A Hybrid Strategy-Based Filling Method for Power Missing Data

Abstract

The present invention relates to the technical field of electric power data cleaning, and more particularly, relates to a method for filling missing electric power data based on a hybrid strategy. Including the following steps: S1, using the improved k-Means clustering algorithm to cluster the data sets containing missing data; S2, improving and constructing the RBF neural network according to the clustering results; S3, training the RBF neural network, and correcting the missing data Do a filling test. This method better solves the clustering problem of data sets with missing attributes, and combines the clustering results to design the RBF neural network to predict and fill the missing values. This method improves the accuracy of missing data filling, is simple to implement, and has an appropriate calculation overhead. It has a high potential for the problem that a large amount of data generated in the process of power system operation and maintenance is missing or damaged due to factors such as physics and software. Practical value.

Description

technical field [0001] The present invention relates to the technical field of electric power data cleaning, and more particularly, relates to a method for filling missing electric power data based on a hybrid strategy. Background technique [0002] With the development of computer science, more and more traditional industries are combined with computer applications. Under the development trend of big data and artificial intelligence, the research on the power industry has produced more new ideas. A large amount of data will be generated in the process of operation and maintenance of the power system, and the problem of data loss will occur due to the influence of physical factors and software factors in the process of data collection, data storage, analysis and classification. Data loss is a complex problem in many research fields. For data mining, the existence of missing values ​​has the following effects: the system loses a lot of useful information; the uncertainty sh...

AI Technical Summary

Problems solved by technology

[0009] Defects of the above-mentioned patent scheme 1: This method designs a solution based on the traditional k-Means algorithm for the problem of missing data filling, which solves the problem to a certain extent, but does not overcome some defects of the k-Means algorithm itself, and uses aggregate There is no way to learn the inherent laws of the data for data filling, and the filling accuracy needs to be improved

[0010] The defect of the above-mentioned patent scheme 2: This method designs an incomplete data filling scheme based on k-Means clustering and deep autoencoder, which solves the problem to a certain extent, but only considers clustering when filling data The resulting full data weighted average does not model the data itself

Moreover, the use of deep autoencoders and backpropagation calculations will increase the computational complexity of the algorithm

Images