Research on Online Monitoring of Transformer Short Circuit Impedance Based on VMD Combined with VFFRLS

Abstract

Aiming at the problems of significant non-stationary characteristics 
and complex strong noise interference in the electrical signals of the primary and 
secondary sides of transformers during actual operation, to achieve high
precision, real-time and reliable online impedance monitoring, this paper 
proposes a joint signal processing and parameter identification scheme 
integrating Variational Mode Decomposition (VMD) and Variable Forgetting 
Factor Recursive Least Squares (VFF-RLS). First, VMD is used to adaptively 
decompose the noisy voltage and current signals at multiple scales, which 
effectively suppresses noise and accurately extracts pure fundamental 
components. In the signal preprocessing stage, the superiority of VMD in 
fundamental component extraction and anti-interference is verified through 
comparative analysis with the method combining wavelet threshold denoising 
and windowed interpolation. Subsequently, based on the classic T-type 
equivalent circuit model of the transformer, the VFF-RLS algorithm is adopted 
to complete the online identification of dynamic short-circuit impedance. 
Simulation results under three representative operating conditions and with SNRs 
of 40–60 dB show that the short-circuit impedance identification error can be 
maintained within 1%, demonstrating high accuracy and stability. This scheme 
provides an efficient and feasible technical support for the health assessment of 
transformer insulation and winding status, early fault warning, and safe operation 
and maintenance.