Open AccessAn Improved CNN-Based Completion Method for Power Grid Middle Platform Data
Author(s) -
Peng Wu,
Mingsheng Xu,
Cheng Li
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1815/1/012034
Subject(s) - computer science , missing data , convolutional neural network , power (physics) , data mining , key (lock) , grid , pattern recognition (psychology) , artificial intelligence , machine learning , mathematics , quantum mechanics , physics , geometry , computer security
The transmission of power data would be likely to be interrupted or interfered, which results in the middle platform data missing. Missing data reconstruction plays a key role in power data processing, based on which the quality and utilization of power data have been enhanced. In traditional power data filling methods, only a single data distribution was considered, the correlation of power data in time and space was ignored. In this paper, an improved Convolutional Neural Network (CNN) method for filling power data was presented, and a CNN structure was designed. Through the unsupervised training of CNN, this method mines the correlation of data from the dimensions of time and space, and efficiently completes the missing data through the constraints of time continuity and space continuity. The completion results show that our method can fill the missing data efficiently, furthermore, experimental evaluations validate the performs of the proposed method.