A novel intelligent fault identification method based on random forests for HVDC transmission lines

Abstract

In order to remedy the current problem of having been buffeted by competing requirements for both protection sensitivity and quick reaction of High Voltage Direct Current (HVDC) transmission lines simultaneously, a new intelligent fault identification method based on Random Forests (RF) for HVDC transmission lines is proposed. S transform is implemented to extract fault current traveling wave of 8 frequencies and calculate the fluctuation index and energy sum ratio, in which the wave index is used to identify internal and external faults, and energy sum ratio is used to identify the positive and negative pole faults occurred on the transmission line. The intelligent fault identification model of RF is established, and the fault characteristic sample set of HVDC transmission lines is constructed by using multi-scale S transform fluctuation index and multi-scale S-transform energy sum ratio. Training and testing have been carried out to identify HVDC transmission line faults. According to theoretical researches and a large number of results of simulation experiments, the proposed intelligent fault identification method based on RF for HVDC transmission lines can effectively solve the problem of protection failure caused by inaccurate identification of traditional traveling wave wavefront or wavefront data loss. It can accurately and quickly realize the identification of internal and external faults and the selection of fault poles under different fault distances and transitional resistances, and has a strong ability to withstand transitional resistance and a strong ability to resist interference.

GADF-VGG16 based fault diagnosis method for HVDC transmission lines

“…Compared with traditional intelligent algorithms, the fault diagnosis model based on improved VGG16 constructed in this paper can fully extract deep fault features. Compared with the fault diagnosis algorithm based on SVM, BP, RF and other shallow neural networks in literature [9][10][11][12], the algorithm proposed in this paper has higher fault recognition accuracy, and the experimental results show that it is not affected by transition resistance, fault type and fault. It has strong anti-interference ability and fault tolerance.…”

“…In recent years, a large number of scholars have proposed to use support vector machines (SVM), Back propagation neural networks (BP), artificial neural networks (ANN), random forests and other methods [9][10][11][12] to study the problem of transmission line fault diagnosis. For example, Johnson et al [9] used the SVM classification mechanism to achieve fault identification and classification of HVDC transmission lines.…”

“…However, whether the method can also correctly identify fault types in high transitional resistance and high noise environments remains to be proven. Wu et al [12] used random forest (RF) neural networks to achieve the selection of fault poles and identification of fault types on HVDC transmission lines. But, the extraction of fault features is more complex with this method.…”

Application of Time-Frequency Analysis with Selection of Suitable Mother Wavelets for Transient States of HVDC Transmission Lines

“…The research analyzes the various methods for DC fault detection, localization, and isolation in CSC and VSC-based HVDC transmission systems in two-terminal and multi-terminal configurations [32]. Research conducted that focused on, a feature reconstruction method for bipolar CSC-based HVDC transmission line high-speed fault protection uses the bagged tree ensemble classifier algorithm, and the wavelet transforms [33]. Research suggests a method for finding DC arc faults in VSC-HVDC transmission lines [34].…”