To address the frequent occurrence of transient noise and the time-varying and unknown background noise in the online comprehensive error monitoring of electronic instrument transformers and merging units in complex substation environments

an online monitoring method based on the variational Bayesian adaptive Kalman filter (VB-AKF) is proposed. The proposed method enables high-precision online error monitoring under complex electromagnetic and non-stationary noise conditions. First

a dynamic state-space model using in-phase and quadrature demodulation is established based on single-frequency weak signal injection. Second

a variational Bayesian inference mechanism is introduced to iteratively estimate the real-time noise variance and adaptively adjust filter gain

thereby enhancing weak signal tracking accuracy under strong interference. Finally

a high-precision automated testing platform is developed. Experiments demonstrate that even with transient noise and a signal-to-noise ratio as low as -20 dB

the amplitude and phase error monitoring accuracies are maintained within 0.02% and 1'

respectively. The developed prototype fully complies with State Grid Q/GDW 11179 standards regarding 0.2S-class measurement accuracy

electromagnetic compatibility

and protocol consistency.