摘要
为了准确识别感应电机定子匝间短路故障,该文提出一种基于骨干微粒群算法优化支持向量机的故障诊断方法,并给出了可行的诊断步骤。该方法首先利用小波包频带能量分解技术,将定子电流信号的各频率分量分解到不同频带,形成感应电机运行状态的特征向量,并以此作为支持向量机的输入向量。采用支持向量机进行分类,并利用无需设置控制参数的骨干微粒群算法和交叉检验优化模型参数,避免了参数选择的盲目性。最后试验结果表明,该方法诊断感应电机定子匝间短路故障能取得良好的效果。
AIn order to accurately recognize the stator winding inter-turn short circuit fault of induction motors, a novel method for fault diagnosis was proposed based on support vector machine (SVM) optimized by bare-bones particle swarm optimization (BBPSO). And the feasible diagnostic steps were also introduced. Firstly, the feature vector of in- duction motor in different conditions was extracted by using wavelet packet. Different frequency components were de- composed to series of frequency bands, which was considered as the input vector of SVM. SVM was used to solve the classification problem, and the parameter-free BBPSO and cross-validation were taken to optimize model parameters, which avoided the blindness of preferences. Finally, the experiment shows that the proposed method is effective to diag- nose the stator fault of induction motors.
出处
《电力系统及其自动化学报》
CSCD
北大核心
2013年第2期65-70,共6页
Proceedings of the CSU-EPSA
基金
教育部科学技术研究重大项目(311021)
关键词
感应电机
定子匝间短路
骨干微粒群优化算法
小波包
支持向量机
故障诊断
induction motor
stator winding inter-turn short circuit
bare-bones particle swarm optimization( BBPSO )
wavelet packet
support vector machine (SVM)
fault diagnosis