在线优化可变基的非线性预测函数控制

Nonlinear Predictive Function Control of Variable Basis Function with Online Optimization

针对基本预测函数控制只能用于线性系统控制的这一不足,给出基于实用随机NARMAX模型的非线性预测函数控制,采用可克服算法病态的非线性递推最小二乘法进行参数估计,利用在工作点处的动态切平面逼近方法,将实用随机NARMAX模型用线性时变CARMAX模型逼近,使非线性预测函数控制转化为线性模型下的预测函数控制,用线性优化算法求解控制输入,避免了复杂的非线性优化问题,并采用直接极小化指标函数优化算法对可变基函数的加权系数进行在线优化,提出在线优化参数的非线性预测函数控制。仿真研究表明,因算法具有优化可变基函数加权系数和预测函数控制功能,系统具有优良的控制响应。

A nonlinear predictive function control based on practical random NARMAX model was developed so that the basic predictive function control can be applied but not limited to linear system control.The model parameter were estimated with the nonlinear recursive least squares method of overcoming algorithmic ill condition,The original practical random NARMAX model was approximated into a linear time-varying CARMAX model with the method of dynamic cutting horizontal approximating at working point to transform the nonlinear predictive function to the linear predictive function control.The linear optimization algorithm improved complicated nonlinear optimization in control input,The weighting coefficient of variable basis function was optimized online with the direct minimization of index function optimization algorithm,and a nonlinear predictive function control with online optimization parameter was proposed.Simulation results showed that the control response of the system was excellent due to the algorithm’s optimized variable basis weighting coefficient and predictive function control.