Open AccessFusion Model of Short Fault Recognition Based on DBN and DNN
Author(s) -
Ming Chen,
Sisi Xie,
Yunan Wang,
Hao Ouyang,
Senlin Lan,
Yefeng Wang
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/632/4/042070
Subject(s) - deep belief network , artificial intelligence , computer science , artificial neural network , backpropagation , pattern recognition (psychology) , matlab , fusion , deep learning , fault (geology) , machine learning , linguistics , philosophy , seismology , geology , operating system
In order to ensure electrical safety, improve the ability to accurately identification of the short fault in the household circuit, we propose a fusion model based on Deep Belief Network (DBN) and Deep Neural Network (DNN). The fusion model consists of an unsupervised pretraining stage and a supervised learning stage. In the pretraining stage, a DBN which denoises the high-dimensional current data and extracts highly abstract features. In the stage of supervised learning, a DNN to learn the relationship between the features extracted and the target classes so that classify the circuit state. And the the initial parameters of DNN are given by pre-trained DBN before the latter stage. Lastly, the model is optimized by backpropagation algorithm, so as to reduce the training time and speed up the convergence speed of DNN. Besides, the current data obtained from MATLAB/Simulink simulation platform is used to train the fusion model and verify its accuracy. Experiments show that the fusion model achieves 95.67% accuracy.