Research On State Evaluation Method Of Intelligent Electric Energy Meter Based On Bayesian Network And Emseble Learning

Smart energy meters play an important part in the construction of "smart grids".In order to promote the process of construction of "smart grid",the State Grid Corporation of China has carried out related research on smart energy meters in China.National standards stipulate that energy meters(grades 1 and 2)need to be recycled and tested after using 8 years,and the tested energy meters need to be "two disassembled and two installed" to continue to use,which leads to a huge amount of waste of energy meters every year,and it is particularly important to use more scientific methods of energy meter evaluation.Data fusion,as a technology to assist decision-making by comprehensive analysis of observational information from multiple sources,is widely used in various industries.In order to reduce the waste of energy meter resources,this paper uses the data fusion technology based on the probabilistic graph model to evaluate the operating state of the smart energy meter.The specific content is as follows:(1)During the data preprocessing process,this paper uses the data entropy-based discretization algorithm to process the original data firstly,and then used simulated annealing to select data features to make better use of the data.(2)Learning the structure of Bayesian networks using an improved artificial bee colony algorithm.This paper introduces the idea of simulated annealing on the basis of the improved artificial bee colony algorithm,and uses the thermodynamic formula to accept the suboptimal solution with a certain probability to optimize the strategy of updating the honey source.(3)In order to improve the generalization ability of the model,this paper uses an integrated algorithm to integrate the Bayesian network model based on the improved artificial bee colony algorithm.Based on this,the k-means clustering algorithm is(4)used to selectively integrate the model.In order to better fuse the prediction results of subclasses and reduce the model’s false positive rate,this paper also uses convex function evidence theory to fuse the prediction results of multiple subclassifiers.In this paper,the model is applied to State Grid Corporation’s smart energy meter data.The experimental results show that the improved artificial bee colony algorithm has a better ability to search for solutions in the Bayesian network solution space than genetic algorithms and multi-group bacterial foraging algorithms and mixed fish Group algorithm;and after integrating the improved version of the artificial bee colony Bayesian network model,selective integration and convex function evidence theory fusion algorithm,the model is lower than the LGB algorithm and XGB algorithm and other mainstream algorithms,the false positive rate is lower,The accuracy is higher,a better prediction effect is achieved,and at the same time,it has a better interpretation.