Research On Intelligent Electric Meters Fault Prediction Technology Based On Data Mining

Smart meter,an important metering device in electric power acquisition system,build the connection between the power grid company and power users.Ensuring the normal operation of smart meters and timely maintenance and replacement of faulty meters are of great significance to safeguard the vital interests of grid companies and power users.At present,the fault types of smart meter are becoming more and more diversified on account of rich functions.It will greatly reduce the operation and maintenance cost of smart meters and improve the maintenance efficiency of smart meters by accurately judging the fault types of smart meters and assisting relevant personnel to make appropriate and reasonable solutions.Based on the collected historical fault data of smart meters and the multi-classification algorithm of data mining,this paper studies the fault type prediction technology of smart meters.The main work of this paper is shown as follows.Firstly,data preprocessing research is carried out on the historical fault data of smart meters.In view of the duplication,absence and abnormality of sample data in intelligent meter fault data set,data cleaning is performed.Combining the correlation analysis between the datasets features and the sample fault type,the features of the smart meter historical fault datasets is screened to reduce the influence of redundant features and unrelated features on the fault classification model construction.A mixed sampling method for imbalanced datasets based on feature correlation analysis is proposed to reduce the model over-fitting phenomenon caused by imbalanced sample sizes of various fault classes from the sample sampling level,and the sampling method proposed in this paper is validated on public datasets.Secondly,the model fusion techniques under different output forms are studied.Due to the limitation of the decision-making mechanism,the single model have insufficient dataset information learning when constructing the electric meter fault classification model.Based on the idea of information complementation,this paper present a multi-classification method of smart meter fault based on model adaptive selection,which solves the problem that the existing model fusion methods can not fuse the two basic classification models whose output forms are the probability values and the class labels of the samples respectively.By using the method proposed in this paper,the model constructed by random forest and LightGBM algorithm with better classification effect of meter fault data is fused,so as to realize complementary information of the model and improve the classification accuracy of the model,and the effectiveness of the proposed fusion method is verified on public datasets.Finally,the base classification model is optimized by combining the feature weights.In order to avoid the influence of the eigenvalue weight of each characteristic attribute of the smart meter fault data set on the model construction,a method of optimizing feature weights by combining genetic algorithm with multi-classification algorithm is proposed.The genetic algorithm is used to optimize the feature weights.Based on the optimized basic classification model,the fault classification model of the smart meter is constructed after two steps of integration.