模糊神经网络控制器的优化设计

模糊神经网络控制器不依赖于被控对象精确的数学模型,又能根据被控对象参数的变化自适应调节控制规则和隶属函数参数,但是模糊神经网络控制器在线修正权值计算量大、过度修正权值还可能导致系统剧烈振荡.针对以上问题,提出了在线修正计算中仅对控制性能影响大的权值进行修正,以减小计算量;根据偏差及偏差变化率大小,基于TS模型自适应调节权值修正步长,抑制控制器输出的剧烈变化,避免系统发生振荡.仿真结果表明模糊神经网络控制器的优化设计方法可以改善系统控制性能.

The Optimal Designing of Fuzzy Neural Network Controller

Fuzzy neural controller is a kind of intelligent controller which does not require accurate model of plant and is able to learn to control adaptively. Nevertheless, the long training time usually discourages its application in industry. Moreover when it is trained on\|line to adapt to plant variations, the over\|tuning may cause system oscillates extensively. In this paper, the optimization of fuzzy neural controller is proposed. Only these linking weights that affect the control performance significantly are updated. In accordance with the error and change of the error of the system, the updating step is adjusted adaptively based on T\|S model. Simulation results show that the training time is reduced greatly and the performance of the system is improved.