Chance Constrained Model Predictive Control And Double-Integrator System Real-Time Model Predictive Control

Model predictive control is one of the most widely applied control methods in process manufacturing industry. Compared with other controllers, its superiority is the ability to handle multi-variable operational problems and inequality constraint condition easily. Rolling optimization strategy and feedback correction action improve the robustness performance of the controlled system. Conventional model predictive controller encounters some difficulties in the design and application, such as the optimization involved costs too much time in online calculation or under the uncertain factor's influences, it is hard to maintain inequality constraint condition and so on. Aiming at these problems, this thesis brings forwards two kinds of new design methods for model predictive controller, and they have achieved remarkable effects in practical application. Specifically, the work of this thesis is described as follows:1. We design an energy minimum model predictive controller for a fixed two-point boundary value, input constrained, dual integral system. In view of dual integral system being a simple system, this thesis uses Pontryagin minimum principle and geometry analysis to off-line calculate the functional relation of optimal control and current state. When implementing the model predictive control stategy, we only need to look up the table to get values of the control signal. So this method reduces the on-line calculation difficulty, and makes it satisfy the requirement of high-required real-time control occasion.2. In practical production process, random disturbance and uncertain model parameter's influences will break constraints condition, eventually result in dropping the control performance. Aiming at this problem, this thesis designs a chance constrained model predictive controller based on set-valued optimal algorithm. In view of the model of MAC being linear, we can transform the original chance constraint programming problem to a quadratic programming problem using set-valued optimization and multi-inclusion's property of robust control strategy. So we can use nominal optimization algorithm to solve the problem. Applying this controller to high-purity distillation process, we get satisfying results.