基于TSA多变量非线性RBF神经网络预测控制

Multivariable nonlinear RBF neural network predictive control based on TSA

为了解决现在有的非线性预测控制方法在线实时求解非线性方法的困难,提出一种基于TSA的多变量非线性RBF神经网络预测控制算法。该算法采用多个RBF神经网络建立非线性系统的过程模型,并作为预测模型。采用树和种子算法(TSA)在线搜索非线性预测控制系统的最优控制律,避免了直接递推控制律时解决复杂的非线性优化问题。CSTR过程的仿真对比结果验证了该算法的跟踪性能和抗干扰能力。

To deal the difficulty for online real-time computing the nonlinear equation with the present nonlinear predictive control method,a multivariable nonlinear neural network predictive control algorithm based on tree and seed algorithm(TSA)is proposed.This algorithm adopts the multi-RBF neural network to construct the nonlinear system,which is used as the predictive model.The optimal control law of the nonlinear predictive control system is searched online using TSA,thereby avoiding the problem of complex nonlinear optimization for directly solving the control law.Simulation results of CSTR show that the proposed control scheme has excellent tracking and resisting disturbance abilities.