论文部分内容阅读
提出了一种非线性系统的模型辨识方法。在只有被辨识系统的输入输出数据的情况下,利用一种无监督的聚类算法来进行结构辨识,从而自动获得模糊规则库,并可以得到模糊系统的初始参数。在聚类的基础上,构造一个与之相匹配的模糊神经网络,用它的学习算法来训练网络得到一个精确的模糊模型,从而实现参数辨识。同时,证明了所构造的模糊神经网络具有通用逼近能力,这个能力在模糊建模和模糊控制方面非常有用。通过对两个非线性系统辨识的仿真结果验证了该方法的有效性。
A model identification method for nonlinear system is proposed. In the case that only the input and output data of the identified system are used, an unsupervised clustering algorithm is used to perform structure identification to obtain the fuzzy rule base automatically, and the initial parameters of the fuzzy system can be obtained. On the basis of clustering, a matching fuzzy neural network is constructed, and its learning algorithm is used to train the network to get an accurate fuzzy model so as to realize parameter identification. At the same time, it is proved that the constructed fuzzy neural network has universal approximation ability, which is very useful in fuzzy modeling and fuzzy control. The effectiveness of this method is verified by the simulation results of two nonlinear system identification.