基于RGB图像生成的CNN-CBAM-BiLSTM配网接地故障分类方法

CNN-CBAM-BiLSTM Distribution Network Ground Fault Classification Method Based on RGB Image Generation

摘要: 准确地识别故障相是实现故障快速隔离、恢复的重要手段。为提高故障选相精度，文章提出了基于RGB图像生成的CNN-CBAM-BiLSTM配网接地故障选相方法。首先利用序列–图像转换方法将馈线的三相电流转换为二维图像对应映射到R、G、B三通道生成RGB图像。然后将生成的RGB图像输入到CNN-CBAM-BiLSTM故障选相模型中，利用CNN提取RGB图像的空间局部特征，进一步利用CBAM自适应聚焦和挖掘RGB图像中蕴含的关键特征信息，最后通过BiLSTM提取RGB图像的时序特征，进而实现高精度的故障选相。仿真结果表明，所提方法具有较高的选相精度，且在网络拓扑和中性点运行方式改变情况下具有较强的鲁棒性。

Abstract: Accurately identifying the faulty phase is an important means to achieve rapid fault isolation and restoration. To improve the accuracy of fault phase selection, a CNN-CBAM-BiLSTM distribution network ground fault phase selection method based on RGB image generation is proposed in this paper. Firstly, the three-phase currents of the feeder are converted into two-dimensional images through a sequence-image conversion method and mapped to the R, G, and B three channels to generate RGB images. Then, the generated RGB images are input into the CNN-CBAM-BiLSTM fault phase selection model. The CNN is used to extract the spatial local features of the RGB images, and the CBAM is further utilized to adaptively focus and mine the key feature information contained in the RGB images. Finally, the BiLSTM is used to extract the temporal features of the RGB images, thereby achieving high-precision fault phase selection. Simulation results show that the proposed method has high phase selection accuracy and strong robustness under changes in network topology and neutral point operation mode.