| Model predictive control (MPC) has been widely used in process industry because of its ability to deal with control problems with multi-variables, complex constraints and strong coupling. Traditional MPC algorithms aim at tracking the set points which are usually provided by the operators according to their experiences dynamically. With the development of MPC technologies, a new MPC algorithm with a steady-state target calculation layer added to the traditional MPC layer, i.e., a two-layered MPC algorithm has occurred. The set-points can be determined by steady-state optimization instead of operators and then passed to the lower layer which is responsible for dynamic control. So, compared to the performance of dynamic control, the performance of steady-steady target calculation layer is much more important because of the fact that if we want to achieve the best economic profit through control, the set-points need to be optimal. However, the research about the improvement of MPC controller’s performance mostly focus on the stability of controller, anti-disturbance capability or performance of dynamic control, attention needs to paid more to the improvement steady-steady optimization layer’s performance.In this paper, the research works were focus on steady state optimization in order to achieve a better two-layered algorithm with higher performance. The aforementioned two-layered MPC algorithm is in a LP-MPC cascade form, i.e., the steady-state optimization is accomplished by solving a LP problem. A priority based soft constraint adjustment method is proposed to solve the infeasibility problem of steady-state optimization, and shadow prices is used to tune the boundaries of steady-state optimization problem’s constraints so that a better global performance of the controller is achieved. In addition, the influence of cost coefficient on the results of steady-state optimization is analyzed from the view of feasible zone, in the end, the application strategy of two-layered MPC algorithm is demonstrated through practical cases.The main work and results of this paper are listed as follows:1. The basic idea of two-layered MPC algorithm which is of LP-MPC cascade form is introduced, and the parameter analysis based economic optimization method is discussed. Then the research contents of this paper:a new two-layered MPC algorithm with high performance is put forward.2. The weight based soft constraint tuning method is introduced to deal with the infeasibility problem of steady-state optimization. Then a new and better priority based method which makes use of priority in descending order to solve the infeasibility problem of steady-state optimization is proposed. Furthermore, the superiority of this very method to the weight based method and a priority method with the ascending order strategy is proved from the view of feasible zone’s size and computation complexity analysis.3. In order to optimize the constraint boundaries of steady state optimization problem, a shadow prices based method is proposed, its theoretical foundation is made of two parts.1) The boundaries of constraints can be fixed through the estimation of disturbance according to the process’s outputs’ history data.2) The analysis for benefit to system performance of each constraint boundaries based on their shadow prices, on the basis, a tuning strategy was proposed. Cost coefficients are defined by costs and values of the process inputs and outputs, and their setting will determine the direction of optimization. So a cost coefficient analysis method is proposed after studying cost coefficient’s effect on optimization from the view of feasible zone.4. Aiming at the applying the two-layered MPC in real process, an analysis method of process object was proposed. And then, in order to solve the CMMO problem, a priority based strategy for multi-objects coordination of steady-state optimization was proposed. In the end, the effectiveness of the two-layered MPC algorithm with high performance is demonstrated by three practical cases. |
PDF Full Text Request