Open AccessClassification Analysis Method for Residential Electricity Consumption Behavior Based on Recurrence Plot (RP) and Convolutional Auto-Encoder (CAE)
Author(s) -
Zhiqing Sun,
Xinyu Deng,
Xianghai Xu,
W.-S. Hou,
Shouxiang 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 - 10.1088/1755-1315/645/1/012075
Subject(s) - cluster analysis , computer science , data mining , dimensionality reduction , feature extraction , smart meter , dimension (graph theory) , encoder , electricity , redundancy (engineering) , artificial intelligence , pattern recognition (psychology) , engineering , mathematics , electrical engineering , pure mathematics , operating system
Load clustering is the foundation of big data mining in power distribution system. It is helpful for power companies to accurately grasp users’ electricity consumption habits, improve power quality and develop demand response. To overcome the characteristic redundancy problem of the high-dimensional load data, the load clustering method based on RP and CAE is proposed. Firstly, the one-dimensional load curves are converted into two-dimensional recurrence plot to realize feature enhancement. Secondly, the advanced feature extraction capability of CAE is used to realize load feature extraction and dimension reduction. Finally, the spectrum clustering (SC) is used to analyze the user’s electricity consumption patterns. The validity of proposed method is verified by Ireland Smart meter dataset.