Research On Degradation Trend Prediction Method Of Smart Electricity Meter Based On FOA-MKSVR

The reliability of smart meter performance is the basis of smart grid accurate measurement.In different climatic environments in China,the performance of smart electricity meters varies greatly under the influence of environment.In this thesis,a multi-kernel Support Vector Regression(MKSVR)model based on the Fly Optimization Algorithm(FOA)for parameter optimization is adopted.The influence of environmental factors and time stress on performance degradation of sm art electricity meters is combined to predict the degradation trend of smart electricity meters,providing a new method to explore the degradation trend of smart electricity meters under the condition of multiple environmental factors.This thesis firstly introduces the research background and significance of the degradation trend of smart meters under the conditions of multiple environmental factors,especially in typical environments.Then analyze the shortcomings of the current domestic research situation and draw out the main content of this thesis.Secondly,the module structure and functional characteristics of smart meters are introduced.Considering the influence of environmental variables on the working state of smart meters,the Pearson correlation analysis method is used to preliminarily extract the environmental variables studied in this thesis,combined with Fuzzy C Means(FCM)The clustering data processing method further extracts the trend characteristics of the degradation data of sma rt meters over time.Then,the basic principle of Support Vector Machine(SVM)is elaborated and extended to Support Vector Regression(SVR)to solve the problem of smart meter degradation trend prediction,and the single-core SVR data prediction based on SVR theory is discussed.method and multi-core SVR data prediction method;Then,in view of the problem that the single-core SVR model is difficult to grasp the overall characteristics of the degradation data of smart meters,a nd the prediction accuracy of the degradation trend of smart meters is not high,a multi-core SVR method is proposed to predict the degradation trend of smart meters.The linear combination of Gaussian kernel functions with strong characteristics is constructed as a new multi-core kernel function,which gives full play to the feature mapping capabilities of different single-core kernel functions.Combined with smart meter degradation data,the method for predicting the degradation trend of smart meters based on the MKSVR model is explored.In the multi-parameter setting problem in the thesis,the FOA algorithm is used to optimize the parameters,and the FOA-MKSVR degradation trend prediction model of the smart meter is established.Finally,based on the smart meter degradation data coll ected by the smart meter experimental base in Xinjiang,the results show that in the case of small samples,compared with the single-core SVR and BP neural network models,the FOA-MKSVR model proposed in this thesis can predict the smart meter degradation trend with higher accuracy and stronger stability.