Diagonal Recurrent Neural Network as an On-Line Identifier of a Nonlinear Energy System

W.A. Caswell, A. Davari, and B. Liu (USA)


System modeling and identification, Energy systems, Circulating fluidized beds, Neural networks


System identification for nonlinear systems is difficult. Complex systems do not always have good theoretical models. The circulating fluidized bed is an example of complex nonlinear energy systems. One method for modeling system identification is using neural networks. Feedback networks have good accuracy offline, but are generally too complex to implement for online system identification. Feedforward networks are quick enough to work as online model, but generally lack the necessary accuracy. In the paper we develop an online system identification model based on the diagonal recurrent neural network method using MatlabĀ® . The model was tested with a cold flow circulating fluidized bed in both and offline and online capacity. The results show that the DRNI works well for modeling the nonlinear system of the circulating fluidized bed

Important Links:

Go Back