Research On Non-intrusive Load Monitoring And Decomposition Based On Deep Learning

In the context of continuous promotion of power reform and construction of global energy Internet,accurate access to user side power information is one of the most critical technologies.However,the traditional intrusion detection method needs to install monitoring devices in the user's internal,which is difficult to deploy.Non intrusive load monitoring system consists of power load information acquisition module,decomposition processing module,communication module,storage module and terminal display module.Just install the device at the entrance of the user's power supply,the operation information of the power equipment can be accurately recorded.When the electrical equipment is turned on,such as air conditioning,the power load information acquisition module senses the change of current and other information,triggers the decomposition processing module to monitor and decompose the load,obtain the load category characteristics and power and other information,and transmit the data to the remote database for storage through the network.The non-invasive load monitoring and decomposition device can be extended to various scenarios including commercial buildings,enterprises and factories,to obtain the active and reactive power data of the internal load,and finally save the information to the remote server through the concentrator and industrial computer.Power companies and other upstream units can conduct data analysis on the received information to truly understand the load details,composition proportion and other information on the user side,so as to provide power suggestions and other services for downstream users.At the same time,the downstream users interact with the upstream power companies through the web page or app to query the power consumption situation,so as to achieve energy saving and safe power consumption.These humanized services will eventually bring convenience to our life and work.This paper first summarizes the research status and application of non-invasive load monitoring and decomposition at home and abroad.Then it discusses the user layer technology: household appliances status recognition,household appliances load decomposition.The deep cyclic convolution neural network is used to identify the status of household appliances,and the gated recurrent units(GRU)and attention mechanism model are used to decompose the load.Then focus on its application in the user layer: smart socket,electricity behavior analysis.When the appliance is inserted into the socket,the socket can identify the type of appliance in a few minutes and judge whether the electricity is used illegally,which can effectively avoid the occurrence of fire;According to the decomposition of the historical operation data of electrical appliances,the user's electricity behavior habits are obtained through data analysis,which reflects the user's health from the side.Finally,this paper discusses the technology of non-invasive load detection and decomposition in the substation level,and proposes a decomposition method based on convolution neural network and support vector machine,which can decompose the total power consumption information into the power consumption information of all walks of life,and expounds its application in the substation level.