Research On The Automation Method For Non-intrusive On-site Load Monitoring

The effective implementation of demand side management is considered to be an important way to solve the problem of energy shortage.According to the load consumption data,demand side management is capable of dispatching and allocating the demand side resources reasonably,so as to improve the efficiency of power utilization.As a key step of demand side intelligent management,monitoring the load in user domain is an effective means to obtain demand side load data.At present,the commonly used methods of load monitoring are divided into intrusiveness and non-intrusiveness.The intrusive method needs to enter the user and measure the power consumption data.Because the power grid has limited authorization for user load and the measurement equipment is inconvenient during the process of installation,maintenance and management,the intrusive monitoring is difficult to be widely used.Therefore,the non-intrusive load monitoring(NILM)technology has been extensively concerned by researchers.By analyzing the total electric data collected at the power entrance,the technology can obtain the power consumption information of each load in the user,so as to grasp the actual consumption situation and detailed energy consumption.In the actual implementation stage of NILM,the current research needs to enter the user to record the load prior data manually.It interferes with the normal power consumption of users,resulting in the whole monitoring process can not be automatically implemented.To solve this problem,this thesis studies an automation method for non-intrusive on-site load monitoring with user-adaption.The contents include the following three parts:(1)The domestic and foreign NILM researches are summarized and analyzed.Aiming at the existing problems in the actual implementation stage,the basic framework of the automation method for non-intrusive on-site load monitoring with user-adaption is proposed.Besides,the non-intrusive user power data acquisition device is built to obtain the electric data,and the typical electrical signal signatures of residential loads are analyzed.(2)According to the knowledge graph theory,the load signature graph is constructed by signature information around the load type ontology,to form the standard load signature system.Firstly,the waveform information is obtained by load signature extraction.Then,the information is integrated by signature fusion.Finally,the graph is improved by signature knowledge processing.Combined with the constructed graph,the combined classification model of support vector machine is established to distinguish the types of load waveform,and the effectiveness of the load signature graph is verified by the measured data.(3)An automation method for non-intrusive on-site load monitoring with user-adaption is studied.Firstly,the independent load waveform is obtained by decomposition model,and the waveforms are clustered based on the extracted signatures.Then,the Bayesian model is used to quantify the signatures probability,and the type label of load waveforms can be judged to form the specific load signatures library for independent user in the operation process.Finally,based on the data in the library,the current optimization model is established to realize the load identification continuously,so as to obtain the load power consumption in real time.The example analysis shows the effectiveness of the proposed method.