基于LSTM-FCN神经网络的船舶电力系统故障识别方法

Fault identification of ship power system based on LSTM-FCNneural network

船舶电力系统拓扑结构日趋复杂,故障种类繁多且不易区分。为确保继电保护动作的正确性,本文基于船舶电力系统故障录波数据,利用全卷积网络(Fully Convolutional Network,FCN)在局部特征提取上的优势,以及长短期记忆网络(Long Short-Term Memory,LSTM)在时序特征提取上的优势,提出了一种基于改进LSTM-FCN网络的故障诊断模型,并应用于船舶电力系统故障识别。依托PSCAD/EMTDC仿真软件对典型船舶电力系统各种故障进行仿真,通过小波变换对采样信号进行预处理。实验结果表明:本文所提出的故障诊断模型能够很好地对船舶电力系统故障进行分类识别。

The topological of ship power system is becoming more and more complex,and the potential fault threat it faces is getting more complex and difficult to distinguish.In order to ensure the action of relay protection,based on fault recording data,this paper proposes a fault diagnosis model based on improved LSTM-FCN neural network and applies to ship power system fault identification,which utilizes fully convolutional network(FCN)in local feature extraction and long short-term memory network(LSTM)in temporal feature extraction.Based on PSCAD/EMTDC simulation software,various faults of single-bus ship power system are modeled,and the sampled signals are preprocessed by wavelet transform.The experimental results show that the fault diagnosis model can identify ship power system faults with high accuracy.