Double-layered Predictive Control System Design

In large process industry,control objectives proposed by the production enterprise are generally realized via hierarchical control system,since single control method is unable to meet all the requirements in most cases.The execution flow of a typical hierarchical control system can be simplified as follows.Upper real-time optimization(RTO)layer obtains the ideal operating points by optimizing the production target globally,then transmits the optimized set-points to the middle dynamic constrained control layer,which solves a multi-objective coordinated optimization problem under an assuring feasible region for the purpose of fulfilling the external targets from RTO to a great extent.The solution of the middle layer turns to be the regulating set-points by PID controller of lower basic control layer.The overall control targets could be actualized with the information coordination of these multiple layers.Model predictive control(MPC)is a highly accepted dynamic control algorithm in the industrial process area,since it can handle different kinds of constraints,consider multiple control demands comprehensively in the form of soft constraints or the performance index,improve the dynamic property in the control course and provide a strong robustness and anti-interference with its receding optimization and feedback compensation mechanisms.These advantages of the model predictive control fit the control requirement of the middle layer well in the hierarchical control system.Also,it is capable of solving the deficiency in handling constraints and lack in economic benefit by traditional control method,which satisfies the practical engineering need.Hence,how to combine the hierarchical control system and model predictive control,how to design corresponding control algorithm to optimize the economic benefit in the production process to a great extent on the basis of realizing the control requirement and how to enlarge the scope of the designed control system are the key points of this thesis.Aiming at these problems,systematic research in the following four aspects have done in this thesis.Firstly,a double-layered model predictive control system based on steady state target calculation(SSTC)is designed,which can be applied in the middle layer of the hierarchical control system.In such control system,SSTC including the feasibility and economic optimization stages is in its upper layer.The optimizing results are realized by model predictive control in the lower layer.Secondly,based on the study of SSTC,a new control strategy which is called dynamic trajectory calculation(DTC)is presented.In the designed control system,a multi-dimensional input increment sequence is selected in DTC instead of one-dimensional steady state input increment vector in SSTC as the optimization variable.Accordingly,the outputs vary from one-dimensional steady-state output increment vector in SSTC to a multi-dimensional output increment sequence in DTC in the constraints formation and performance index.As consequence,the tracking targets of model predictive control in the lower layer transfer from a single steady-state output target in SSTC to a series of future dynamic trajectories in DTC.Then,the complexity of the control object is considered.The former SSTC and DTC is used for the multi-input-multi-output control object.Based on such research work,the influence of measurable disturbances are considered,accordingly,the control object is extended to the multi-input-multi-disturbance-multi-output.The applicability of the two predictive control systems is enhanced.Finally,aiming at the practical production process,the dynamics of control course is considered furthermore.The constraints related to the amplitude and set-points from upper layer are prioritized offline in the designed double-layered predictive control system.Due to the production process and target are varying continuously,the prioritized constraints and set-points might be changing.The concept of time-varying multi-prioritized constraints is proposed,which can be applied in more complex production process.All research works in this thesis are based on theoretic analysis and algorithm derivation,then select the important production processes in the petrochemical field to be the control object.Validate the effectiveness of the two proposed double-layered predictive control systems by simulation.