Studies On Fault Diagnosis Of Commutation Failure In Hvdc Systems

HVDC transmission is one of the ultra-long-range, large-capacity power transmission systems, it plays an important role in Chinese economic development and brings enormous economic benefits. Commutation failure is one of common fault in HVDC transmission system, if inexact control measures are taken after commutation failures, it will lead to continuous commutation failure and the direct current power transmission interruption ultimately. Fast and accurate commutation failure diagnosis technology can ensure accurate control measures will be taken, it will be beneficial to the stability of the HVDC system.This paper introduces the domestic and international HVDC transmission system and research results of the commutation failure diagnosis, summarizes the current commutation failure diagnosis approaches and analyses the problems of the existing methods first.Secondly, from the commutation process of HVDC system, analyzes the related factors of the commutation failure, gives the reasons and harms of commutation failure and proposes commutation failure prevention measures.Thirdly, analyses the application of the wavelet analysis and singular value decomposition in fault feature extraction and the existing problems. This paper proposes a new fault feature extraction method used wavelet-singular value decomposition and establishes commutation failure diagnosis model based on wavelet-singular value and support vector machine. PSCAD/EMTDC simulation software is applied to simulate commutation failure of the HVDC transmission system, fault DC currents' wavelet-singular value are selected as learning samples of the SVM. SVM toolkit in MATLAB is used to solve the diagnosis model based on partial binary tree SVM multi-classification algorithm and parameter selection method based on the similarity matrix method. The validity of the method in this paper is verified by training and testing the learning samples.Finally, PSCAD/EMTDC software is used to establish the Tian-Guang HVDC transmission system's detailed electromagnetic model according to the configuration data of it. The SVM's training samples are acquired by simulation experiment and used to train the SVM. The testing samples is selected as the typical faults happened in Tian-Guang HVDC transmission system in recent years, the fault DC current's wavelet-singular values are input the trained SVM to test and output the diagnosis results. The diagnosis results are consistent with the actual situations, the approach's validity and practicality in this paper are all verified further.