基于ACNN的配电网故障识别与定位研究

Research on Distribution Network Fault Identification and Location Based on ACNN

针对目前配电网故障诊断时存在计算时间长、定位精度低的问题,提出了一种基于自适应卷积神经网络的配电网故障识别与定位模型。通过自适应卷积神经网络训练电力数据特征,从而有效提取配电网电力故障特征信息,并基于全连接层对故障进行分类,从而实现端对端的故障检测。通过双端故障定位模型实现故障精确定位。实验结果表明,与DBN模型相比,所提ACNN模型综合性能更优,检测精度提高7.12%时,模型训练时间减少了42.7%。

Aiming at the problems of long calculation time and low location accuracy in current distribution network fault diagnosis,a distribution network fault identification and location model based on adaptive convolutional neural network is proposed.The adaptive convolutional neural network is used to train power data features,effectively extract power fault feature information from distribution networks,and classify faults based on the full connection layer to achieve end-to-end fault detection.Accurate fault location is achieved through a two-terminal fault location model.The experimental results show that the proposed ACNN model has better overall performance compared with the DBN model.When the detection accuracy is improved by 7.12%,the model training time is reduced by 42.7%.