Research On Intelligent Identification Algorithm For Non-intrusive Residential Load Monitoring

The great demand of the smart grid for the electrical data of the residential load has promoted the vigorous development of the residential data monitoring system.However,the traditional monitoring system represented by intrusive monitoring system has shortcomings in equipment complexity and installation.The research on non-intrusive load monitoring system has gradually become the mainstream.In the non-intrusive load monitoring system,the mixed signal of the power network contains single load signals,so it is important to identify the single load signals from the mixed signal by load identification algorithm,which has great significance to non-invasive load monitoring.First of all,the paper summarizes the researching status of non-intrusive load identification algorithm at home and abroad,studies the advantages of non-intrusive monitoring system,and introduces the architecture and analysis methods of non-intrusive load monitoring system.Using the data collected in non-intrusive mode,transient and steady-state data analysis of residential load was carried out.On this basis,the paper extracts multiple steady state and transient load characteristics,and a non-intrusive residential load identification algorithm based on genetic optimization is proposed.The algorithm based on active power and current effective value,uses genetic algorithm to optimize the different combinations of active power and current values to obtain an optimal solution.If the measured data contains the power and current value of a certain load,the load is running,otherwise the load is not running.The identification algorithm based on feature space division is further studied,and multiple load features are reduced to get a new load feature space.The least square error algorithm is used to design the discriminant function to divide the space for different loads,and identifying the load according to the area to be identified.Finally,based on the characteristics of load frequency domain,load recognition methods based on the periodogram theory and frequency domain matching filter are proposed.On the one hand,the periodogram theory is used to get the frequency response of the load system,and make detailed theoretical analysis and simulation verification for the selection and determination of the load characteristic model.On the other hand,the matched filter only allows the frequency of the load system function to pass,and the rest of the load frequency components are filtered by the filter.Based on this feature,the paper constructs a matched filter,filters the mixed signal,and realizes identification according to the similarity between the filtered output signal and single load signal.Through the research,this paper proposes the non-intrusive intelligent identification algorithms based on load characteristics extraction and based on frequency domain characteristics.Experiments show that the algorithms can effectively identify the load components in residential electricity network,and are simple to implement.The research of non-intrusive intelligent algorithm has certain reference value for non-intrusive load monitoring system.