Research On Model Free Controller Design Method Based On Neural Network

As the complexity, nonlinear and uncertainties of modern industrial processes, structure or parameters of the controlled object will change in the actual process control, which may make it difficult to achieve the accurate mathematical model. Therefore, model free adaptive control method which does not base on the mathematical model of the controlled object is an important aspect of automatic control.Model free adaptive control method is an adaptive control method which does not need mathematical model of controlled process. It does not contain any mathematical model information of controlled process, only I/O data of the controlled system should be available when the controller is designed. The method has strong parameter and structure adaptability.In this thesis, model free adaptive control method based on neural network identifier is put forward, using neural network to identify the controlled object and using the partial derivative of neural network output to control input to replace pseudo-partial-derivative, which is based on model free adaptive control method. Better control effect of the proposed method is proved by the simulation research.The main contents of the thesis are taken as follows:Firstly, the theory of neural network is introduced. For neural network direct inverse controller design method, several specific examples are given for simulation analysis. According to the simulation results, the method has good control performance for reversible system, and can not achieve a good control effect for irreversible system.Secondly, for irreversible system, two improved methods of neural network designing controller are given. According to the simulation analysis for irreversible system, the improved controllers achieve a better control effect and have good control performance for irreversible system.Finally, model free adaptive control method is analyzed, and the control law derivation is given. Using BP neural network to identify the controlled object, model free adaptive control method based on BP algorithm is put forward. According to the simulation analysis of specific examples, control performance of model free adaptive control method based on BP algorithm is better than that of model free adaptive control method.