Study On Advanced Process Control Technology Based On Model Predictive Control

As a typical representation of advanced process control(APC) strategy, model predictive control(MPC) is a newly promising technology with broad application landscape in process industries, such as chemical and petrochemical plants in the last two decades. From the point of maximizing the enterprize production eco-nomic benefits and implementing optimal operation of process production view, a new method for fulfilling process control by integrating economic performance with high control quality is provided for process industries in this thesis. Through accu-rate identifying the dynamic performance of present process system, MPC strategy can make the plant implement total control, the considerable economic benefit can be obtained by implementing optimal operation in enterprize production, the best approach for improving the process economic potential can be found. This can provide enterprize with decision of production for improving its capacity of compe-tition. And, APC can fulfill optimal process control through artificially employing the information of process system and management of enterprize. A control strategy with super performance can be designed and implemented, meanwhile, the indus-trial process can be maintained under superior quality by combining measurement indices and results which reflect process economic performance, not only for the processes which is nonlinear or with strongly interaction, but also for the process with large dead time.MPC is a such control strategy based on the behaviors and information which reflecting the dynamics of process. It is needed for constructing the objective func-tion of controlled process and for confirming the constrain limitations of key vari-ables which are the reflection of safety of process operation and fulfilling the emis-sion standard that ruled by environment regulations. A optimal control method of receding region control is used to solve the optimal control sequences that the process works under the optimal condition. Advanced process control based on MPC has been an active research area in process control community since it in-volves many technologies and knowledge of process, including system identification and parameter estimation, constructing the performance objective function, con-trol performance assessment and monitoring, numerical analysis computation and optimization. Taking the current trends of the technologies emerged both at home and abroad into accounts, the primary research results of theoretic and engineering aspects of APC based on MPC in this thesis will be summarized as follows:(1) For nonlinear model predictive control(NMPC) which uses artificial neural network(ANN) to predict the dynamic behavior of nonlinear process, the lower efficiency of optimization algorithm in rolling optimization results in imprecise model prediction, bad real-time control performance and degradation of con-trol quality. A novel NMPC based on adaptive network weights ANN model which used modified differential evolution(MDE) algorithm. The implemen-tation of the proposed strategy has four steps. Firstly, a process nonlinear model is constructed by using ANN method; Secondly, the network weights of this ANN model are adapted on-line through the optimization algorithm named MDE, the prediction capability of the process model can be improved highly. Thirdly, the ANN model is used in the objective function of MPC, the computation of control increment quantity in NMPC can be make more precise; In the end, the validity and feasibility of the proposed method are illustrated by simulation example of distillation column which is a nonlinear process.(2) For the problem of degradation of robustness performance in the implemen-tation of MPC for tracing their set-points precisely due to uncertain dis-turbances and noises, a method named model predictive robust zone con-trol(MPRZC) is proposed based on adapting variable weighting parameters in objective function. By using the characteristics that the variable weights can punish the errors and manipulation aggressiveness, at the every sample time interval, the MPC controller manipulates the slack parameters in objective function adaptively according to output variable errors when the uncertain disturbances exist in the process, for a key controlled variable, the MPRZC can be implemented other then nominal tracing set-points control, such vari-ables can be maintained within a zone which corresponds to production spec-ifications. The robustness and stability of the MPC can be obtained by using this proposed method when the uncertain disturbances or time variant pa-rameters exist in the process. The effectiveness of this method is illustrated by using a case study simulation.(3) For the total control performance degradation caused by inappropriate con- sidering the interactions of important channels among different sub-process and uncertain communication mechanism among different sub-MPC controller in distributed model predictive control(DMPC) framework, a coordination method based on probability density function(PDF) coordination rule which is used to coordinate multi-MPC controllers. The whole implementation is comprised of four stages. Firstly, the optimal values of key variables com-puted from upper RTO function level are sent to lower MPC level as the setpoints; Secondly, the optimal control profiles are calculated according to real-time dynamic behavior in MPC function level. However, these optimal control profiles are global optimum because the important interactions among the key sub-process are omitted when the process is modeled. In the end, such control quantities are sent to coordination level, these control quantities are coordinated through PDF coordination rule and sent to lower MPC level for implementation. The effects of control performance caused by non-global op-timal control variables can be decreased. The validation of this proposed method is illustrated via a simulation example and an industrial case study.(4) For large-scale complex process, the optimal economic performance during the production process can not be guaranteed timely caused by the often chang-ing factors of production, such as raw materials due to frequent market price change, an advanced control strategy named hierarchical control based on the integration predictive control with dynamic real-time optimization(DRTO) is proposed. The DRTO located the upper level is not the normal steady-state optimization mechanism, but dynamic optimization mechanism. Namely, the process optimization operation considering the enterprize planning, schedule, information dealt with by the controller is no longer based on normal steady-state optimization, but the dynamic optimization related with the change enterprize production factors, such as raw material price, manufacture safety, environment regulation, et.al,. When such factors change, this DRTO will carry out the optimization for planning and scheduling the total production, the optimal set-point of key process variables are derived. Such an optimiza-tion is based on a dynamic model of process, the computed optimal values are global optimum in dynamic range. The effectiveness of the proposed method is validated through a case study.