Applications Of Feature Weighted Fuzzy C-Means Clustering And Genetic Algorithm Optimization For Load Identification In NILM Systems

An improved fuzzy clustering non-invasive load monitoring method based on genetic algorithm for feature weight optimization is proposed. The non-intrusive load monitoring research needs to extract the features of electrical appliance waveform data, which has the problems of large number of features...

Full description

Saved in:
Bibliographic Details
Published in:International Conference on Wavelet Analysis and Pattern Recognition (Print) pp. 72 - 77
Main Authors: Li, Peijie, Du, Qintao, Fan, Yiliang, Huang, Yijie, Liang, Xin, Zhang, Weile
Format: Conference Proceeding
Language:English
Published: IEEE 02.12.2020
Subjects:
Clustering algorithms
Data mining
Feature extraction
Feature weight coefficient
Filtering algorithms
Genetic algorithm
Improved fuzzy clustering
Manuals
Non-intrusive load monitoring
Optimization
Wavelet analysis
ISSN:2158-5709
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:An improved fuzzy clustering non-invasive load monitoring method based on genetic algorithm for feature weight optimization is proposed. The non-intrusive load monitoring research needs to extract the features of electrical appliance waveform data, which has the problems of large number of features and redundant features. In order to achieve a good recognition effect, when using Fuzzy C-Means to recognize, the traditional method often needs to filter the features, but this process is complicated and does not fully consider the different influences of different features on the model performance. In this paper, the Fuzzy C-Means algorithm is improved, considering that different features have different influences on the clustering recognition effect, and each feature is given importance weight coefficient. The genetic algorithm is then used to optimize the feature weights in order to find the best model performance Combination of feature weight coefficient. Experimental results show that this processing method can effectively improve the performance of the classifier, and at the same time does not require manual tedious feature selection process.
ISSN:2158-5709
DOI:10.1109/ICWAPR51924.2020.9494385