参数未知系统直接广义预测控制

Direct Generalized Predictive Control for Systems With Unknown Parameters

传统的广义预测控制(GPC)算法需要在线递推求解Diophantine方程及矩阵求逆,因此计算量很大,不适合对实时性要求高的控制系统。文中对参数未知线性系统提出一种直接广义预测控制(DGPC)方法。该方法直接辨识广义预测控制器参数θu,即基于广义误差估计值对控制器参数和广义误差估计值中的未知向量θe进行自适应调整。理论证明了该方法可使广义误差收敛到原点的一个小邻域内。仿真结果验证了此方法的有效性。由于实时性提高,并且不需要被控对象模型,该方法更利于实际应用。

Traditional GPC is not suited for the application because of the high load of computation resulting from online recursion of Diophantine,matrix inversion,etc.Therefore a direct generalized predictive control（DGPC） method for a class of linear system with unknown parameters is presented in the paper.In DGPC the controller parameters θu and the unknown vectors θe in the estimation of generalized error are adjusted adaptively.It is proved that the proposed method can make the estimation of generalized error converge to a small neighborhood of the origin.Simulation results demonstrate the effectiveness of this method.Moreover,the method not only is of good real-time but also doesn＇t need mathematic model of plant,so it even more favors the practical application.