基于多模型的非线性系统广义预测控制

对于复杂的离散时间非线性系统, 提出一种基于多模型的广义预测控制方法. 通过在平衡点附近建立线性模型, 并用径向基函数神经网络来补偿匹配误差, 形成了非线性系统的多模型表示, 然后采用模糊识别方法作为切换法则, 并结合广义预测控制构成了多模型广义预测控制器. 通过对连续发酵过程的计算机仿真,表明了该方法的有效性.
关键词: 非线性系统；多模型；广义预测控制；径向基函数神经网络
Abstract: A multiple model based generalized predictive control is provided for complex nonlinear discrete time system. The RBFNN, i.e. radial basis function neural network, is used to approximate the matching error of the local linear model, and the nonlinear system is modeled by the multiple linear model and neural network at different equilibrium operating point. A fuzzy recognized method and generalized predictive control algorithm are used to set
up the multi-model generalized predictive controller. From the simulation of continuous fermentation process, it can be seen that the controller proposed in this paper can give a better control performance for nonlinear system.
Key words: nonlinear system; multi-model; generalized predictive control; RBFNN