Research On Diagnosis Methods Of Abnormal Electricity Consumption Based On Data Mining

With the development of our country's economy,energy shortage has become a very serious problem.The basic industry of electricity,as an important part of the energy industry,has increasingly become a topic of national concern.Driven by interests,many abnormal electricity activities,such as electricity theft and leakage,continue to occur.These actions will not only seriously threaten economic development,but will also disrupt normal power supply and use of electricity.Although the power supply companies are constantly improving their own scientific and technological level,and various computer communication network technologies are applied in power operation management,the traditional methods are still relatively backward.The power system storage equipment has accumulated a large amount of energy metering data.It has great significance in improving the poor monitoring results of frequent defaults and misjudgments of traditional methods to make full use of these data,and analyse these data through methods for data mining and analysis to assist power companies to grasp the characteristics of user power consumption and abnormal power consumption behavior.In response to these problems,this paper develops an analysis and research of abnormal power consumption behavior diagnosis methods based on data mining methods.(1)Preliminary exploration and analysis of abnormal electrical characteristics.This paper first discusses the related data mining technology theory,and analyzes the application situation in its power field,and constructs an abnormal power consumption detection closed-loop application mechanism based on data mining.From the aspects of causes and methods,it made a correlation analysis of abnormal electricity consumption behaviors,and initially explored the characteristics of abnormal electricity consumption,and analyzed the differences in the characteristics of abnormal electricity consumption and normal electricity consumption.(2)Preliminarily screening of suspected users of abnormal electricity based on FCM cluster analysis.For the problem of selecting the optimal number of clusters,this paper proposes a cluster effectiveness evaluation index which called WCCVI based on weighted combination,and uses this index to determine the optimal cluster number,and obtains the cluster number when the index reaches the optimal value.Based on this,the power load was used for cluster analysis and the user typical load characteristic curve was obtained.In addition,this paper also proposes a weighted form of similarity measure,combining Euclidean distance and correlation coefficient,and uses it as an indicator to examine the similarity between the load characteristic curve of the user to be detected and the typical characteristic curve.The set threshold value is used to determine whether the user is suspected of having abnormal power usage,thereby screening users with abnormal power usage.(3)Abnormal electricity diagnosis based on fuzzy neural network.In this paper,a fuzzy neural network algorithm that is more suitable for the research content of this paper is selected,and an abnormal power consumption evaluation index system is designed.Then,based on the preliminary screening of suspect users through FCM cluster analysis,a diagnosis model is built using fuzzy neural network and model training is performed.test.At the same time,the diagnostic results were analyzed and evaluated,and compared with other models.After testing,it is found that the abnormal power consumption diagnosis program based on FCM load clustering analysis and FNN diagnosis model has an ideal effect on the identification of abnormal power consumption behavior.The error rate and misjudgment rate are all ideal wthin the acceptable range,and better than the effect of other methods.