Research Of Neural Networks Control For Uncertain Systems

In the practical control, systems always have uncertainty because of existing the followingfactors: parametric error, unmodeled dynamics or external interference, etc. So the ideal controlindex is difficult to be reached. For settling the above question, it is needed to designcorresponding controller that can make the uncertain systems keep stable or satisfy ideal designindex. Neural networks (NN) have the stronger capability of parallel processing, approachingarbitrary nonlinear mapping and self-organizing learning, etc. These characters can solvepreferably the uncertain problems of systems. We shall acquire good control effects if NN wereapplied to uncertain systems. In this paper, linear uncertain systems, nonlinear uncertain systems and delay-time uncertainsystems were researched respectively, designed corresponding controllers based on NN. Intime of studying linear uncertain systems, integrated NN control, robust control with slidingmode (SM) variable structure control. A state feedback controller was designed based onMatrix Inequality;modeling for the uncertain variables by NN;last, NNSM controller wasdesigned. A sufficient condition that can guarantee the linear uncertain systems to be stable inthe sense of Lyapunov was given. The simulation results indicate that the control effect isbetter than that of the single control algorithm. The design of NN controller was studied for a class of nonlinear uncertain systems.According as Lyapunov theory, a sufficient condition that can guarantee the controller to existwas given. It was proved that the systems were stable in the sense of Lyapunov. The simulationresults indicate that the controller is effective. For reducing the tracking error, enhancing thespeed of response, the original controller was improved. The simulation examples indicate themodified controller can satisfy the demand of enhancing control performance.The stability problems for uncertain systems with time-delay were analyzed in this thesis.Firstly, a class of nonlinear time-delay systems with uncertainty was considered. A scheme ofdesigning NN controller based on Radial Basis Function (RBF) was proposed. It can guaranteethe time-delay systems to be stable in sense of Lyapunov. And the simulation examples verifythe effectiveness of the controller. Then, a class of nonlinear uncertain systems with time-delaywas studied. Based on original robust H_∞ control, NN control and SM control were alsoadded into the control, so the robust H_∞ performance was melted into it. The NNSMcontroller based on roust H_∞ was designed. It was proved that the controller could guaranteethe time-delay systems to be stable in sense of Lyapunov. The simulation results indicate thecontrol performance was enhanced obviously compared with original controller.