摘要
探讨了利用最小二乘支持向量机(LS-SVM)进行非线性系统辨识的方法,LS-SVM用等式约束代替传统支持向量机中不等式约束,求解过程从解QP问题变成解一组等式方程.将得到的LS-SVM模型应用到非线性预测控制,提出了基于LS-SVM模型的非线性预测控制算法.通过CSTR过程仿真表明,最小二乘支持向量机学习速度快,在小样本情况下具有良好的非线性建模和泛化能力.基于LS-SVM的预测控制算法具有很好的控制性能.
An identification method of nonlinear systems using least squares support vector machine (LS-SVM) is proposed. The constraints of inequalities in the classical SVM approach are replaced by equality-type constraints in LS-SVM. The LS-SVM solution follows directly from solving a set of linear equations instead of quadratic programming. A kind of nonlinear predictive control scheme based on the LS-SVM model is presented. Simulation results for a CSTR process show that LS-SVM can be trained fast. The LS-SVM has good ability of modeling nonlinear process and good generalization under small data set available. The nonlinear predictive control strategy based on LS-SVM model shows satisfactory performance.
出处
《控制与决策》
EI
CSCD
北大核心
2004年第4期383-387,共5页
Control and Decision
基金
国家863计划项目(AA413130).
关键词
最小二乘支持向量机
非线性建模
预测控制
非线性控制
Computer simulation
Identification (control systems)
Lagrange multipliers
Least squares approximations
Nonlinear control systems
Parameter estimation
