Study On Non-Intrusive Load Identification Method Based On Behavioral Signatures

Non-Intrusive Load Monitoring(NILM)technology is able to decompose the energy consumption and operation status for a certain power supply area by monitoring and analyzing the voltage and current data from a centralized power supply point and provide energy management system with valid basic energy management information.The current NILM technology,along with the problems of high requirements for the sampling devices and large computation,is hard to be applied to the huge amount of residents.To fix this problem,the paper surveys basic signatures of residents’ consumption,proposes a variable to represent such signatures.The statistical description method of load signature parameters is proposed according to the uncertainty signatures of load behavior and electrical parameters.Besides,a non-intrusive load identification method based on behavioral signatures which can be applied to smart meters is proposed.The main research contents and results are as follows:Firstly,the paper proposes the signatures to represent load behavior according to the signature of residents’ load.Without extra data,such signatures can be directly computed by sampled current and voltage data from a centralized power supply point with its sampling time.Load behavior signatures can be combined with simple electrical signatures to provide high resolution for different load,which avoids identifying detailed electrical signatures(such as harmonic,transient etc.)for the current method.Secondly,the paper proposes a statistic method on the signatures above according to consumption behavior,power and multifarious electrical signatures of a same kind load in residents’ load.To make the expression of load signatures truthfulness,this paper uses probability function to represent the signatures of different load and membership function to describe the importance in confirming different load.Then propose a method of load identification based on this.Thirdly,because of the problem of power pulse and ripple noise affecting the identification and matching of load on-off events from original sampled data,the paper proposes the event detection algorithm based on power increment and matching algorithm based on power increment and current waveform to improve the accuracy of event detection and matching,effectively.Finally,the method is validated by the international BLUED dataset,laboratory data and actual household data.The success rate of event detection is above 99%,the success rate of event pairing is over 92%,and the success rate of load identification is over 85%.As a conclusion,it has the advantages of low sampling rate,easy extraction of load features,and low computational load,covering the recognition of common household appliances.And the algorithm of this paper can be implemented in smart meters by being verified in MATLAB.