In this paper, a sequential quadratic programming method combined with a trust region globalization strategy is analyzed and studied for solving a certain nonlinear constrained optimization problem with matrix variables. The optimization problem is derived from the infinite-horizon linear quadratic control problem for discrete-time systems when a complete set of state variables is not available. Moreover, a parametrization approach is introduced that does not require starting a feasible solution to initiate the proposed SQP trust region method. To demonstrate the effectiveness of the method, some numerical results are presented in detail.
2
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Predictive control based on linear models has become a mature technology in the last decade. Many successful real-time applications can be found in almost every sector of industry. Nonlinear predictive control can further increase the performance of this easy-to-understand control strategy. One of the main problems of implementing nonlinear predictive control is the computational aspect, which is of most importance in real-life applications. In this paper, suboptimal nonlinear predictive control strategies are proposed and compared. The nonlinear predictors are built based on neural identification methods or by white modelling. The use of diophantine equations, which is a common technique to calculate the optimal contribution of the noise model, is avoided by using a more natural method. The comparison between the control algorithms is made based on a simulated discrete multivariable nonlinear system and a continuous stirred tank reactor.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.