基于小波包能量熵和DBN的MMC-HVDC输电线路单极接地故障定位方法

可靠精确的故障定位技术对MMC-HVDC输电系统的稳定运行至关重要。针对高阻接地故障的定位精度低的问题,提出了一种基于小波包能量熵(wavelet packet energy entropy,WPEE)和深度信念网络(deep belief network,DBN)的输电线路单极接地故障定位方法。通过对不同条件下故障波形中所包含的故障信息进行分析,运用小波包能量熵提取双端故障电压波形中深层故障特征并构造新的特征矩阵。基于新的特征矩阵搭建DBN模型并通过粒子群寻优(particle swarm optimization,PSO)算法对其模型参数进行寻优,最终利用DBN回归机制实现精确的故障定位。利用电磁暂态仿真软件PSCAD/EMTDC搭建了±250 kV双端MMC-HVDC系统模型并在线路上进行不同位置、不同过渡电阻的单极接地故障模拟仿真。为了对比测试模型的定位精度性能,与基于支持向量机(support vector machine,SVM)的故障定位方法相比较。实验结果表明,该方法在20 kHz的低采样频率下可以精准可靠地定位过渡电阻高达4000Ω以内的直流线路单极接地故障。

Single Pole Grounding Fault Location Method of MMC-HVDC Transmission Line Based on Wavelet Packet Energy Entropy and DBN

Reliable and accurate fault location technology is very important for the stable operation of MMC-HVDC transmission system.Aiming at the difficulty of locating high-resistance ground faults,this paper proposes a method for locating single-pole ground faults of transmission lines based on wavelet packet energy entropy( WPEE) and deep belief network( DBN). By analyzing the fault information contained in the fault waveform under different conditions,the wavelet packet energy entropy is used to extract the deep fault features in the double-ended fault voltage waveform and construct a new feature matrix. Based on the new feature matrix,a DBN model is constructed by particle swarm optimization( PSO) algorithm on the matrix’s model parameters,and finally the DBN regression mechanism is used to achieve accurate fault location. In this paper,the electromagnetic transient simulation software PSCAD/EMTDC is used to build a ± 250 k V double-ended MMC-HVDC system model and simulate single-pole ground faults with different positions and different transition resistances on the line. In order to test the localization performance of the proposed model,the proposed method will be compared with the fault localization method based on support vector machine( SVM). Experimental results show that this method can accurately and reliably locate the DC-side single-pole ground faults with a resistance of up to 4 000 Ω at a low sampling frequency of 20 k Hz.