection performance of traveling wave protection with du/dt as the core. And if the historical fault data is reusable to increase PCA cluster point clusterings, PCA fault identification element can continue to improve. This work is supported by National Natural Science Foundation of China(No. 51267009)and National Natural Science Foundation of China-Yunnan(No. U1202233).