Power Transformer Partial Discharge Signal De-noising Based on Variational Mode Decomposition

Abstract

Partial discharge (PD) signals is one of the key means to judge the transformer fault, how to effectively remove interference of PD in the field is the key in PD monitoring. Based on the Hilbert transform, this paper proposes a new algorithm of signal decomposition of Variational Mode Decomposition (VMD) by using the de-noising in transformer PD signal, VMD can decompose signal into several modes, and each of them is compact around a center pulsation. The simulation results show that the proposed method can well remove the narrow-band periodic interference and white noise, and keep the partial discharge signal, verify the validity of VMD. Introduction Power transformer is one of the important equipments of the power system, its reliable operation has great significance for safety, economic operation of power system. The reliability of the power transformer mainly depends on the insulation condition, and one of the main causes of transformer insulation aging and damage is PD, so transformer PD on-line monitoring has high theoretical and practical value to improve power system reliability and economy. One of the main problems of PD on-line monitoring is the interference, including continuous and periodic narrowband interference and white noise which are the most serious [1]. Recently wavelet transform (WT) has been used to PD de-noising, noise and PD signals can be separated into different coefficients by WT. By applying thresholds to the coefficients, the noise will be discarded and the PD signals will be kept for reconstructing the de-noised PD signals.[2] Empirical Mode Decomposition (EMD), an adaptive technique, decomposes a signal into a series of IMFs, which have different frequency scales. But the EMD method has the mixed mode phenomena, so that it’s hard to extract PD signals from narrowband interference [3]. In this paper, a new method for transformer PD signal de-noising is proposed, the proposed VMD method can decompose the input signal into a series of sub-signals (modes), and each mode are tightly around a center pulsation. VMD can filter the white noise out adaptively, and can decompose the rest into a series of modes. Extract the mode containing PD signals then strip the PD signals from noisy signals and achieve a better result of signal de-noising. Brief review of VMD theory and algorithm VMD theory. VMD process is variational problem solving, and involves three important concepts: the classic Wiener filter, Hilbert transform and frequency mixing. Variational problem. We assume each mode to be mostly compact around a center pulsation, and describe variational problem as seeking k modes uukk, minimizing the estimated bandwidth of each mode, and its constraint is the sum of each mode is equal to the input signal f. Specific configuration steps are as follows: 1) For each mode uukk, compute the analytic signal by means of the Hilbert transform in order to obtain a unilateral frequency spectrum: International Conference on Intelligent Systems Research and Mechatronics Engineering (ISRME 2015) © 2015. The authors Published by Atlantis Press 819 ( ( ) )* ( ) k j t u t t δ π + (1) 2) For each mode, shift the mode’s frequency spectrum to “baseband”, by mixing with an exponential tuned k j t e ω − to the respective estimated center frequency: ( ( ) )* ( ) k j t k j t u t e t ω δ π −   +     (2) 3) Compute the squared L2-norm of the gradient of the demodulated signal, estimate the bandwidth, the resulting constrained variational problem is the following: