Intelligent electric meter fault prediction method based on multi-classifier fusion

Abstract

An intelligent electric meter fault prediction method based on multi-classifier fusion is disclosed. Aiming at the characteristics that intelligent electric meter fault data is large in scale, high in dimension and complex in structure and has errors and abnormal data, missing value filling and abnormal value replacement are performed on an original data set by adopting a normal distribution completion and box diagram method; redundant and irrelevant features are eliminated by calculating correlation coefficients between feature attributes and fault types to form feature subsets; a mixed sampling strategy of over-sampling minority samples and under-sampling majority samples is constructed, and the problem of fault data imbalance is solved. The accuracy of processing the fault data of an intelligent electric meter by a support vector machine (SVM), a BP neural network and a random forest algorithm is calculated, and a confusion matrix representing the performance of each classifier is constructed; and in consideration of the recognition capability of each classifier for different fault types, weights are distributed for each classifier, further a multi-classifier decision function is constructed, and the category with the maximum weight sum is taken as a fault prediction result of a sample.

Description

technical field [0001] The invention relates to a fault prediction method for smart electric meters, in particular to a fault prediction method for smart electric meters based on multi-classifier fusion. Background technique [0002] As an important component of the electricity consumption information collection system, the smart meter undertakes the task of power collection and measurement transmission. With the development of the current society and the improvement of the regional economic level, the coverage of the electricity consumption information collection system continues to expand. The smart meter Faults are characterized by suddenness, difficulty in reproduction, complexity, and multi-facetedness. In addition, due to the different sources of smart meters, the equipment originals and manufacturing processes selected by many domestic suppliers are different, so the types of faults that may occur in smart meters after installation are different. When a fault occurs, ...

AI Technical Summary

Problems solved by technology

Aiming at the complex structure of smart meter fault data, the present invention fills in missing values ​​and replaces outliers in the original data set, eliminates redundant and irrelevant features, forms feature subsets, constructs a mixed sampling strategy, and solves the problem of unbalanced fault data; Construct a confusion matrix representing the performance of classifiers such as support vector machine (SVM), BP neural network and random forest algorithm, assign weights to each classifier, and then construct a multi-classifier decision function, and take the weight and the largest category as the fault prediction of the sample result

Images