Research On Optimal Control Methods Integrated With Modeling And Optimization On Polyvinyl Chloride Production Process

Polyvinyl Chloride(PVC)production process is a large-scale industrial process and has very complicated chemical reactions.This production process mainly has the following characteristics: strong coupling,non-linearity between variables,large time lags of system parameters,and slow time variability.It is a very common complex controlled object.In the production process,there will also be many uncertain factors.If traditional control methods are still used,industrial indicators cannot be met.Due to the emergence of advanced control methods,these complex objects can be better controlled.This paper proposes a neural network soft-sensor modeling method based on data dimension reduction methods on polyvinyl chloride production process.The seven commonly data dimension reduction methods were used to reduce the dimension of high-dimensional input data for the proposed soft-sensor model of the PVC production process,that is to say to reduce the field data involved in the aggregation process to three-dimensional data.Simulation research and prediction of vinyl chloride conversion rate were achieved by using a variety of methods,including the gradient method,clustering method and orthogonal least square method for RBF neural network and the training algorithm for dynamic fuzzy neural network.According to the simulation results,it is proved that the neural network soft-sensor modeling method based on data reduction strategy for PVC production process can accurately predict the indicators of PVC polymerization production process.Based on the basic principle of PVC stripping process,the subspace modeling method,the decoupling method of muti-variable coupling system and PID controller parameter setting method,a decoupling control method of PVC stripping process based on subspace modeling is proposed.Then the subspace modeling method was used to model "top temperature-slurry flow" and "bottom temperature-steam flow" in the PVC stripping process.The diagonal matrix decoupling and feed-forward compensation decoupling methods are used to decouple the TITO system to obtain two SISO systems,which are controlled by two PID controllers respectively.PID parameters are tuned by using four engineering tuning methods to obtain step response curves.For different controlled objects,four PID controller parameter adjustment methods will show different effects.The polyvinyl chloride distillation process is introduced.For the controlled distillation process,the PID parameter optimization algorithms are introduced in details.From the optimization point of view,the optimal values are found in the three parameter spaces of Kp,Ki,and Kd,which makes the control effect of the system be optimal.The particle swarm optimization(PSO)algorithm,bacterial foraging(BF)algorithm and the combination of the two algorithms(PSO-BF)are adopted to carry out the analysis and comparison on the PID controller for the distillation model.The influence of PID controller parameters optimization algorithms on system performance are analyzed.The parameters and curves optimized by three optimization algorithms are obtained.The simulation results show that these three optimization algorithms have their overemphasis on accuracy,speed and stability.