Fault Prediction of Intelligent Electricity Meter Based on Multi-classification Machine Learning Model

Abstract

Aiming at the characteristics of intelligent electricity meter fault data, such as large scale, high dimension, complex structure, error and abnormal data, a fault prediction method of intelligent electricity meter based on multi-classification machine learning model was proposed. Firstly, the normal distribution completion and box diagram method were used to fill the missing values and replace the outliers in the original data set. By calculating the correlation coefficient between feature attributes and fault types, redundant and irrelevant features are eliminated to form feature subset. To solve the problem of unbalanced fault data, a mixed sampling strategy was constructed, which oversampled a few samples and undersampled a majority of samples. Secondly, the prediction accuracy of intelligent electricity meter fault data processed by support vector machine (SVM), BP neural network and random forest algorithm was calculated, and the confusion matrix representing the performance of each classifier was constructed. Considering the recognition ability of each classifier for different fault types, weight is assigned to each classifier, and then the multi-classifier fusion decision function is constructed. Finally, public data sets and actual electricity consumption data are used as samples to verify the effectiveness of the proposed method.