基于覆盖聚类的TS模糊神经网络热工对象建模

Fuzzy Neural Network Modeling Based on TS Model and Cover Cluster

热工过程具有非常复杂的动态特性以及强耦合、大延迟和不确定等特征。控制过程需要较为精确的模型,但是常规的建模往往并不能满足要求,因此提出一种改进型的TS模糊神经网络建模方法。首先基于一种覆盖聚类算法对离线数据进行分类,初步得到模糊神经网络的前件和后件参数,再利用卡尔曼滤波算法调整后件参数和动态梯度算法调整隶属函数的宽度和中心,最后把得到的前件参数和后件参数进入在线网络,若进入网络的实时数据不属于所有的类,则应增加聚类中心和规则。

Considering the complex dynamic behaviors,nonlinearities,strong coupling and time-varying characteristics of modern thermal process,the traditional modeling method does not satisfy the demand of accuracy.Based on the fuzzy theory and neural network,a new method is provided.In this paper,firstly,the offline data is sorted by cover cluster in the model identification method and then cluster centers were used in fuzzy net.Secondly,the Kalman filter is used to identify the consequent parameters of fuzzy rules.At last,gradient algorithm is used to adjust antecedent parameters.Finally,two simulation examples of thermal process are given to prove the validity of the network.