Non-intrusive electrical appliance load identification method based on CNN and K-means clustering

Abstract

Aiming at the problem that the current non-invasive load monitoring can only identify a single household appliance,and the recognition rate of multiple household appliances running at the same time is low,a non-invasive household appliance identification method based on CNN and K-means clustering was proposed.Firstly,the improved CUSUM edge detection algorithm was used to detect the time of the obtained user power data and extract the power waveform of the load switching event.Secondly,the extracted power waveform was filtered by Gaussian filtering method,and the processed waveform was transformed into pixel image as the load feature library.One part was used as the training set to train the CNN model improved by K-means algorithm,and the other part was used as the test set to test the model recognition accuracy;Finally,the experimental platform was used for actual test and analysis.The experimental results show that the recognition rate of the model for seven kinds of household appliances is 100%,which verifies the effectiveness of the model.The convolutional neural network is improved by K-means algorithm to increase the difference between similar load characteristics and improve the accuracy of load identification,which provides a reference for the development of non-invasive load detection technology.

Keywords

• electrical measuring technology and its instrumentation; non-intrusive load monitoring; convolutional neural network (cnn); k-means clustering; cusum edge detection