基于神经网络的非线性系统约束预测控制

Neural network-based constrained generalized predictive control for nonlinear systems

利用非线性激励函数的局部线性表示,提出一种可用于非线性过程的基于神经网络模型的约束广义预测控制算法。该算法将非线性搜索转化为只对当前控制增量的约束,避免了非线性优化求解,并不需要很多的计算量。文中给出了仿真结果。

This paper presents a kind of contrained generalized predictive control algorithm for nonlinear systems based on a neural network model by using the local linearization of a nonlinear activation function.The algorithm converts the nonlinear searching into the constraint on present control increment and avoids the complicated nonlinear optimization.The computation burden is not very serious.A simulation result is also presented in the article.