Generalized Predictive Control of Neural Network Based on LM Optimization
Starting from the establishment of generalized predictive model based on neural network, LM optimization algorithm is applied to the perdictive control model for study in order to solve these problems that the training speed of the BP network is slow and it is easy to trap into the local minimum.Generalized predictive control and neural network which has the capability of approaching any nonlinear function are combined to forecast the future outputs of the system.LM algorithm is used instead of gradient descent method to optimize controller parameters and it makes full use of Jacobian matrix information identified by neural network.The result of Matlab simulation indicates that the neural network using LM algorithm has the feature of fast convergence rate,model of high precision and good robustness,which is more suitable for real-time nonlinear control.
S. M. Wang et al., "Generalized Predictive Control of Neural Network Based on LM Optimization", Applied Mechanics and Materials, Vols. 66-68, pp. 2164-2169, 2011