Neural model predictive controller for multivariable process

In this paper, control of a non-minimal quadruple tank process, which is non linear and multivariable is reported. A nonlinear Model Predictive Controller (NMPC) is developed using a Recurrent Neural Network (RNN) as a predictor. The process data is obtained from the laboratory scale experimental setup, which is used in training the RNN. The network trained is used in controlling the quadruple tank, solving the least square optimization problem with a quadratic performance objective. The control system is implemented in real-time on a laboratory scale plant using dSPACE interface card and Matlab software. The quality of controller using NMPC is compared with dynamic matrix control (DMC) for reference tracking and external disturbance rejection.