【摘要】 水电机组的轴心轨迹能够反映机组不同的运行状态，为了提高轴心轨迹的识别率，准确判断机组运行状态，本文提出方向梯度直方图(Histogram of Oriented Gradient, HOG)结合由瞬态搜索优化(Transient Search Optimization, TSO)算法优化的支持向量机(Support Vector Machine, SVM)的方法。将轴心轨迹信号经改进小波阈值方法去噪后，生成轴心轨迹图像，之后提取图像HOG特征，经主成分分析(Principal Components Analysis, PCA)降维处理后，利用TSO-SVM对降维后的特征进行分类识别。结果表明所提方法能够很好地识别不同状态的轴心轨迹，具有识别准确率高和识别速度快的特点。更多还原

【Abstract】 The shaft orbit of hydropower unit can reflect the different operating states of the unit. In order to improve the recognition rate of the shaft orbit and accurately judge the operation state of the hydropower unit, the method of histogram of oriented gradient(HOG) combined with support vector machine(SVM) optimized by transient search optimization algorithm(TSO) was proposed. After denoising the shaft orbit signal by the improved wavelet threshold method, the shaft orbit image was generated, and then the HOG feature of the image was extracted. After the principal components analysis(PCA) dimension reduction processing, the TSO-SVM was used to classify and identify the features. The results show that the proposed method in this paper can well identify the shaft orbit of different states with high recognition accuracy and fast recognition speed.更多还原