Winding condition assessment is an essential task for operating transformers. In this paper, a novel feature extraction method based on vibration analysis is proposed, which converts the vibration monitoring data with load information into a vibration image. Then, a deep learning approach based on convolutional neural network (CNN) is used to classify the images belong to different classes. In the laboratory experiment, free vibration tests are performed on an on-load winding model, which are used to verify the relationship between the natural frequency and the electromagnetic force under different clamping forces. During the field experiment, transformers are divided into three categories, including normal, degraded and anomalous, and the proposed scheme is trained and tested by using the vibration samples acquired from more than 100 operating transformers. The performance of the CNN classifier under different input sizes is investigated, which achieves 98.3% overall accuracy. Besides, the confusion matrices obtained by other methods are compared, such as artificial neural network (ANN), support vector machine (SVM) and naive Bayes classifier (NBC). The results show that the proposed scheme including the vibration image extraction method and the CNN classifier offers superior performance in winding fault diagnosis
3 September 2018 in IEEE Transactions on Instrumentation and Measurement
Windings are the most critical components in power transformers, and the related mechanical degradation assessment has attracted an increasing attention in recent years. In this paper, a novel feature extraction method based on an amalgamation of variational mode decomposition (VMD) and wavelet transform (WT) is encouraged, which correlates the nonlinear structural parameters of windings with the mechanical degradation. First, the winding vibration mechanism is explored, and the relationship between the transient vibration of windings and the structural parameters is studied. Next, the vibrations are decomposed into multiple modes using VMD, and the feature vector is extracted from the modes by means of WT. After that, three classification algorithms are employed, including support vector machine, naive Bayes classifier, and artificial neural network. In the experiment, the winding structure degradation was simulated via adjusting the clamping force. The vibrations from three winding units are used to train and test the classifiers, and their performance is evaluated and compared. The results demonstrate that the proposed feature extraction method with an appropriate classifier is effective for winding degradation assessment.
Study of the vibration mechanisms of transformer windings may lead to useful applications of vibration techniques to transformer online diagnosis. In a power transformer, the clamping force provides a boundary mechanical constraint to ensure the integrity of the winding. Thus, the looseness of the clamping force is an important health indicator for a transformer. Although the effect of clamping force on a winding’s stiffness and natural frequencies is known, the effect of time-varying load current on these natural frequencies remains unsolved. In this paper, this effect is investigated by studying the mechanical frequency response function of an on-load single-phase winding under different clamping forces and variation of the harmonic amplitude of in-service transformers with load current. Then, a gated recurrent unit (GRU) neural network is used to explore the relationship between current sequence and vibration sequence for operating transformers. This study shows that the electromagnetic force induced by the load current affects the vibration response of the winding structure, especially when the looseness of the clamping force is significant. A potential application of the observed phenomenon for online detection of winding conditions is also illustrated.
Weigh-in-motion (WIM) systems are used to measure the weight of moving vehicles. Aiming at the problem of low accuracy of the WIM system, this paper proposes a WIM model based on the beetle swarm optimization (BSO) algorithm and the error back propagation (BP) neural network. Firstly, the structure and principle of the WIM system used in this paper are analyzed. Secondly, the WIM signal is denoised and reconstructed by wavelet transform. Then, a BP neural network model optimized by BSO algorithm is established to process the WIM signal. Finally, the predictive ability of BP neural network models optimized by different algorithms are compared and conclusions are drawn. The experimental results show that the BSO-BP WIM model has fast convergence speed, high accuracy, the relative error of the maximum gross weight is 1.41%, and the relative error of the maximum axle weight is 6.69%.
4 April 2024 in Journal of Magnetism and Magnetic Materials
This paper proposes an operation modal analysis (OMA)-based approach to obtain a transformer’s mechanical condition from vibrations during the transformer’s transient processes, such as short-circuit and power-off conditions. Such processes generate a short transient vibration response, which yields a spectrum with poor frequency resolution. Vibration before the transient process could be included to increase the signal length; however, this would introduce a forced vibration component into the spectrum, making it challenging to distinguish the modal frequencies. To overcome these problems, a time–frequency analysis-based algorithm is proposed to extract the modal frequency spectrum from the vibration signal with the high-frequency resolution, providing clearer insight into the mechanical condition of the transformer. A faulty-state indicator is then proposed based on the similarity between the extracted modal spectrum and the initial modal spectrum obtained under a healthy state. To validate the proposed method, laboratory experiments were conducted under short-circuit and power-off conditions. Two mechanical faults—the looseness of the winding clamping and winding deformation—are introduced. The results show that both faults will cause variations in the modal frequency spectrum, leading to significant decreases in the indicator value. In summary, the proposed method can effectively evaluate a transformer’s mechanical condition in an OMA setting.