谐波源的神经网络系统辨识与样本建模  

Neural Network System Identification and Sample Modeling of Harmonic Sources

作　　者：郭宝英[1] 郑思凡[2] 

机构地区：[1]仰恩大学工程技术学院,福建泉州362000 [2]黎明职业大学机电工程与自动化学院,福建泉州362000

出　　处：《三明学院学报》2017年第4期61-68,共8页Journal of Sanming University

基　　金：福建省中青年教师教育科研项目(JAT160588);泉州市科技局项目(2014Z138)

摘　　要：针对当前电网谐波的诊断与分类识别均依赖于大量的样本采集作特征提取,采用系统辨识的方法以三相直流励磁同步发电机为对象,以三相直流励磁电压和定子感应电动势为激励,对发电机的正常发电过程及暂态震荡与短路故障进行系统辨识。文中根据问题的实际情况确定了神经网络的类型结构,训练算法等参数,并给出了仿真实现。仿真结果表明,所训练的BP神经网络具有一定泛化能力并可用于激励参数异常的故障样本建模。Considering that the current power quality diagnosis and classification are dependent on a large number of sample acquisition for feature extraction, the normal generation process, transient concussion and short-circuit fault are made system identification by adopting system identification and taking three-phase direct current excitation generator system as research object. Then the type structure and training algorithm etc of the neural network are determined according to the actual situation of the problem and the simulation implementation is given. The simulation results show that the trained BP neural network has some generalization ability and can be used to model the fault samples with abnormal input parameters.

关 键 词：系统辨识 LEVENBERG-MARQUARDT 同步励磁发电机 VC维 BP神经网络 

分 类 号：TM711[电气工程—电力系统及自动化]