基于混合神经网络的交直流混联电网交流线路故障定位

AC Line Fault Location for AC/DC Hybrid Power Grids Based on Hybrid Neural Networks

摘要:

随着电网交直流混联程度的不断加深，复杂的电网结构导致故障的精确定位愈加困难。针对目前定位方法所存在问题，提出一种基于同步挤压小波变换结合混合神经网络的交直流混联电网交流线路故障定位方法。首先，对交流线路单端电压数据通过同步挤压小波变换处理后，提取其时频矩阵中部分低频区域的能量，构建特征向量；然后，将特征向量构建数据集，并输入至混合神经网络中进行训练与测试，实现故障定位。实验结果表明，所提方法有较高的定位精度与一定的抗噪性能，受不同故障类型与过渡电阻的影响较小。

Abstract:

With the increasing degree of mixing of AC/DC power grids, the complex grid structure makes it increasingly difficult to accurately locate the faults. Aiming at the problems of the current location methods, a fault location method for AC line of AC/DC hybrid power grid based on synchrosqueezing wavelet transform (SWT) combined with hybrid neural network is proposed. Firstly, the single-ended voltage data of AC line is processed by SWT, and then the energy of some low-frequency regions in the time-frequency matrix is extracted to construct feature vectors. Finally, the feature vectors are constructed as a data set, which is input to the hybrid neural network for training and testing to realize fault location. The experimental results show that the proposed method has high positioning accuracy and certain anti-noise performance, and is less affected by different fault types and transition resistances.