点击次数：

发表时间：2022-10-09

发表刊物：电机与控制学报

卷号：23

期号：2

页面范围：53-61

ISSN号：1007-449X

摘要：Focusing on the issue of open-circuit fault diagnosis of power switches in the neutral point clamped(NPC) three-level STATCOM, this paper proposes a hybrid fault feature extraction technique. By analyzing the change law of output current after power switch open-circuit fault of the three-level STATACOM, the normalized average current feature extraction method was applied on the basis of wavelet packet energy spectrum feature extraction method. The normalized average values of the three-phase currents were calculated in one cycle. Fault signal indicators were extracted from both frequency domain and time domain, and then the fault feature vectors were constructed. Support vector machine(SVM) was adopted to classify the faults. The training samples were obtained by changing the load and then the SVM was trained by using grid search method. The proposed technique was verified by using Matlab offline simulations and dSPACE real-time simulations, respectively. The results show that comparing with the single wavelet packet energy spectrum feature extraction technique, the hybrid feature extraction technique increases the accuracy of fault diagnosis by 38.88% and 49.43%, respectively, and reduces the diagnostic time by 7.38% and 6.61%, respectively. ? 2019, Harbin University of Science and Technology Publication. All right reserved.

备注：新增回溯数据