【摘要】 动车组变压器是保障高速铁路稳定运行的核心设备，频率响应法是目前检测变压器绕组状态的有效方法。为提升车载变压器绕组状态诊断的准确性，结合暂态信号与频率响应法提出基于频率响应与图像特征提取的动车组变压器绕组状态诊断方法。搭建试验车载变压器绕组故障模拟平台，获取不同故障类型和故障位置的频响曲线，利用类Gram矩阵结合幅频和相频曲线信息，再利用密度分层法转换为伪彩色图，提取对应的灰度共生矩阵和灰度差分矩阵特征值，根据鹈鹕优化支持向量机方法对绕组故障进行诊断。试验结果表明：车载变压器绕组故障发生时，伪彩色图能够反映出故障信息，有利于图像分析和特征提取，采用基于频率响应与图像特征提取的动车组变压器绕组状态诊断方法能够识别车载变压器绕组的典型故障类型和位置。更多还原

【Abstract】 The trainset transformer is the core equipment to ensure the stable operation of high-speed railroads, while the frequency response method is an effective method to detect the state of the transformer windings. In order to improve the accuracy of the diagnosis of on-board transformer winding state, combining the transient signals with the frequency response method, this paper proposed a trainset transformer winding state diagnosis method based on the frequency response and image feature extraction. Firstly, a simulation platform was built for testing on-board transformer winding faults to obtain the frequency response curves of different fault types and fault locations, and to combine the amplitude-frequency and phase-frequency curve information by using the Gram-like matrix set. Secondly, the density hierarchy method was converted to a pseudo-color image to extract the corresponding grayscale covariance matrix and grayscale difference matrix eigenvalues, and to finally diagnose the winding faults according to the pelican-optimized support vector machine method. The experimental validation results show that in the case of on-board transformer winding fault, the pseudo-color image can reflect the fault information, which is conducive to image analysis and feature extraction. The diagnosis method for the state of trainset transformer winding based on frequency response and image feature extraction can identify the typical types and locations of the on-board transformer winding faults.更多还原