Nonlinear continuous stirred tank reactor (CSTR) identification and control using recurrent neural network trained Shuffled Frog Leaping Algorithm

This paper presents the design of a recurrent neural network trained Shuffled Frog Leaping Algorithm (RNN-SFLA) for identification and tracking control of a nonlinear continuous stirred tank reactor (CSTR). The RNN is applied to approximate unknown dynamic of system and SFLA is used to train and optimize the connection weights of RNN. In the proposed control scheme, neural control system synthesis is performed in the closed-loop control system to provide appropriate control input. For this, the error between desired system output and output of control object is directly utilized to tune the network parameters. The capability and efficiency of the proposed method is illustrated by the temperature control of a nonlinear CSTR. The simulation results show that RNN-SFLA controller has excellent dynamic response and adapt well to changes in reference trajectory and system parameters.