A Simplified Recurrent Neural Network Trained by Gbest-Guided Gravitational Search Algorithm to Control Nonlinear Systems

Abstract

This paper presents a feedback control strategy using a SimplifiedRecurrent Neural Network (SRNN) for nonlinear dynamical systems. As anenhancement for a previously reported modified recurrent network (MRN),the proposed SRNN structure is used as an intelligent Proportional-IntegralDerivative(PID)-like controller. More precisely, the enhancement in theSRNN structure was realized by employing unity weight values between thecontext and the hidden layers in the original MRN structure. The newlydeveloped Gbest-guided Gravitational Search Algorithm (GGSA) wasadopted for optimizing the parameters of the SRNN structure. To show theefficiency of the proposed PID-like SRNN controller, three differentnonlinear systems were considered as case studies, including a control valve,and a complex difference eq.. From an extensive set of evaluation tests, whichincludes a control performance test, a disturbance rejection test, and ageneralization test, the proposed PID-like SRNN controller demonstrated itseffectiveness with regards to precise control and good robustness andgeneralization abilities. Furthermore, compared to other Neural Network(NN) structures, including the original MRN and the Multilayer Perceptron(MLP) NN, the SRNN structure attained superior results due to the utilizationof a reduced set of parameters. From another study, the GGSA accomplishedthe best optimization results in terms of control precision and convergencespeed compared to the original Gravitational Search Algorithm (GSA).