摘要
为了避免Diophantine方程求解和矩阵求逆运算,提高广义预测控制算法的实时性,对一类参数未知多变量非线性系统提出一种径向基函数神经网络的直接广义预测控制(GPC)算法。该算法将多变量非线性系统转换为多变量时变线性系统,然后利用径向基神经网络来逼近控制增量,对控制器参数向量,即网络权值中的未知向量基于跟踪误差进行自适应调整。理论证明,该方法可使跟踪误差收敛到原点的一个小邻域内。仿真结果验证了此算法的有效性。
For avoiding resolving Diophantine function and inverse matrix in order to improve the real-time ability of the algorithm,presented radial basis function（RBF） neural network direct GPC for a kind of multi-input multi-output（MIMO）nonlinear system with unknown parameters.It turned the MIMO nonlinear system into a MIMO time-varying linear system,then used RBF neural network to approximate the function of control increment,and adjusted the controller parameters vectors adaptively based on tracking error.It proved that the proposed method could make the tracking error converge to a small neighborhood of the origin.Simulation results demonstrate the effectiveness of this method.
出处
《计算机应用研究》
CSCD
北大核心
2010年第7期2513-2516,共4页
Application Research of Computers
