基于最小二乘支持向量机的多变量逆系统控制方法及应用

为提高多变量、非线性和强耦合系统的动态特性和解耦能力，解决逆模型辨识问题，讨论了基于最小二乘支持向量机(least squares support vector machines，LS-SVM)的多变量逆系统解耦控制方法。通过分析LS-SVM的函数拟合特性，离线建立被控对象的非线性逆模型，将得到的逆模型直接串接在原对象之前，原系统被解耦成多个独立的单变量伪线性子系统。为克服直接逆模型的建模误差，提高系统鲁棒稳定性，提出了复合控制方法，其中直接逆模型作为前馈控制器，而用PID控制器作为反馈控制器。文中还分析了球磨机控制系统的特点，并进行了仿真控制研究，仿真结果表明该复合控制方法不依赖于系统的精确数学模型，且解耦能力强、鲁棒稳定性好、跟踪精度高。

Decoupling Compound Control Method Based on Least Squares Support Vector Machines Multivariable Inverse System and Its Application

To improve the dynamical property and decoupling capability for multivariable, nonlinear and strong coupling system,and to solve the problem of inverse model identification, the decoupling control method of multivariable inverse system based on least squares support vector machines(LS-SVM) was discussed. The fitting characteristic of LS-SVM function was analyzed, and the nonlinear inverse model of the controlled object was offline built with LS-SVM. The LS-SVM inverse model was cascaded before the original object, thus, the original system was decoupled into several independent single variable pseudo-linear subsystems. To overcome the modeling error of direct inverse control and improve the robust stability, the compound control system was also proposed, in which the inverse model was used as feed-forward controller, and PID controller was used to realize feedback control. The characteristics of ball mill control system were analyzed, and the ball mill control system was simulated and studied, the simulative results show that the combined method does not depend on the accurate mathematical model and has strong decoupled ability, good robust stability, high tracking accuracy.