An improved recurrent neural network and its application in nonlinear control

This paper proposed a new recurrent neural network (RNN) model. Combining RNN and state feedback predictive control, this paper proposed a predictive control algorithm based on RNN. The algorithm uses the output of RNN as predictive model to calculate the predictive value of state and output which were then calibrated by measurable process variables, at last, using the calibrated values to work out the control law of the whole system through optimal control theory. This algorithm was applied in wastewater treatment control system and some simulations according to different disturbances and set values were presented to illustrate its utility.