东北大学信息科学与工程学院，辽宁省 沈阳市 110819
College of Information Science and Engineering, Northeastern University, Shenyang 110819, China
In view of the fault characteristics of multi-terminal HVDC (MTDC) transmission lines, such as rapid rising speed， large peak value of fault current and difficulty in fault location, a fault diagnosis method for MTDC system with both rapidity and accuracy is proposed. Firstly, the amplitude and frequency characteristics of fault signal waveforms of MTDC transmission line faults are analyzed. The extraction methods for fault amplitude and frequency features of MTDC transmission line faults are studied separatedly based on amplitude variation characteristics of signal waveforms and wavelet packet analysis.Then, the fault diagnosis method of MTDC transmission system based on amplitude-frequency characteristics is formed. Secondly, the parallel convolutional neural network (P-CNN) with fault classification and fault location branch is constructed, and the training method of P-CNN based on transfer learning is proposed. Finally, the simulation verifies that the fault diagnosis method of MTDC system based on P-CNN meet the fast requirements, and the parallel structure is more accurate and expandable than other artificial intelligence fault diagnosis methods.
WANG Hao,YANG Dongsheng,ZHOU Bowen,et al.Fault Diagnosis of Multi-terminal HVDC Transmission Line Based on Parallel Convolutional Neural Network[J/OL].Automation of Electric Power Systems,http://doi.org/10.7500/AEPS20191124003.