Data-based construction of feedback-corrected nonlinear prediction model using feedback neural networks

We propose to fit a recurrent feedback neural network structure to input-ou tput data through prediction error minimization. The recurrent feedback neu ral network structure takes the form of a nonlinear state estimator, which can compactly represent a multivariable dynamic system with stochastic inpu ts. The inclusion of the feedback error term as an input to the model allow s the user to update the model based on feedback measurements in real-time uses. The model can be useful in a variety of applications including softwa re sensing, process monitoring, and predictive control. A dynamic learning algorithm for training the recurrent neural network has been developed. Thr ough some examples, we evaluate the efficacy of the proposed method and the prediction improvement achieved by the inclusion of the feedback error ter m. (C) 2001 Elsevier Science Ltd. All rights reserved.