基于支持向量机的超导限流器故障电流模式识别研究

Research of rapid pattern recognition of fault current based on support vector machine for HTS three-phase saturated core fault current limiter

对于饱和铁芯型超导限流器，指出短路故障电流需要快速识别，设计和制造了三相小样机和相应的基于labview和NI板卡的实时检测控制系统．提取了短路故障电流的2个重要特征：电流实时幅值，电流变化率，根据特征，分别采用神经网络感知机模式分类，线性核的支持向量机和径向基函数核的支持向量机，离线在matlab环境下训练，找出最优分类面．对几种方法进行比较实验，实验数据验证表明了RBF核支持向量机具有最好的识别效果．但是该方法难以在FPGA中实现，而线性核支持向量机是综合识别效果和可实现性2个指标的最佳选择．

The necessity of rapid recognition of fault current is presented. The SCFCL and its real-time data acquisition and control system based on labview and board of NI are designed and fabricated. Two important characteristics such as value of current and variable rate of current are extracted. According to the characteristics, the methods of neural network perceptions, support vector machines （SVM） with linear kernel function or radial basis kernel function are individually adopted to recognize the fault current. The off-line training in Matlab is done to find the optimized classification surface. By comparison among the methods with acquisitive experiment data, the support vector machine with RBF kernel function is proved to be more effective than another two. But it is hard to put into practice in FPGA. While the S VM with linear kernel is the best choice among them in recognition effectiveness and easy-to-use in FPGA.