一种基于数据挖掘的模糊建模方法

A New Method for Fuzzy Modeling Based on Data-mining

为提高非线性系统模糊建模的速度和精确度,提出一种快速有效的基于数据挖掘的非线性系统模糊建模方法。该方法先采用改进的减法聚类结合模糊C-均值聚类进行结构辨识,在解决初始化问题的同时减少计算量,进而提高建模速度;然后利用带动态遗忘因子的递推最小二乘法进行后件参数辨识,减小动态误差,提高建模精度。将提出的方法应用于Box-Jenkins煤气炉建模和透平膨胀机建模两个例子,仿真结果验证该方法的有效性。

With the goal of improving the speed and accuracy of fuzzy modeling for nonlinear system,a fast and effective method based on data-mining is proposed.In the process of fuzzy structure identification,an improved subtractive clustering combined with fuzzy c-means clustering are introduced to speed up the modeling by solving the problem of initialization of FCM and reducing the calculation of clustering simultaneously.Then the consequence parameter identification is obtained by the recursive least-square algorithm with dynamic forgetting factor,which can low the dynamic error and improve the accuracy.The simulation results of Box-Jenkins gas furnace and Expander show the effectiveness of this approach.