Research Oon Non-intrusive Family Load Decomposition Based On User Behavior Analysis

Globally,the electric power industry is the basic industry of national economic development and people’s daily life.The orderly operation and efficient utilization of electric power is the key to guarantee the national economy and people’s livelihood.In recent years,both developed countries and China are devoting themselves to the development of smart grid.Smart grid is transformed by computer,communication technology and electronic equipment,which makes the power network more economical,environmentally friendly and efficient.At the same time,the two-way interaction between the power side and the supply side is realized.Load monitoring is the premise of realizing "intelligent power consumption" and"green power consumption".Non-Intrusive Load Monitoring(NILM)can use NILM technology to get specific details of power consumption when only the total power consumption data are given.It can provide users with information and data reference for formulating energy-saving policies,so as to better carry out energy-saving activities.Move.After investigating the achievements of predecessors,this paper finds that most of the existing solutions use high frequency data,which requires higher storage and sampling of monitoring equipment,single load characteristics,and low recognition accuracy for low-power and multi-mode electrical appliances.In view of its shortcomings,this paper continues to carry out research work,puts forward innovative solutions,expounds its basic theory and detailed steps,and uses open data sets to carry out example experiments to verify the feasibility of the method.Firstly,the background and significance of load monitoring are introduced,the advantages and disadvantages of intrusive and non-intrusive load monitoring schemes are compared,and the advantages of non-intrusive load monitoring are explained.Referring to the experience of domestic and foreign scholars,the principle and load characteristics of non-intrusive load monitoring are introduced in detail in the second chapter,and the existing household power data sets and evaluation criteria of commonly used algorithms are given.In the third chapter,a non-intrusive load monitoring method based on behavior habits and neural networks is proposed.While using traditional electrical load characteristics,user behavior information(time)is innovatively used as a new input feature.A clustering-based state labeling method is proposed to solve the problem of pattern labeling for electrical appliances,which is validated by open data sets.In the fourth chapter,a non-intrusive load monitoring method based on Long Short-Term Memory(LSTM)is proposed,which not only considers the current characteristics of the load itself,but also establishes a current template feature library,and makes full use of the power consumption information contained in the historical data of time series data.The decomposition method uses the decomposition model based on LSTM,which can effectively transfer the past information to the calculation of the current moment and improve the accuracy of load identification.