Design Of Non-intrusive Load Monitoring System Based On Steady-state Harmonic Characteristics

Residential electricity consumption behavior monitoring is an important content of smart grid construction,and also an important basis for realizing reasonable distribution of electric energy.Electricity load classification is an important support of residential electricity behavior monitoring,through the load classification to understand residents for users of electricity information,so as to predict the peak period of electricity and total electricity,and the information feedback to the grid system,grid system can finely power supply plan,finally achieve the purpose of saving energy.In this paper,a non-invasive load monitoring system is designed to monitor residents' electricity consumption behavior.This system only needs to install power monitoring equipment at the power inlet bus of the residents,collect current,power and other signals,and provide data support for power load monitoring.However,there are two difficulties in the study of non-invasive load monitoring technology.Second,the non-invasive load monitoring method has a poor classification effect on household electrical load and electrical equipment with similar characteristics.In this paper,the steady-state current harmonic amplitude data are collected and the current harmonic amplitude database is established based on the commonly used residential electricity load.For the problem that the classification effect of low-power residential electricity load and electrical equipment with similar characteristics is poor,this paper puts forward a feature difference enhancement algorithm to process the amplitude data,processes the steady-state current harmonic amplitude data through the enhancement function,strengthens the characteristics of various residential electricity load,and obtains the enhanced database.The probabilistic neural network(PNN)is used as a classifier,and the training network of two databases without intensive processing and after intensive processing is used to obtain the training model respectively,and the two models are applied to the classification of power load.The experimental results show that the steady-state harmonic is taken as the characteristic value of power load classification,and the classification accuracy is more than 88% by using PNN network.The classification accuracy is significantly improved after the intensification of the characteristic value of power load with small load and similar characteristic value,and the best case has reached 100%.