Non-intrusive Load Disaggregation And Device Research And Development Based On HMM

In order to help users establish a complete electrical energy monitoring system and achieve accurate measurement of electrical energy consumption,it is essential to analyze the information of household electricity consumption.Compared with the intrusive load monitoring technology that installs sensors at each load,the non-intrusive load monitoring technology only needs installing simple measurement equipment at the electricity entrance to realize the electricity consumption information for the main load of the household load monitoring.This detection mode has the advantages of low investment cost,low actual engineering cost,low implementation difficulty and high operability,which is closer to the actual application of the project and has better engineering application prospects.Therefore,non-intrusive load monitoring has become the mainstream of research in the field of load monitoring.The non-intrusive load monitoring technology is realized through a non-intrusive load disaggregation method: the total power consumption information is collected through related hardware devices and extract load characteristics,obtaining the power consumption information of each electrical appliance by applying the load disaggregation algorithm.Most of the existing non-intrusive load monitoring technologies are based on the edge cloud collaboration method: the terminal hardware collects data,and the cloud is responsible for load disaggregation operations to obtain load disaggregation results.This method needs to consume a large amount of network bandwidth to transmit the electricity signal information read by the user terminal and the disaggregation result information fed back by the server.The timeliness of the algorithm is difficult to be guaranteed.In addition,to transmit the electricity signal from the user terminal to the server side,it is difficult to guarantee the privacy of the users.In view of this situation,this paper proposes an improved load disaggregation algorithm based on Hidden Markov Model(HMM),and develops a set of load disaggregation devices,which can completely achieve non-invasive load disaggregation locally.The main work of this article is as follows:(1)The load disaggregation method based on the hidden Markov model is studied.For the traditional hidden Markov model,it is necessary to manually specify the number of load operating states and it is difficult for the equipment with complex operating states to accurately determine the number of operating states manually.Tus,an adaptive hidden Markov model(Adaptive HMM)is proposed.This model builds a probability distribution function on the load power,analyzing the distribution of the maximum point of the function,and automatically determine the number of operating states of each load and the power corresponding to each typical state..A comparative experimental analysis of the improved method on the REDD dataset is carried out.The results show that the disaggregation effect of the proposed adaptive hidden Markov model is more accurate and superior to the traditional hidden Markov model.(2)Developed a non-intrusive load disaggregation device,using Raspberry Pi as the core computing platform,designing related supporting hardware and software modules,and used the improved load disaggregation algorithm proposed in this paper to complete a parameter setting without manual setting Automatic load disaggregation device: The power information is calculated by collecting the voltage and current signals on the user side,and the load disaggregation step is performed to obtain the load disaggregation result.(3)In order to verify the effectiveness of the device,several common household electrical appliances were selected,their power characteristic values were collected and a load characteristic database was established,and multiple sets of load disaggregation experiments were designed.The analysis of the results showed that the non-intrusive load disaggregation device developed by the thesis has better recognition accuracy,and can effectively achieve user-side load disaggregation.