Study On Non Intrusive Load Decomposition Of Residential Electricity Based On HMM

As the problem of energy saving becomes more and more serious and the development of smart grid,energy consumption monitoring technology has become the focus of attention.Non-intrusive load monitoring is the key technology of power consumption monitoring,the essence of which is load decomposition,unlike intrusive load monitoring,the NILM only installs a sensor at the subscriber's entrance,collects and analysis the total current and the terminal voltage to monitor the power consumption and the working state of each or every type electrical appliances in the room,there is no need to add power data monitoring equipment for each appliance,and because of less investment and convenient installation,it has great research value and application prospect.Based on this background,a non-intrusive load decomposition method based on Hidden Markov model(Hidden,Markov,Model,HMM)is proposed,and using public data sets to take experimental validation,then the accuracy and effectiveness of the method are demonstrated.Firstly,this thesis introduces the principle of HMM,and describes its application in NILM,the five basic parameters:sets of state values,sets of observed values,transition probabilities,observation probabilities and initial probabilities of the model are analyzed,and the significance of these parameters in this method is also introduced.For load breakdown scenarios for multipurpose appliances,using the appropriate state representation method,the task of load decomposition can be implemented without changing the HMM structure.Then,the existing method of state determining by utilizing the statistical law of the electrical appliance data is improved,which make it applicable to a more general situation.A state representation method for multipurpose appliances with different states is proposed,the method is based on binary coding theory,its principle is clear and the calculation is simple.And it is suitable for the load decomposition scene of multipurpose appliance.Thirdly,the proposed load decomposition method is experimentally verified by using public data set AMPds.In order to get generalized results,10-cross test is adopted,and the steady state current of seven kinds of common electric appliance is selected as load feature.Two scenarios are considered respectively:two states of switching and more than two states,the performance evaluation of load identification results is carried out by the basic accuracy calculation method,F-value and its improved method,and the evaluation results are analyzed.Finally,energy consumption is estimated by load decomposition results,the specific power consumption and energy consumption ratio of each appliance in each set of experiments are calculated respectively.Through the analysis,the load decomposition method proposed in this thesis has higher accuracy and effectiveness in the state identification and energy consumption estimation of electrical appliances.