【摘要】 牵引变压器的安全运行关乎电力系统的稳定，为识别变压器绕组故障，本文提出基于随机暂态信号捕获分析的牵引变压器绕组状态在线感知与智能评估方法。首先获取牵引变压器绕组的端口暂态电信号，得到变压器绕组的频响曲线，然后结合频响曲线的幅频特性和相频特性转化为灰度图，根据灰度图提取图像的灰度梯度共生矩阵的纹理特征，根据提取的8个纹理特征计算特征关联度(FCD)来识别绕组故障，最后在一台试验牵引变压器上进行局部翘曲的故障模拟。结果表明，由端口暂态电信号的频响曲线转化得到的灰度图在不同故障程度下存在差异，绕组不同故障位置的特征关联度(Features Correlation Degree, FCD)数值有明显不同，且不同故障程度有其相应规律，根据灰度图提取的8个图像特征与相应的特征关联度，可区分牵引变压器局部翘曲故障情况下故障位置与故障程度。更多还原

【Abstract】 The safe operation of traction transformer is critical for the stability of power system.To identify transformer winding faults, this paper proposes an online sensing and intelligent assessment method for the traction transformer winding state based on the stochastic transient signal capture analysis.Firstly, we obtain the port transient electrical signals of the traction transformer windings and get the frequency response curves of the transformer windings.Then, combining with the amplitude-frequency and phase-frequency characteristics of the frequency response curves, the grayscale maps are generated.Texture features are extracted from these grayscale maps using gray gradient covariance matrices, and the feature correlation degree(FCD) is calculated according to the eight extracted texture features to identify winding faults.Finally, a local buckling fault simulation is performed on a test traction transformer.The results show that the grayscale maps obtained from the frequency response curves of the port transient electrical signals exhibit differences under different fault degrees.The FCD values for different faulty windings are significantly different, and there are corresponding laws for different fault degrees.The eight image features and the their respective FCD values, extracted from the grayscale maps, can distinguish the fault location and fault degree in the case of partial warping faults of traction transformer.更多还原