Study On Key Technologies In Non-intrusive Residential Load Monitoring For Intelligent Power Utilization

Monitoring and analysis of appliances state and electricity behavior of power consumers provide data support to fine power service such as demand response,energy efficiency management and electrovalency policy,is a key link in realization of flexible and interactive intelligent power utilization.At present,load state monitoring data mainly depends on installation of arge number appliances state monitoring device inside of consumers.This method has certain interference to user's production and life and brings economic problems to grid caused by high development costs,it's difficult to popularize in resident consumers.Key technologies in non-intrusive residential load monitoring(NILM)for intelligent power utilization are studied in this paper,to get accurate power load data and realize non-intrusive monitoring of residential load by solving the above technical problems in load data compression and transmission,load feature extraction and recognition on the base of data acquisition and processing capability of existing power system.Facing the technical requirement in mass meticulous load data transmission,accurate identification of wide variety of electrical load and electricity service behavior mining,the study on key technologies in non-intrusive residential load monitoring for intelligent power utilization in this paper includes four aspects,which are NILM data compressed sensing and transmission method,NILM feature extraction method,NILM identify algorithms and non-intrusive abnormal electricity behavior detection.Specific contents are as follows:Aiming at the requirement of data sampling frequency and communication efficiency for NILM,data compressed sensing(CS)model is established by analyzing the feasibility of power load data compression and build data sparsity and observation matrix.A NILM data reconstruct method is presented the of non intrusive load monitoring,and improved iterative threshold reconstruction method is used to reconfiguration the compressed sensing data.Wireless sensor network(WSN)optimal clustering model for NILM data is built based on CS by analyzing quantitative relation between WSN cluster routing and data compression ratio.Optimal network size and cluster number are calculated.Then an optimal WSN clustering routing algorithm for NILM compressed data is proposed to realize high efficient transmission of NILM data.To solve the problem of overlap of some electrical equipment load characteristics in NILM,a method of extracting reactive current harmonic load characteristics is presented in this paper.In this method,the sampling current is divided into two parts: active and inactive,and the reactive current is analyzed and processed in frequency domain,so as to improve the characteristic difference of similar electrical load.Aiming at the problem that in some electric power scenario load characteristics of low power appliances are obscure and difficult to identify,a method of extracting the difference features based on improved fuzzy clustering is proposed.The method first extracts the difference characteristics of the load,and then determines the number and type of electrical equipment according to the inter cluster entropy and improved fuzzy clustering.NILM has a high demand for accuracy and timeliness of identification algorithms.To solve the problem of the accuracy of single feature selection in load decomposition and recognition,a load identification algorithm based on multi feature genetic objective function optimization is proposed.A multi feature optimization model is established through genetic encoding of the load state and introduced the new features of the load to improve and optimize the genetic objective function.Then accurate decomposition and recognition of different electrical state change are realized by genetic iterative.In order to overcome the shortcomings of the recognition accuracy and efficiency caused by the random weights and thresholds of the traditional neural network algorithm,a genetic optimization based neural network for NILM is proposed.By establishing neural network genetic iterative optimization model,the network training error is used as the objective function of genetic iteration,and the weights and thresholds of neural network are optimized to obtain better network performance and load identification effect.On the basis of non intrusive load data transmission and monitoring identification,deep excavation of the load status data of the non intrusive load monitoring is studied in this paper,and application of the intelligent load behavior analysis is realized.This noninvasive detection method for abnormal electrical behavior based on NILM results reflect the user behavior and the electricity load change electricity behavior model,put forward a feature selection strategy based on correlation evaluation.Extreme learning machine algorithm is used to train and learn optimized selection features in order to realize the detection of abnormal power consumption.