支持向量机回归建模方法简单,有完备的理论基础,所反映的直接是系统的非线性特征,因而与神经网络等非线性回归方法相比具有独特的优点。因此,本文提出的非线性预测控制结构采用支持向量机回归建立非线性系统的预测模型,采用LM(Levenberg-Marquardt)非线性最小二乘优化方法作为非线性滚动优化方法。实验证明了这种预测控制具有较好的非线性控制效果,而且能够较好地抵消不可测干扰。 Support vector machine regression modeling method is simple and has a complete theoretical foundation, which reflects directly the nonlinear characteristics of the system. Therefore, it has unique advantages over nonlinear regression methods such as neural networks. Therefore, the proposed nonlinear predictive control architecture uses support vector machine regression to establish a predictive model of nonlinear systems, and adopts the Levenberg-Marquardt nonlinear least squares optimization method as a nonlinear rolling optimization method. Experiments show that this kind of predictive control has good non-linear control effect and can better counteract unmeasured interference.