基于二次仿射传播聚类的非线性系统多模型LSSVM建模

Multi-model LSSVM modeling for nonlinear systems based on twice affinity propagation clustering

针对复杂非线性系统单模型建模存在计算量大和精度差的问题,提出一种采用仿射传播聚类的多模型LSSVM建模方法,通过仿射传播聚类对样本数据按输入集和输出集二次聚类划分,并分别建立LSSVM子模型,非线性系统模型通过子模型加权合成.将该方法应用于两电机变频调速系统的张力和速度模型辨识,仿真结果表明,该建模方法具有较高的精度,能准确拟合系统的非线性特性.

For complex nonlinear systems, a single model suffers from heavy burden calculation and bad accuracy. A modeling method based on multi-model least squares support vector machines（LSSVM） using affinity propagation clustering algorithm is proposed. The clustering algorithm is successively applied to group the data into clusters according to input set and output set, and LSSVM is used to construct sub-models. The model of nonlinear system is constructed by weighting all LSSVM sub-models. The proposed method is used to identify tension and speed model for two-motor variable frequency speed-regulating system. Simulation results show that the proposed method can be used to fit nonlinear characteristics of the system with high precision.