最小支持向量机在系统逆动力学辨识与控制中的应用

为克服支持向量机(support vector machine，SVM)在线辨识过程需要较大的内存开销的问题，该文将递推最小二乘法(recursive least square，RLS)与最小二乘支持向量机(least squares support vector machine，LS-SVM)回归相结合，利用RLS在线调整支持向量机的权向量和偏移量，实现了系统逆动力学模型的在线辨识。在获得逆动力学模型的基础上，设计了一种基于逆动力学递推最小二乘支持向量机的控制算法，利用RLS在线调整控制器参数。过热汽温辨识和控制的仿真结果表明，辨识出的逆动力学模型具有较高的精度，所设计的控制器能获得较好的控制性能和有较强的适应能力。

Application of RLS-SVM in Identification and Control for Inverse Dynamics of System

To overcome the large memory expense in the process of on-line identification by utilizing support vector machine(SVM), least squares support vector machine (LS-SVM) was combined with recursive least square(RLS), the weigh vector and bias were adjusted on-line by RLS algorithm, and on-line identification of inverse dynamic model of system was realized. Based on the inverse dynamic model acquired, a control algorithm based on recursive least squares support vector machine (RLS-SVM) of inverse dynamics was designed. The parameters of controller were adjusted on-line by RLS algorithm. The simulations on superheated steam temperature identification and control system show that the inverse dynamic model identified has high precision and the controller designed has good control performance and strong adaptability.