Determining faults in high voltage direct current transmission by using numerical protection methods

In this study, different signal analysis and machine learning methods were used to help to predict the location of faults occuring on HVDC transmission lines. The fault current and voltage signals used for fault prediction were obtained by simulation in Matlab. The fault current and voltage signals obtained from each km of transmission line simulation were analyzed by using wavelet transform, Stockwell transform and Hilbert-Huang transform methods. Thus, the distinctive features of transients are obtained. In this methods to reduce the size of feature vectors, different criterions such as; energy, entropy, standard deviation, mean magnitude, the mean of the instantaneous frequency and amplitude are employed. In the fault estimation process, support vector machines which are based on statistical learning theory, artificial neural networks and extreme learning machines methods are used. By investigating impacts of those different signal analysis methods on fault estimation process, optimal fault location algorithm is determined. The obtained results show that the proposed approach is very successful in fault estimation and location process. Key Words: HVDC transmission line faults, discrete wavelet transform, Stockwell transform, Hilbert-Huang Transform, support vector machines, artificial neural networks, extreme learning machines.