Study On Sliding Mode Control Method And Its Application For Coke Oven Gas Collection Process Based On Data-Driven Technique

Coke oven gas collection system plays a vital role in the coking process.Gas collection process by effectively recycling coke gas is an important procedure in the iron and steel production,which not only saves energy but also reduces environmental pollution.The gas-collector pressure of coke ovens is an important parameter in the gas collection process,and its stability directly affects the quality of coke and gas,the life-time of ovens as well as the production environment.Therefore,it is vital to study the control problem of gas collecting process of coke ovens.The coke oven gas collection process is a highly complex industrial process,which features strong coupling,nonlinearity,time-variation,large disturbance,therefore hard in modeling.Consequently,it is difficult to achieve satisfactory control performance only with the conventional control methods.Thus,for the gas-collector pressure control problem,a data-driven sliding mode decoupling control strategy is developed,providing a new approach in solving the gas-collector pressure control problem.The main contributions of this dissertation are fivefold:(1)To deal with the complex mechanism,various coupling factors and unmeasurable disturbance in coke oven gas collection process,this dissertation proposes a novel hybrid nonlinear dynamic mathematical model.Different from the previous linear models of the gas collector pressure system,the proposed model contains both gas collector pressure system(GCPS)and blast blower recovery and cooling system(BBRCS).Firstly,the first principle modeling method is adopted to deal with the disturbance,nonlinearity and coupling,in GCPS.Secondly,the neural network modeling method is applied to cope with the complex technological process in BBRCS.Thirdly,by applying the Runge-Kutta method,the two established mode! are combined and the coupling between the two subsystems are established and analyzed.Finally,the numerical simulations and analysis are performed for the established model in terms of dynamic characteristics of the gas collection process.The simulation results show that the gas collection process is a non-affine nonlinear system with strong coupling,time-variation and large disturbances,and it is nearly impossible to obtain its accurate model or nominal model.(2)Because of the problems of time-variation,coupling,large disturbance in the unknown gas collection process,an adaptive second-order sliding mode decoupling control approach with data-driven sliding surface is proposed.Firstly,by utilizing the non-parametric dynamic linearization technique(NDLT)and the extended state observer(ESO),a novel data-driven sliding surface is designed which can establish the relation between the desired trajectories and the control inputs.Then,an adaptive second-order sliding mode decoupling control law based on the proposed sliding surface(DSS-SSMDC)is then derived to deal with couplings,uncertainties and external disturbances.Moreover,the theoretical analysis also demonstrates that the states of closed-loop system are asymptotically stable.Finally,a numerical example of gas collection process is given to prove the effectiveness of the theoretical results.(3)As there exists measurable disturbance and the PI controller has been widely applied in gas collection process,a hybrid variable structure PI control(HVSPIC)approach is further proposed.The methodology is developed based on the framework of data-driven sliding mode control(SMC)with PI control,decoupling control and feedforward control.Firstly,the PI controller stabilizes the pressure of gas collectors.Then,the extended state observer(ESO)is utilized to facilitate the decoupling controller design.Finally,a feedforward signal is also incorporated into the controller design to suppress the disturbance from blast blower.By applying the SMC strategy,the robustness of the whole control approach is ensured and the superiority of the PI control scheme could also be reserved.Moreover,the asymptotic stability could also be guaranteed in theory.In addition,the simulation results show that the proposed controller has satisfactory tracking performance regardless of the problems of couplings,time-variation and disturbances.(4)As the measurement disturbance exists extensively in the coke oven gas collection process and the stability can hardly be guaranteed for the available control methods under this condition,a data-driven robust output tracking control(DROTC)scheme is proposed.By combining the advantages of sliding mode control(SMC)and data-driven control(DDC),the pressure control problem in gas collection process could be solved with the proposed approach.Unlike the conventional DDC approach,the proposed controller is based on SMC framework,where a novel hybrid control structure and a new data-driven sliding surface are developed to facilitate the controller design.On one hand,the couplings,disturbances and uncertainties are suppressed owing to the application of the SMC technique and the extended state observer(ESO),in which robustness of the control system could also be ensured.On the other hand,the stability of the DROTC system with bounded measurement disturbances could also be guaranteed by applying the DDC law.Finally,the numerical results demonstrate that the developed controller exhibits stronger robustness,less conservativeness and better adaptability than that of the DROTC controller.(5)Based on the distributed control system(DCS)of the actual coke oven gas collection process and with the help of the OLE-for-process-control(OPC)technique,the proposed DROTC approach is applied to the actual gas collection process.The results demonstrate that the developed controller has better tracking control performance than that of the PI controller and fuzzy controller,and it has less complex structure,less parameters,lower conservativeness,better decoupling ability and stronger disturbance rejection ability,which can provide a better approach in solving the gas-collector pressure control problem.