Research On Non-intrusive Load Sensing And Identification Method For Industrial Users

Obtaining production energy efficiency information through load monitoring and conducting targeted demand side management is an important means of rationally allocating user production resources and improving energy utilization.Industrial load is an important demand side load resource,and policy of the targets of carbon peak and carbon neutralization also puts forward new requirements for green and low-carbon industrial production.Therefore,it is necessary to monitor the electricity consumption of industrial users,obtain their production details and energy efficiency levels,and assist in production planning and low-carbon transformation of them.Both intrusive load monitoring and non-intrusive load monitoring(NILM)are commonly used load monitoring methods.Due to the fact that industrial loads commonly work in high-voltage power environments and have strict requirements for electricity safety,intrusive monitoring is not suitable for industrial users.Therefore,industrial power consumption monitoring adopts the NILM method,which collects power consumption data from low-voltage side of the user and monitors the power consumption load through data processing,and can restore the power consumption details of industrial production and obtain the energy consumption of the user.Considering that the current research on load monitoring of industrial user mainly focuses on monitoring their switching actions by using the electrical features of loads.But switching behavior of industrial loads is not frequent,and there are fewer features available for identification.It is difficult to effectively restore the specific details of production by electrical features of loads.To solve this issue,this paper studies a non-intrusive load sensing and identification method for industrial users.The main content includes the following three parts:(1)This paper summarized and analyzed the current domestic and foreign research results research achievements in the field of NILM.Aiming at the problems that can be improved in the existing research work,a non-intrusive load sensing and identification framework for industrial users is proposed.Besides,the monitored events and their features are analyzed by using actual industrial information and data,and the principles of load event detection and identification are explained based on the analysis results.(2)According to the manifestation of industrial load events,corresponding event detection and waveform extraction methods are designed for load state changes event and mode changes event.Firstly,considering the fluctuation and noise of power consumption data,the intersection feature of the power curve and its mean value is introduced based on the power features,and the state change event and mode change events of industrial loads are detected through intersection feature difference.Then,according to the manifestation of the event,the steady-state waveform of the state change event and the transient waveform of the mode change event are extracted.The effectiveness of the event detection and waveform extraction method is verified and analyzed through industrial measured data.(3)According to the behavioral features and electrical features of load events,the construction of industrial load feature database and load identification methods based on representation learning models are studied.Firstly,a load knowledge map is constructed based on the prior information of industrial users,and then the knowledge map is vectorized through a representation learning model to build a load feature database.Finally,the identification indicators of behavior features and electrical features are calculated based on the feature of the event,and the identification of load events is realized based on comprehensive analysis to achieve the production information restoration of target users.The effectiveness of the load feature database construction method and the identification method is verified through example analyses.