基于RBF神经网络的直接广义预测控制

针对广义预测控制算法需要在线递推求解 Diophantine 方程及矩阵求逆等计算量大的缺陷,对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算法.该算法将多变量非线性系统转化为多变量时变线性系统,用三次样条基函数逼近系统广义误差向量中的时变系数,然后利用径向基神经网络来逼近控制增量表达式,并基于广义误差估计值对控制器参数向量即网络权值向量θu和广义误差估计值中的未知向量θe进行自适应调整.仿真结果验证了此算法的有效性.

Direct generalized predictive control based on RBF neural network

Direct generalized predictive control(GPC) based on radial basis function(RBF) neural network method for a class of multiple-input multiple-output(MIMO) nonlinear system with unknown parameters is presented to overcome the high load of computing of traditional GPC as on-line recursion of Diophantine,matrix inversion etc.In this method,the MIMO nonlinear system is turned into a MIMO time-varying linear system,then a group of cubic spline functions are used to approach the MIMO time-varying coefficients of generalized error,and a RBF neural network is used to approximate the function of control increment,and both the controller parameters vectors and the unknown vectors 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 effecti-veness of this method.