The Big Data Fusion Algorithm for Online Evaluation of Transformer Metering Error

Abstract

This paper proposes an online evaluation algorithm for transformer metering error based on big data fusion, which combines principal component analysis (PCA) and support vector machine regression (SVR) technology. The algorithm first reduces the dimension of the actual value of the transformer secondary signal through PCA and extracts the main components to establish a principal component model. Then, SVR is used to perform regression analysis on the reduced-dimensional data to predict the actual metering error of the transformer. In addition, this paper also considers the influence of mechanical vibration on the transformer metering error. By introducing the simulation of vibration factors in the simulation platform, the significant influence of vibration on the error is verified. By constructing a simulation model of a 500kV power system and generating relevant error data sets for algorithm training and optimization, the simulation results show that the error evaluation model based on PCA-SVR can achieve high-precision error prediction under different load and grid fluctuation conditions, especially under high dynamic load and grid change conditions. Compared with traditional methods, the proposed algorithm has significantly improved computational efficiency and has real-time evaluation capabilities, which can meet the needs of high-frequency data analysis in smart grids. In addition, the anti-interference ability and dynamic adaptability of the model have also been verified.