Data for: A CNN Recognition Method for Early Stage of 10kV Single Core Cable Based on Sheath Current

Published: 31 March 2020| Version 2 | DOI: 10.17632/fn38hf353w.2
Contributors:
RUI LIANG, Peng Chi

Description

Here we provide the raw data which is obtained from the simulation in PSCAD/EMTDC software and the processed data which contain 4 cable states and 54 characteristics. The raw data contain 4 rar files, load change.rar, shield grounding.rar, shield grounding after the reduction of insulation.rar, and reduction of insulation between core and shield.rar. There are 153 out files in load change.rar and each of the other 3 rar files has 289 out files. The simulation time of each out file is 0.5s and the sampling frequency is 100kHz. We set 6 fault trigger and load changing moments: 0.073s, 0.147s, 0.210s, 0.284s, 0.357s, 0.440s. The duration of every fault trigger and load changing is 0.04s. The out file contains 10 columns in total. The first column corresponds to the simulation time. From the second column to the last, it is the sheath current of three phase (ABC) in first end of cable and the last end, and the load current of three phase (ABC). The csv file is the processed data. It contains 7038 samples which have been randomly disorganized. Every row is one sample and the first column is the index. The xlsx file shows the results of comparative experiments when activation function, the CNN structure, and the number of iterations are different.

Files

Categories

Electrical Engineering, Power Cable, Recognition, Deep Learning

Licence