Study of How to Identify Nonlinear System Based on Neural Network by MATLAB Toolbox

Aimed at the bulky nonlinear temperature system with time-delay, a DFOPDT model is established. The relation between the classical model and the disperse model of the system is deduced. An algorithm is presented which is an effective way to solve unknown parameters of the model by neural network. The network can soon converge through the iterative calculation in which the value of unknown parameters is predicted continuously. It also solves training a network with unknown output by the function ´train´ in MATLAB neural network toolbox. The system model is identified by the method through step response data, which proves that the method has a good convergence and high accuracy.