The time difference of arrival (TDOA) localization method has been widely applied in partial discharge (PD) localization in substations based on an ultra-high-frequency (UHF) antenna array. However, TDOA errors obtained from different UHF antennas can result in a poor localization accuracy. In this paper, a novel nonlinear transform localization algorithm based on regularization for PD in substations is proposed. First, to avoid the problem of difficult solutions and heavy calculations caused by solving nonlinear localization equations, the equations are transformed from nonlinear to linear by eliminating the second-order terms. Then, for the ill-conditioned problem of the localization equations caused by the antenna array positions and time difference errors, the equations are optimized by centralization and row balance to reduce their ill-conditioned degree. Finally, the Tikhonov regularization method is used to solve the localization equations with an ill-conditioned problem, and then, the PD position is determined. The experimental results show that the proposed localization algorithm can improve the accuracy of PD localization in substations and that the localization errors are within 2.23 m.
21 November 2014 in Circuits, Systems, and Signal Processing
With the development of both distribution automation and management system in distribution grid, the fault location methods based on wide-area measurement show great potential. Timely fault location based on wide-area measurements depends largely on the computing time of the algorithm. In addition, the meters layout method also affects the cost and accuracy of fault location. In this article, the efficiency of sparse Bayesian learning (SBL) algorithm and the meter configuration strategy for distributed grid fault location have been improved. Different from the previous calculation for all parameters in the multidimensional vector of Bayesian learning, the proposed method uses the $\boldsymbol {K}$ -means algorithm to extract partial parameters in the vector for update, which improves the computational efficiency without reducing the reconstruction accuracy. Then, for the sparse meter configuration using the principle of sparse recovery, which can be optimized by improving the coherence principle of the sensing matrix, the simulation results show that the improved Bayesian learning of the proposed $\boldsymbol {K}$ -means algorithm significantly improves the computational efficiency, regardless of low- or high-dimensional signals. At the same time, use the low coherence principle of sparse recovery to guide the meter configuration in distribution grid, which could increase the overall fault location accuracy of about 4% compared to recent work. To verify the effectiveness, the performance of the method is tested on IEEE standard 69 bus distribution grids. Simulation results indicate that the suggested strategy performs well for various scenarios.
The detection of partial discharge (PD) is a crucial method to evaluate the insulation status of transformers. The main difficulties of the current localization algorithms are the complexity of the solution and sensitivity to time delay errors. This article proposes a PD localization method in transformers based on linear conversion and density peak clustering (DPC). First, to reduce the complexity of solving the localization equations, the nonlinear localization equations are transformed into linear localization equations by eliminating the second-order terms. Then, to reduce the influence of time delay errors on localization accuracy, the initial localization values are located by multiple acoustic emission (AE) sensors. Finally, the optimal PD coordinates are determined by clustering the initial location values using density peaks clustering algorithm with automatic finding centers (AFC-DPC). The experimental results show that the proposed method can improve PD localization accuracy in transformers, and the average localization error is 5.30 cm.
6 September 2024 in IEEE Transactions on Dielectrics and Electrical Insulation
Different types of partial discharges (PD) in gas insulated switchgear (GIS) cause different degrees of GIS insulation damage, and correct PD identification is critical to GIS insulation status. This paper proposes a multi-model fusion PD pattern recognition method based on one-dimensional adaptive transfer deep residual shrinkage network (AT-DRSN) and improved deep residual network (IDRN), which fully utilizes PD time-domain waveform images generated from field inspection. Firstly, the finite-difference time-domain (FDTD) method and mathematical models are used to simulate GIS partial discharges, and the time-domain waveform image datasets of four typical partial discharge defects are established. Secondly, the network model is built based on time-domain feature sequences and time-order feature parameters respectively, and effectively combined with transfer learning method, and finally, the AT-DRSN and IDRN models are weighted and fused for PD pattern recognition. The results show that the model proposed in this paper can effectively achieve high-precision diagnosis of small-sample GIS PDs, with a recognition accuracy of 98.3%, and the accuracy under strong noise is still greater than 95%, which has higher accuracy and anti-interference performance compared with other methods, and is of reference value for small-sample PD recognition in the field.
S-transform (ST) is an excellent tool for time-frequency filter. There are two factors that influence filtering performance: Inverse s-transform (IST) algorithms and the pass-regions in time-frequency domain. A novel matrix IST algorithm is derived and an adaptive Singular value decomposition (SVD) method for solving the pass-region problem is proposed. The former can avoid reconstructing errors in time-frequency filtering; the latter is effective to distinguish the pass-region of signal from noise. Filter can be realized by removing the smaller singular values and keeping the larger singular values. An additive noise perturbation model is built in ST time-frequency domain and the effective rank of noise perturbation model based on matrix IST is analyzed. Simulation results indicate that the proposed SVD method can provide higher precision than the existing ones at low signal-to-noise ratio and does not need to compute the noise statistics property. Illustrative examples verify the effectiveness of proposed method.