Study And Application Of Nonlinear Model Predictive Control Fast Algorithm

Linear model predictive control (LMPC) algorithms have been widely used in the multivariable process industry system with constraints. But linear models are often inadequate to describe highly nonlinear processes and moderately nonlinear processes which have large operating regimes. As a result, the research on nonlinear model predictive control (NMPC) is of great significance theoretically and practically.LMPC solves an online quadratic programming and an analytic globally optimal solution can be obtained per sampling time. But a nonconvex nonlinear programming problem needs to be solved online in NMPC. It is difficult to be an analytic solution and the computation burden increases exponentially with the dimension of the decision variable. The computation burden makes it difficult to implement NMPC in real-time.The major objectives of the dissertation are to study nonlinear modeling methods and computational methods of solving general nonlinear programming in NMPC. In order to reduce the online computational load, several NMPC fast algorithms are developed. The main contents are depicted as follows:1) A survey on the theoretical and practical development and status of nonlinear model predictive control is discussed. The characteristics and main principle of nonlinear model predictive control are analyzed, the causes of on-line heavy computational burden are pointed, and some strategies to improve computational efficiency of NMPC are summarized.2) A one-step nonlinear model predictive control based on affine nonlinear system is proposed. The algorithm has analytic solution, low computations and above all it is simple. Also a modified one-step NMPC algorithm based on disturbance model is presented. Computer simulations on model mismatching verify the proposed method.3) In order to avoid the conventional iterative optimization methods are very sensitive to the initialization of the algorithm, genetic algorithm is used to optimize the control sequence in NMPC. In order to reduce the computational load, a suboptimal nonlinear model predictive control algorithm is proposed and a proof of nominal stability of the closed-loop system is also given. Computer simulations on continuous stirred tank reactor are carried out to corroborate the effectiveness of the method And compared with traditional NMPC, suboptimal NMPC has also been validated the superiority.4) To reduce the severe computational burden, stair-like NMPC algorithm is proposed with combining NMPC with stair like control. Simulation results prove the correctness of the method and effectiveness of the proposed controller.5) Nonlinear model predictive control strategy comes from the real industrial process, and goes back to solve the control problem existed in real process. By using hydrodynamics theory, the mechanics principle of tank system is analyzed and the system nonlinear mathematics model is established. Three proposed algorithms are applied into the plant and the effectiveness of three algorithms is demonstrated and the nonlinear system identification method is also verified. In this dissertation, stair-like NMPC algorithm is also applied to the simulation study of the sensor motion control system in mobile wireless sensor networks.The main results and some problems to be solved in the future are presented in the conclusion.