一种基于CNN与FFT‑ELM的输电线路故障识别与定位方法

A method based on CNN and FFT‑ELM for fault identification and location of transmission lines

及时、准确地检测输电线路故障类型与位置是提高电力系统可靠性最重要的问题之一,为此提出一种基于卷积神经网络(convolutional neural networks,CNN)与基于快速傅里叶变换(fast Fourier transform,FFT)的极限学习机(extreme learning machine,ELM)分类模型并行的输电线路故障识别及定位方法。首先,以故障电压时序图作为输入,构建CNN;然后,利用FFT将时域故障电压数据分解,提取各频段的电压峰值与相角作为故障特征样本;接着,以提取的故障特征样本集作为输入,构建ELM网络;最后,通过特征融合层将2个神经网络进行融合,输出故障类型和定位结果。实验结果表明,此方法对输电线路故障识别的准确率为99.95%、故障定位误差在500 m以内、平均误差为263.5 m,可靠性优于其他模型。

It is one of the most important problems in power system reliability to detect the fault types and locations of transmission lines in time and accurately.Th is paper presents an approach for fault identification and location of transmission lines based on convolutional neural networks(CNN)paralled with extreme learning machine(ELM)based on fast Fourier transform(FFT).First,CNN is constructed with fault voltage sequence diagram as input.Then FFT is used to decompose the fault voltage data in time domain and extract the peak voltage and phase angle of each frequency band as fault feature samples.The ELM network is then constructed by taking the extracted fault feature sample set as input.Finally,the two neural networks are fused by the feature fusion layer to output the fault type and location results.Experimental results show that the accuracy of the method is 99.95%,the error of fault location is less than 500 m and the average error is 263.5 m;the reliability of the method is better than other models.