Study On Decomposition Model Of Power Load With Nonintrusive Method Based On Improved HMM

The development of smart grid technology has reduced impact of energy consumption on environment.The smart grid’s power load monitoring technology realizes remote real-time monitoring of users’ electricity consumption.Among them,the non-invasive load monitoring(NILM)technology can realize the internal power system by analyzing the power load data at the power entrance.The automatic monitoring of the operating status of electrical equipment and the consumption of electrical energy is of great significance for the economic operation and dispatch management of smart grids.It has become a research hotspot in the field of NILM to use the constantly developing hidden Markov model(HMM)and its extended models to solve the load decomposition problem in NILM.The basic idea is that an NILM problem can be expressed as a decoding problem of HMM.By using the decoding algorithm in HMM,the state of each power device in the power system can be obtained by solving the total monitored power data obtained.At the same time,the NILM method based on HMM also has the problems of not considering the influence of the duration of the resident state of the power equipment and the interference generated from unmodeled power equipment.The main research work is as follows:(1)The standard Viterbi algorithm has the problem that the number of states that need to be calculated and traversed during decoding progress is large.The state of the power equipment will not change between two consecutive load events in the power system,by using this feature,that the number of calculating times and traversing times is reduced,thereby this thesis improved the Viterbi algorithm;(2)There exists a problem in the NILM method based on HMM which does not consider the influence of the duration of the the power equipment’s state and the recognition result will be interferenced by the unmodeled power equipment.This thesis proposed the factor hidden semi-Markov model(FHSMM)to solve the NILM problem,this method can fully consider the state residence time in a power device,and can accurately identify multiple power devices of the same type of power equipments,while not requiring to model all the power equipments,and greatly reduces the calculation difficulty of the power load model;(3)By using synthetic load data and the public BLUED data set to compare the improved Viterbi algorithm and FHSMM.Experimental results show that the improved Viterbi algorithm has lower time complexity and wider application scenarios.Experimental results show that,compared with the AFAMAP algorithm,the FHSMM method,the NILM method based on FHSMM is superior to the AFAMAP algorithm in stime efficiency and recognition accuracy.