Research On Residential Load Modeling And Decomposition Technology For Non-intrusive Monitoring

As an important power consuming end,resident users can respond to the demand of smart grid through scientific planning and allocation of power load,which is an important platform for energy conservation work.As one of the key technologies of load management,load monitoring can provide detailed consumption information to users,guide them to formulate energy-saving schemes,and achieve the goal of reducing energy consumption and electricity expenditure.The traditional intrusive load monitoring has a large economic cost,and the installation and maintenance process will bring inconvenience to users,which is not suitable for the long-term development of power system.Therefore,non-intrusive load monitoring(NILM)has gradually become the mainstream of research because of its unique advantages.It only relies on the mixed signals collected at the entrance of the power system and uses the signal analysis and processing method to realize load identification,which greatly reduces the cost and does not affect the normal life of users.At present,most NILM methods are based on load features,and there are few studies on load mdeling.Besides,most of the algoritnms are still in the stage of theoretical simulation,and lack of practical application.Therefore,this thesis takes NILM as the background,focusing on residential load modeling and decomposition algorithm.The main work can be summarized as the following three parts:(1)The existing NILM methods at home and abroad are summarized,and the shortcomings of existing research are analyzed.The framework of NILM system is designed,and the typical features of residential load are analyzed,which can extract unique and highly identifiable features for each load and provide a basis for subsequent load modeling.(2)Based on the analysis of load features,the model of residential load is established and the principle of model matching is given.This thesis establishes two models for residential load from different perspectives and gives the matching principles respectively.On the one hand,considering the overall structure of the load,the load is regarded as an independent system which can process signals,and the system model is established for it.On the other hand,considering the law of load power consumption,the electricity consumption characteristics of residential load are summarized and summarized,and the electricity consumption model is established.(3)This thesis studies non-intrusive residential load decomposition and proposes three different decomposition algorithms.The first algorithm takes the current signal as the sample,decomposes the mixed current signal by local mean decomposition(LMD),and matches the separated current with the load system model to realize load identification.The other two algorithms take power signals as samples.On the one hand,in order to solve the problem of identification with fewer training samples,a new method of load decomposition based on graph signal processing(GSP)is proposed.On the other hand,in order to increase the practicability of the algorithm and reduce the identification complexity,a method combining NILM practical application scenarios is proposed to reduce the load space,and the method of event detection and matching of power consumption model is used to realize residential load decomposition.Finally,the effectiveness of the proposed method is verified by simulation experiments.