Study On Modeling And Advanced Control Strategy Of Industrial Cut Tobacco Drying Process

Industrial drying is an energy-intensive process,and most industrial drying processes have low energy efficiency.With rising energy costs and increasingly fierce global competition,the drying process' s energy consumption and quality performance must be improved.Researchers often focus on the mechanism and simulation of drying process,but seldom on the operation and control of drying process.The main cost of drying process is not in the initial investment(design and assembly),but in the daily operation of drying process optimization.Control strategies are essential to improve energy efficiency and obtain the ideal quality of dried products.The improvement method and strategy is to establish a reasonable drying model and use effective control strategies to optimize the drying process.In industrial systems,most optimal control strategies are designed based on models.In the control regulation problem,the model's predicted values are used to generate optimal control actions;in the estimation problem,the optimal state and parameter estimation of the system are generated based on the coordination of the model's predicted values and the actual industrial measurement data.The system model is helpful to understand the internal mechanism behavior of the system and the basis of collaborative optimization of the whole control system.Therefore,the drying process modeling is the first problem to be solved in the control research.Most of the drying models are established from the perspective of technology,and the main function is to simulate the drying process and understand the changing laws of physical variables in the drying process.The models are very complex and many involve high-dimensional partial differential dynamic models,so it is difficult to use these models for real-time control strategy of drying process.As a control strategy,researchers pay more attention to the influence of models on the subsequent control strategy implementation and the linear models are too simplified to accurately describe the complex drying process,and the coupling between variables is rarely considered.The nonlinear model based on the first principle/mechanism(energy,mass and momentum balance)can not only accurately describe the complex dynamic characteristics of the drying process,but also can easily transplant and extend the use of these models in other drying processes/conditions.In order to obtain the optimal drying conditions,better quality performance and higher energy efficiency of the drying process,an optimal control strategy based on the first principle model is usually the first choice.The main research work of this paper is as follows:Firstly,this article establishes a fourth-order nonlinear first principles model based on the objective analysis of the relevant variable factors of the actual drying process(cut tobacco drying process),the principal component analysis of the drying process data,and the mechanism analysis of the drying process.The first-principles model is general,not only has potent model portability and expansibility for other drying processes and can build highly complex and accurate system models.Secondly,because of the unmeasurable or difficult to measure state variables and physical parameters in the industrial drying process,it is estimated through the moving horizon estimation algorithm that can handle nonlinearity and constraints,avoiding much effort to measure and verify some difficult to measure physical parameters during the drying process and provide accurate models for subsequent optimization control.Finally,aiming at the control problem of insufficient degree of freedom in the industrial drying process,three control strategies were designed to optimize the control system,and all of them achieved good control results.The main innovations of this paper are as follows:(1)A nonlinear moving horizon estimation strategy for cut tobacco drying process is designed.Because the cut tobacco drying process is a nonlinear model,and there are unknown parameters and state variables and system constraints which are difficult to measure,it is difficult to deal with the nonlinear problems and system constraints by conventional timedomain estimation strategy,and it is not possible to simultaneously estimate the unknown parameters and state variables of the drying process model.The nonlinear moving horizon estimation strategy(L1-Norm Moving Horizon Estimation,L1-Norm MHE)designed in this paper can explicitly deal with system nonlinearity and various constraints.Based on the time-domain moving window,the optimization strategy only uses the nearest time-domain window data to estimate the system state variables and unknown parameters.Compared with the full-information time-domain estimation strategy,this strategy has accurate estimation results and reduces the optimization calculation load and calculation time.The simulation results show that L1-norm MHE is more robust and accurate than other algorithms in the face of complex conditions(data outliers,noise,data drift).The estimation strategy provides real-time state and unknown parameter estimation for the optimization control layer of the subsequent industrial drying process,and improves the optimization control effect of the drying process.(2)A zone nonlinear model predictive control strategy for cut tobacco drying process is proposed.As the cut tobacco drying system is a multivariable non-square model(the number of manipulated variables of the model is less than the number of output variables),there is insufficient control freedom.Conventional model predictive control often leads to steady-state error in output variables,and control accuracy and product quality may be significantly affected.The zone model predictive control(Zone Model Predictive Control,ZMPC)proposed in this paper controls the cut tobacco outlet moisture w at the setpoint value without deviation.Other controlled output variables does not need to be strictly controlled on the setpoint value,and the control requirements of other output variables are relaxed as long as they are within the given setpoint zone.The strategy of relaxing the setpoint value improves the control freedom of the system to a certain extent,satisfies the control requirements of the system's key output variables,and eliminates the steady-state error of the output variables.The zone model predictive control's unique feature is the use of zone reference trajectory,only when the model's predicted value exceeds this zone reference trajectory,the optimizer will change the manipulated variable.Compared with the traditional setpoint tracking model predictive control,the zone model predictive control(ZMPC)has better tracking performance and robust performance,as well as the minimum action economic characteristics of the controller.The key is to keep the cut tobacco outlet moisture track setpoint without any steady-state error.(3)A priority multi-objective nonlinear model predictive control strategy for cut tobacco drying process is studied.Aiming at the tobacco drying process of the nonlinear multivariable model,there is a contradiction of mutual coupling and competition among the system's output variables.Besides,the control system is a non-square model with an insufficient degree of freedom,so how to satisfy the most critical output variable of the system under the limited operating variables is an urgent problem to be solved by the non-square system.This paper proposes to introduce the priority multi-objective optimization strategy into the model predictive control strategy framework.The use of priority multi-objective optimization is the best solution strategy to deal with the competition between system objectives.The drying process controlled output variables are optimized in ascending order of priority,and the process requirements of the controlled output variables with high priority are met first.Given the controlled output variables that may be subject to different target constraints,after determining the specific controlled output variables' priority,this variable's target constraints are divided in descending order of priority.The target constraints with low priority are first relaxed.Once the optimization is feasible,stop relaxing other high-priority target constraints,and finally make the system output variables move along the best target trajectory.Through the simulation verification of the control strategy,the priority multi-objective control strategy prioritizes meeting the target requirements of the cut tobacco outlet moisture,and reduces the other three output variable control objectives to achieve the optimal control effect of the drying process.Compared with the zone model predictive control,the designed control strategy more reflects the industrial operator's subjective desire for the drying process.(4)A two-layer nonlinear model predictive control strategy for cut tobacco drying process is developed.Further analysis of the drying process model reveals that the compatibility and uniqueness of the input and output steady-state values of the system are all caused by the unreasonable setpoint value of the upper optimization(Real Time Optimization,RTO)and insufficient degrees of freedom,which leads to the inability to solve the steady-state relationship between input and output.The root cause of the drying process output variable's steady-state error is that the setpoint value of the output variable is unreasonable.In response to this root cause,this paper adds a steady-state objective optimization layer(Steady State Objective Optimization,SSTO)between the upper optimization(RTO)and the control layer and reoptimizes the setpoint values of the output variables in combination with the current stage of the process,and the two-layer nonlinear model predictive control control(Steady State Objective Optimization-Model Predictive Control,SSTO-MPC)strategy is developed.Through control strategy verification,SSTO-MPC control strategy has better tracking ability and antiinterference ability than traditional model predictive control.Compared with zone model predictive control and priority multi-objective model predictive control strategy,SSTO-MPC control strategy has stricter theoretical optimization operation setpoint,the implementation is more scientific and reasonable.