Transformer diagnosis using probabilistic vibration models

Identification of the nonlinear vibration system of power transformers

“…To demonstrate the potential impact of the FNN Hammerstein (FNNH) model, its performance is compared with that of the transformer electro-mechanical (TEM) model [4]. The electro-mechanical model, resulting in an noninvasive method and without need of prior measurements under abnormal operation conditions [3,7,8], has been widely used for vibration based transformer monitoring and diagnosis, table 4 shows the MSE between the actual and estimated system outputs by the FNNH model and TEM model at 100 Hz. The significantly reduced MSE for the FNNH model is due to its accurate estimation of amplitude and phase of the vibration signature.…”

“…This led to the development of nonlinear models of transformer dynamics (e.g. a probabilistic model [7] and an artificial neural network model [8]), capable of addressing the vibration at the higherorder harmonics. Ibargüengoytia et al presented the probabilistic relationship between operational conditions (e.g.…”

“…Ibargüengoytia et al presented the probabilistic relationship between operational conditions (e.g. current, voltage, and temperature) and tank vibration at the main harmonics using Bayesian nonlinear modelling techniques [7]. In their approach, statistical models of vibration of a healthy transformer were established for each sensor at several major harmonics.…”