基于RBF神经网络的非线性控制系统

针对时滞、非线性多变量耦合系统控制中串级控制存在滞后，并且PI控制参数初始设置困难，很多时候只能手动控制的缺陷，本文采用内模控制方隶．通过RBF神经网络训练获得内部模型，同时利用最小二乘模型降解，简化解耦矩阵的求取，实现了两输入两输出系统的解耦控制。仿真结果表明．该方案可以消除变量间的耦合，并且解决了时滞问题、降低过程超调量，使得控制系统更加平稳，改善了过程控制的品质。同时．当对象特性发生一定改变时．系统具备良好的鲁棒性能。

Non-linear Control Systems Based on RBF Neural Network

For Multi-varies, time-delay and coupling systems, traditional PID method has its disadvantages, such as not in time, difficult to set initial PI parameters. This paper applied IMC method with RBFNN to realize decoupling control. RBFNN is used to retain the IMC model; while LMS used in reducing model order, simplify retain of decoupling matrix. Simulations demonstrate that this method has effects on decoupling, time-delay settling. The performances of control system have certain improvements. Meanwhile, when the plant character has a little change, system has good performance and robustness.