基于规则选取和补偿运算的模糊系统建立方法

为了解决模糊系统建立过程中规则数目的选择问题,提出了一种模糊系统建立方法。该方法利用K均值法计算聚类中心,不需要预先给出聚类的数目,聚类的数目根据聚类准则函数的收敛性决定,同时在算法结构中引入补偿因子以增强算法的稳定性,再结合梯度下降法辨识后件参数,从而得到模糊系统。与传统的模糊系统建立方法进行比较,该方法提高了辨识精度且能自动生成模糊规则,避免了规则数目选取的盲目性。最后将该方法用于辨识单元机组的协调控制系统,仿真结果表明了该算法的有效性和快速性。

New method to establish fuzzy systems based on rule selecting and compensatory calculation

To solve the problem that how to select the rule number in the establishment of fuzzy system,a new establishment method is suggested.The cluster center is calculated using K-means algorithm.The number of clusters is determined by the convergence of cluster principle function.A compensation factor is introduced to improve the stability.The gradient descent algorithm is used to identify parameter of consequent part of rule base.This method is applied to identify the coordinated control system of power unit.Simulation results prove the validity of this method.