Research On Non-intrusive Household Appliances Load Monitoring Technology

As an important development direction of smart grid,non-intrusive load monitoring technology establishes a communication bridge between power companies and residential users,realizes the friendly interaction,and it has a significant role to improve the safety,reliability and stability of power systems.However,with the continuous advancement of technology,the types of household appliance loads are emerging endlessly,which makes it difficult to identify non-intrusive loads.Especially when identifying loads with similar characteristics,misjudgment often occurs,which reduces the accuracy of identification.In addition,when detecting the load switching events of household appliances,there is a problem that the positioning time accuracy is not high,and it cannot be flexibly used to detect most load switching events.The general process for implementing non-intrusive load monitoring is: collecting raw data,data pre-processing,load event detection,feature extraction,load feature database and load identification.Based on the monitoring process and the above problems,this thesis takes common household appliances as the research object and conducts research on non-intrusive load monitoring technology.The specific contents are as follows:1.First,this thesis discusses the classification of the load characteristics of household appliances,clarifies the calculation method of each characteristic parameter,and studies the implementation principle of non-intrusive load monitoring.In the stage of collecting raw data,this thesis builds a non-intrusive appliance load data acquisition platform in detail,selects representative appliance loads,collects voltage and current data of single load and mixed load,and compares the operating characteristics of appliance loads,and explains the load Identification of feasibility.2.In the stage of data pre-processing,this thesis uses wavelet transform to denoise the raw current data.By comparing the effect of denoising,a wavelet function with better denoising performance is selected.In the stage of load switching event detection,in order to improve the positioning time accuracy,this thesis proposes an improved load switching detection method based on the wavelet detail coefficient ratio and compares it with the traditional event detection method based on current and active power ratio.This method effectively solves the problem of selecting the time window length and has higher positioning time accuracy.3.In the stage of feature extraction,in view of the low accuracy of appliance load identification,based on the extraction of traditional load characteristics,this thesis introduces the wavelet steady-state characteristics and wavelet transient characteristics into the load identification feature quantity,and establishes five types of household appliance loads feature database.In the stage of load identification,this thesis studies the load identification algorithm based on multi-class support vector machines in detail,designs a non-intrusive appliance load identification process.And compares the load identification experiments with different characteristic parameters,it illustrates that the wavelet feature extraction can increase distinguishability,and improve the accuracy of non-intrusive appliance load identification.