摘 要:针对矢量控制交流调速系统,该文提出并设计了一种基于再励学习的模糊神经网络速度控制器。详细介绍了基于遗传算法的神经网络权重在线训练方法,仿真对比了输入空间的划分即模糊子集数量对模糊神经网络控制器的训练及其控制效果的影响。仿真结果表明该速度控制器能通过在线训练方式获得最优参数以适应被控对象的参数变化,能使系统获得优良的动态和静态性能。 关键词:模糊神经网络;遗传算法;再励学习;交流调速
Abstract: For vector control AC drive system, the thesis presented a fuzzy neural network speed controller based on reinforcement learning. The thesis introduced the on-line training method of neural network weight based on the genetic algorithm, simulated to compare the division of the input space, which is the number of fuzzy sets, so can see its impaction to the training and the control effect of fuzzy neural network controller. The simulation results show that the speed controller through the online training mode can obtain optimal parameters to adapt to changes in the parameters of the controlled object, and enable the system to obtain good dynamic and static performance. Key words: fuzzy neural network; genetic algorithm; reinforcement learning; AC speed adjustment
|