基于自适应高斯基表示的神经网络在电力系统故障和振荡识别中的应用

Application of Neural Network Based on Adaptive Gaussian Representation in Discrimination of Fault and Oscillation in Power System

结合最优联合时一频处理无交叉项干扰及神经网络自学习分类识别的优点,提出了一种在有色噪声干扰下识别电力系统故障和振荡的方法。将经过自适应高斯基表示(Adaptive Gaussian Representation,AGR)分析处理的电力信号特征向量输入神经网络分类器进行识别。待辨识输入向量不仅表征了原信号的基本信息,而且没有交叉项,运算简单。仿真结果表明,此方法能正确分类识别有色噪声干扰下的系统故障和振荡,提高了电力系统微机保护在系统振荡中检测故障的灵敏性和精确性。

Based on the advantages of non-cross term interference by optimal joint time-frequency processing method and classification discrimination of neural network, the authors propose an approach to distinguish power system fault from oscillation under colored noise disturbance. In this approach the eigenvectors of power signals to be discriminated, which are analyzed and processed by adaptive Gaussian representation （AGR）, are input into neural network classifier. The input vectors to be discriminated can characterize the basic massage of original signals and there are not cross-terms, so its calculation is simple. Simulation results show that the proposed approach can correctly classify and discriminate the fault and oscillation of power system under colored noise disturbance, and the sensitivity and validity of fault detection by microcomputer based protection during system oscillation are improved.