高压直流输电线路单端智能故障定位方法

Single-ended Intelligent Fault Location Method for HVDC Transmission Lines

针对当前高压直流输电线路故障定位方法中存在的问题,提出一种基于S变换组合特征能量和改进一维卷积神经网络-门控循环单元混合神经网络模型的单端智能故障定位方法。首先,对直流输电线路故障电压、电流数据分别进行S变换,提取特征频率范围内的能量;然后归一化电压、电流特征能量并构成组合能量特征向量;最后将组合能量特征向量形成的数据集输入改进的一维卷积神经网络-门控循环单元模型进行训练和测试,实现故障定位。结果表明,该模型具有较高的定位精度和较好的鲁棒性。

Aimed at the problems existing in the fault location method for high voltage direct current(HVDC)transmis⁃sion lines,a single-ended intelligent fault location method based on the combined characteristic energy of S-transform and the improved one-dimensional convolutional neural network-gated recurrent unit(1D-CNN-GRU)hybrid neural net⁃work is proposed.First,the fault voltage and current data of HVDC transmission lines are S-transformed to extract the energy in the characteristic frequency range.Then,the voltage and current characteristic energy is normalized to form a combined energy eigenvector.Finally,the dataset formed by the combined energy eigenvector is fed into the improved 1D-CNN-GRU model for training and testing,thus achieving the fault location.Results show that the proposed model has a high location accuracy and a good robustness.