基于FastICA的电流传感器相位差测量方法

电流传感器的相位差易受环境的影响，为提高电力绝缘在线监测系统的可靠性和准确度，文中提出了一种在线监测电流传感器相位差的测量方法. 该方法引入独立分量分析（independent component analysis,ICA）对电流传感器的输出信号进行分离. 给出一种补偿观测信号与源分量数目的方法，建立ICA 的数学模型. 针对FastICA 算法每次分离结果误差不同的局限性，用一个关于混合矩阵的评价函数选取多次分离结果中相位测量误差较小的结果. 实验结果显示：对于信噪比为10.9 dB 的信号，评价函数能使相位测量误差小于0.06± 的样本接受率从51.4% 提高到81%.

Phase Difference Measurement for Current Sensor Based on FastICA

Phase difference of a current sensor is susceptible to environmental impact. In order to improvereliability and accuracy of the on-line insulation monitoring system, a phase difference measurement methodfor current sensor on-line monitoring is proposed. In the method, independent component analysis (ICA) isused to extract the test signal from the output of current sensor. An ICA model is established, and a methodgiven to compensate the number of observed signals and its components. To solve the problem that phaseerrors are different in different test runs, an evaluation index of the mixing matrix is proposed to find the mostaccurate result. Experimental results indicate that for a signal with SNR=10.9 dB, the percentage of phaseerrors below 0.06± is raised form 51.4% to 81% by using the evaluation index.