Importance of denoising in dielectric response measurements of transformer insulation: An uncertainty analysis based approach

In Data Acquisition denoising is an important issue. In the present work, importance of denoising in the case of real-time data acquisition for transformer condition monitoring using dielectric response measurements in time domain has been studied using uncertainty analysis approach. The uncertainty analysis of the recorded data is also very important for proper decision making in dielectric response analysis of transformers. Detailed descriptions of the real-time dielectric response measurement system (namely, Polarization–Depolarization current and Recovery Voltage measurements) as well as real-life experimentations are given. The paper establishes the fact that the uncertainty due to the noise is more significant than the traditional expanded uncertainty of the data acquisition setup. Experimental results and comparisons of performances of different filtering schemes show that a hybrid-filtering technique could reduce the uncertainty in the acquired data efficiently. A parameter called Uncertainty Envelope is observed to establish this fact. The procedure to calculate the uncertainty envelope for the data corrupted with non-stationary noise and also its importance are explained in detail. Results also show that without denoising and uncertainty analysis of the data one may arrive into a wrong interpretation about the insulation condition in dielectric response analysis.