电气设备在线监测系统动态误差来源分析方法

摘要

On the basis of the dynamic testing error modeling theory , the characteristics and source of the dynamic testing error are analyzed for on -line monitoring system of electrical equipment .Aiming at the disadvantages of Fourier transform and wavelet transform in dynamic measurement error , a new method combining empirical mode decomposition ( EMD) with Fisher discriminant analysis method is put forward for dynamic measurement error decomposition .First of all, EMD method is used for adaptive measuring error signal decomposition .Then, the time domain auto-correlation, cross-correlation and frequency domain characteristic information of the decomposed intrinsic mode function (IMFs) are extracted, and the feature space is constructed.Finally, the Fisher distance criterion is introduced to classify the feature space and determine the source of each decomposed IMF .In order to verify the effectiveness of the proposed method , the additional error experiments are carried out on the UHF partial discharge monitoring system , and the proposed method is used to decompose the total error signal of the system.Experiment results show that the proposed method can effectively find the source of the error , and it is highly feasible and applicable .%运用动态测试误差建模理论，对电气设备在线监测系统的误差特性以及误差来源进行分析。针对傅里叶变换、小波变换等方法在分解动态测量误差时存在的不足，提出经验模态分解法与Fisher距离判别算法相结合的方法。首先，利用经验模态分解法对测量误差信号进行自适应分解；其次，提取每条分解子曲线的时域自相关、互相关以及频域特征信息，构建特征空间；最后，引入Fisher距离判据对构建的特征空间进行分类，最终确定每条误差分解子曲线的来源。为验证所提方法的有效性，对超高频局部放电监测系统开展附加误差实验，并用所提方法对系统的总误差信号进行分解与溯源。结果表明，该方法能够有效地追溯到误差产生的源头，具有较强的适用性和可靠性。