An Algorithm of Nonlinear Model Predictive Control Based on BP Network

An algorithm of nonlinear predictive control method based on BP neural networks is proposed. According to the nonlinear characteristics of the controlled plant, the method builds its predictive model using BP networks. Furthermore, utilizing prior knowledge reduces the scale and raises the learning speed of the network. Then GA as a global optimization method is used to optimize the control trajectory, which makes up for the neural network’s disadvantage of easily getting trapped in local minimization. The results of mulitivariable nonlinear systems show that the method can accelerate convergence and is easy to implement.