Pattern Recognition of Partial Discharge Based on VMD-CWD Spectrum and Optimized CNN With Cross-Layer Feature Fusion

Abstract

In order to improve the recognition accuracy of partial discharge (PD) by making full use of the time-frequency characteristics of PD signals and employing deep learning theory, a kind of PD pattern recognition method based on variational mode decompositon (VMD)-Choi-Williams distribution (CWD) spectrum and optimized convolutional neural network (CNN) with cross-layer feature fusion is proposed in this paper. Firstly, a PD signal is decomposed into several components by VMD algorithm, and the CWD analysis of the obtained components is carried out to obtain the VMD-CWD time-frequency spectrum. Secondly, the cross-layer feature fusion and optimization CNN (CFFO-CNN) is constructed by introducing cross-layer connection and optimization algorithm. Thirdly, the VMD-CWD is regarded as the input vector to train CFFO-CNN to learn and extract the intrinsic features of the spectrum. Finally, the trained network is used to recognize the PD types of the test samples. The proposed method is compared with traditional recognition methods such as BP neural network (BPNN) and support vector machine (SVM), as well as some commonly used deep learning algorithms. The experimental results indicate that the recognition performance of the proposed method is significantly better than that of existing recognition methods with accuracy up to 99.5%. It is proved that CFFO-CNN has superior feature extraction ability, which can extract the internal features of the VMD-CWD spectrum independently with higher recognition accuracy and wider application prospect. INDEX TERMS Variational mode decomposition (VMD), Choi-Williams distribution (CWD), feature fusion, convolutional neural network (CNN), partial discharge, pattern recognition.

Partial Discharges Classification Methods in XLPE Cable: A Review

“…According to Gao et al [15], the insulation failure of power cables could occur because of several factors: equipment performance, such as cable defects during manufacturing, insulation deterioration and malfunctions; human effects, such as potential human error, quality of workmanship, installation and handling; abnormal system conditions, such as overcurrent and overvoltage from system maloperation and lightning; damages caused by road digging and thermal, mechanical, and earth movements . In addition, Rohani et al recently stated that insulation degradation occurs due to the aging process, environmental factors, mechanical damages, operational stress and manufacturing defects [16].…”

Localisation of inter-layer partial discharges in transformer windings by logistic regression and different features extracted from current signals

“…In LR, the sigmoid function is used in the relationship between output (y) and input (x) system variables, which models the probability of y belonging to each problem category. Equation (18) provides the logistic model for multiple predictors, where the…”

“…Examples include use of neuro‐fuzzy networks [14], trained in an unsupervised way by orthogonal transformations of PD signals, application of a linear combination of digital filters based on wavelet coefficients and principal component analysis [15], and use of linear regression models based on statistical and energy features extracted from terminal signals [16]. Besides, deep learning techniques have also been applied for PDs pattern recognition, as performed in [17, 18]. Such methods have achieved good performance in PD location and pattern recognition, especially when applied with previously selected attributes.…”

A singular Riemannian geometry approach to Deep Neural Networks I. Theoretical foundations

“…In the last few years Deep Neural Networks (DNNs) has played a pivotal role in machine learning, both in theoretical studies and practical applications. Thanks to the growing computational power of GPUs in the last decade, the domain of applications of DNNs has extended to a wide variety of complex tasks, as speech recognition [1,2,3], machine translation [4,5,6], image analysis [7,8], autonomous driving [9] and pattern recognition [10,11]. The vast majority of the theoretical results about Shallow and Deep Neural Networks are obtained in the Euclidean setting, namely viewing input data as points immersed in R n , for n large enough [12,13,14].…”