| Autor: |
Sun, Xuelei, Tian, Xingjun, Wang, Hexiang, Wang, Nanqing, Lu, Ning, Song, Jinchuan |
| Předmět: |
|
| Zdroj: |
International Transactions on Electrical Energy Systems; Dec2021, Vol. 31 Issue 12, p1-25, 25p |
| Abstrakt: |
Summary: This article studies the identification and classification of atypical load current in metro DC traction system. Due to the diversity of metro vehicles and the variability of operating conditions, the DC traction system presents nonlinear dynamic features. During the vehicle braking process, the divergent current usually appears with a small frequency. Actual situation shows that the divergent current belongs to an atypical load current, and it is not a fault current. However, the divergent current has a strong similarity with remote short‐circuit current in terms of current increment and slope, which leads to di/dt‐∆I (DDL) protection error tripping. DDL protection cannot identify and classify atypical load currents, which indicate that the existing protection scheme needs to be further improved and optimized. In order to identify and classify atypical load current quickly and accurately, a hybrid algorithm is proposed by variational mode decomposition (VMD) multidimensional entropy and kernel extreme learning machine (KELM). The energy entropy can effectively identify short‐circuit current and atypical load current, and the setting threshold is 96. Making energy entropy and sample entropy as eigenvectors, a KELM model is applied to classify atypical load current and the accuracy is 97.5%. RT‐plus test platform based on closed‐loop system verifies the effectiveness of the new hybrid algorithm, which is competent to be a backup algorithm for feeder protection with strong immunity to improve the safety and reliability of DC traction system in a metro. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
