| 引用本文: | 杨苹,周少雄,胡斌,马艺玮.双馈风力发电机系统的自抗扰神经网络的励磁控制[J].控制理论与应用,2012,29(2):251~256.[点击复制] |
| YANG Ping,ZHOU Shao-xiong,HU Bin,MA Yi-wei.Active disturbance-rejection neural networks excitation-control of double-fed induction generator[J].Control Theory and Technology,2012,29(2):251~256.[点击复制] |
|
| 双馈风力发电机系统的自抗扰神经网络的励磁控制 |
| Active disturbance-rejection neural networks excitation-control of double-fed induction generator |
| 摘要点击 2090 全文点击 1978 投稿时间：2011-06-06 修订日期：2011-08-17 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/j.issn.1000-8152.2012.2.CCTA-SG110682 |
| 2012,29(2):251-256 |
| 中文关键词 双馈风力发电机 励磁控制 自抗扰控制 BP神经网络 |
| 英文关键词 double-fed wind power generation excitation control active disturbance-rejection controller back propagation neural networks |
| 基金项目 广东省科技计划资助项目(2010A010200004); 广东省教育部产学研结合资助项目(2009B090300424); 2010年粤港关键领域重点突破招标资助项目(20100107-3). |
|
| 中文摘要 |
| 在双馈发电机传统控制方式的基础上, 将自抗扰控制技术和BP神经网络相结合结合, 应用于双馈风力发电机并网运行的控制上, 提出了一种新的双馈风力发电机并网运行控制方案. 该控制方案具有内外两个控制环, 内环通过BP神经网络实现双馈风力发电机的转子d-q轴电流控制, 外环通过自抗扰技术实现双馈风力发电机定子侧的有功、无功控制. 由于自抗扰控制器利用一阶跟踪微分器和扩张状态观测器对系统扰动进行动态跟踪补偿, 在此基础上输出双馈电机转子交--直轴电流的参考值, 然后将该参考值作为BP神经网络训练样本的输入, 训练后的BP神经网络可以更好地逼近实际转子电压输出量. 论文设计并实现了该方案的具体控制算法. 仿真测试表明: 该控制方案具有优良的动态性能, 对系统的内外扰动具有较强的鲁棒性, 在没有精确的发电机参数情况下依然可实现并网系统的稳定运行. |
| 英文摘要 |
| On the basis of traditional control for the double-fed induction generator (DFIG), the active disturbancerejection method and the back-propagation neural-network control method are combined together and applied to the control of DFIG operating in the power grid. The control strategy consists of the internal and external control loops. The internal loop controls the rotor current of DFIG using the back propagation (BP) neural-network controller, while the external loop controls the active power and the reactive power by the active disturbance-rejection controller (ADRC). The ADRC tracks system dynamic disturbances by a first-order differential tracker (TD) and an extended state-observer (ESO), and outputs the reference quadrature current and the direct current of DFIG rotor. Reference values are fed to the BP neural network input for training; the trained BP neural networks precisely approximate the actual rotor output voltage. The algorithm of this strategy is developed and simulated; simulation results show that this strategy is with excellent dynamic performances, and the system is robust to the internal and external disturbances, providing a stable operation for DFIG in the power grid without knowing the exact generator parameters. |
|
|
|
|
|