Research Of Non-intrusive Power Load Identification Method Base On ARM Cortex_M3 And SVM

With the rapid development of the economy,the electricity consumption of the whole society is increasing,and the problem of energy shortage is becoming more and more serious.In order to plan the use of electric energy more efficiently and reasonably,the non-intrusive power load monitoring and identification system is applied,large to large data.From large data centers and factories to small home power supply systems,the system can monitor and identify the power load,so that users can know the operating status of their own appliances in time,and provide information support for taking corresponding measures.As the core technology of non-intrusive power load monitoring system,most of the calculations are done by remote servers.There are not many load identification methods based on embedded systems.The existing system has poor real-time performance,and the accuracy of recognition in the complex scene is not high,and scalability are not strong.In response to this problem,this paper proposes a non-intrusive power load identification method based on ARM Cortex_M3 and SVM(Support Vector Machine).This paper designs the structure of the non-intrusive power load identification system based on ARM Cortex M3 and SVM,including data acquisition and processing module,power supply module,load identification and decomposition module and NB-IoT module.The workflow is as follows:firstly,the data acquisition module is used to collect the power parameters of the power grid;the second extraction can be used as the characteristic parameter of the load identification;Then the SVM algorithm is optimized,the SVM multi-classifier is trained in the upper computer,and the prediction part and training model of the algorithm are deployed to Cortex_M3 to realize online identification and power decomposition of the power load;Finally,the analysis results are displayed on the user terminal and uploaded to the cloud server through the NB-IoT.After experiment and test,the non-intrusive load identification method proposed in this paper can realize the functions of monitoring,identification and decomposition.In the experimental environment,the error of data acquisition results does not exceed 1.5%,the overall recognition accuracy of power load is higher than 96%,and the accuracy of server-based load identification is not much different and the error of the power decomposition result does not exceed 2%,which meets the design requirements and achieves the expected goal.The main work of this paper is as follows:(1)A total of more than 20,000 pieces of fourteen single-load and sixteen sets of mixed load power data are collected,and feature parameters are extracted through data analysis.(2)According to the characteristics of feature parameters,the SVM multi-classification algorithm is tailored and optimized to run in Cortex_M3,which realizes embedded online load recognition.(3)Designed the data acquisition and processing methods,and compiled the data acquisition and SVM classification algorithm program code;(4)The system data acquisition,load identification and power decomposition were measured,and the data and results were displayed through the NB-IoT.The main innovation of the thesis is that the SVM power load identification algorithm is successfully deployed in Cortex_M3 to move the power load identification process to the embedded system,which has certain application value.