基于改进模糊C回归聚类的水轮发电机组的模糊辨识
Fuzzy Identification of Hydro-turbine Generating Units Based on Modified Fuzzy C-regression Clustering Algorithm
-
摘要: 针对水轮发电机组精确建模的难题,提出了一种基于改进模糊C回归聚类的T-S模糊模型辨识方法。考虑到样本输出值与聚类超平面输出之间的误差值指标的重要性,对于模糊C回归聚类算法进行了改进。该算法将误差值的倒数赋给对应的样本隶属度,构建新的权重矩阵用于更新聚类超平面,从而加速聚类朝向最优聚类超平面的收敛;提出一个新的超平面型隶属度函数,直接利用超平面辨识前提参数;应用带遗忘因子的递推最小二乘算法在线辨识模型的结论参数。以三个常用的数学实例及某水电站水轮发电机组为对象,进行T-S模糊模型的建立,并与其他辨识方法进行比较。结果表明,所提出的模糊辨识方法具有较高的辨识精度,辨识所得模型具有较强的泛化能力。Abstract: A Takagi-Sugeno(T-S)fuzzy model identification approach based on modified fuzzy c-regression clustering aiming at the difficult problem of high-precision modeling of hydro-turbine generating units is proposed. Considering the importance of the error index between the sample output value and the output of the clustering hyperplane,the fuzzy c-regression clustering algorithm is improved. The algorithm assigns the reciprocal of the error value to the corresponding sample membership degree,and constructs a new weight matrix to update the clustering hyperplane,thereby accelerating the convergence of the clustering process towards the optimal clustering hyperplane. A new hyperplane-shaped membership function is proposed,which directly uses the hyperplane to identify the premise parameters. The recursive least squares algorithm with forgetting factor(FFRLS)is used to identify the consequent parameters of the model online. Taking three commonly used mathematical examples and a hydro-turbine generating unit of a hydropower station as the objects,the corresponding T-S fuzzy models are established and compared with other identification methods. The results indicate that the proposed fuzzy identification method has high identification accuracy,and the identified model has strong generalization ability.