Study Of Nonlinear Control Based On ANN Inverse System Method And Soft Sensing

Usually, the nonlinear system is linearized in order to be controlled by the method of linear control theory. As a feedback linearization method, the inverse system method is widely applied in various fields for its advantage of clear physical idea, simple structure and facile realization. But in the practical engineering applications, most of the systems are nonlinear with large scale, complicated structures and frequently varying parameters which hinder the application of the inverse system method. Besides, inverse system method often requires the feedback of immeasurable variables, which also hinder its application. In order to solve the practical problems existing in the engineering applications, on the basis of the research on inverse system method and ANN, nonlinear control and dynamic feedback soft sensing are studied in this paper.First, this paper presents the inverse system theory and method analysis, and the inverse system method is applied to design the control law of the reagent in CSTR. Compared with the control law designed by the staged-transformation, the results show that the control law designed by the inverse system method has a better closed loop dynamic performance which proves the inverse system method is effective and feasible to control the nonlinear system.Second, for the nonlinear control, directly immeasurable variables are always fed back. On the basis of studying ANN inverse system method, this paper supposes an inner estimator exists in the original system and gives the conditions of its existence and reversibility so as to solve the soft sensing of the directly immeasurable variables, which makes the inverse system method become easier to be applied.Finally, combined with the soft sensing thought, the ANN inverse system method is applied to the generator's excitation voltage control law design, and immeasurable dynamic feedbacks are estimated using measurable variables and its derivatives. This paper presents the concrete design scheme, control realization and the research and analysis of the dynamic soft sensing. Simulation results show the validity of the ANN inverse method in the nonlinear control.