Online Optimization Correction Technology Of Model Parameter And Its Application Research In Pvc Production Process

In most process industries, parameters of fundamental process model derived from first principles often significantly changed with time. To implement efficient process optimize operation, model parameters often demanded online estimation. There are two methods for online estimation, filter and moving horizon estimation. The former often used to handle unconstrained linear system. However, because polymerization process was characterized by mechanism complex, nonlinear, varying parameters and uncertainty, so there were some difficulties in the application of filter methods. MHE was a practical method for solving online optimizes correction in nonlinear models. It transformed estimation into optimize problem, while contained the constraint condition of system at the same time. MHE used fixed number data, reduced the dimension of optimize problem and computational cost.This article mainly includes the following:(1) This paper researched particular mechanism model of Vinyl Chloride suspension polymerization, establish detailed polymerization process mechanism model. But there were challenges of applying detailed polymerization process mechanism model in online optimize correction. Though make reasonable assumptions on detailed mechanism model, a reduced mechanism model was obtained. It also provided a model basis for online estimation of model parameter.(2) The advantages and disadvantages of different online estimation method were discussed. Moving Horizon Estimation (MHE) method was choused as basic algorithm for online estimation of nonlinear models. But process was characterized by highly nonlinear, hard derivation, so a combination of Monte Carlo and MHE was proposed. MHE was used to obtain objective function and MC was used to optimize of the objective function. Referenced to model characteristics of the polymerization process, we proposed a benchmark model to verified proposed algorithm. The efficient of benchmark model is estimated and predicted online using moving horizon estimation (MHE) method, and achieved good result.(3) Different calculation methods of Arrival Cost were summarized and unscented Kalman filter (UKF) was proposed to approximate calculate Arrival Cost in the objective function. Firstly, commonly used Symmetric sampling strategy was employed in UKF. The new MHE method based on UKF was used to online parameter estimation and made good performance. By means of analyses of the application results, a new sampling strategy based on adaptive switching of two strategies was proposed, and used it in the online estimation of benchmark model. Compared to another two methods, it overcame existed problem and achieved better result.(4) Because polymerization process was characterized by mechanism complex, nonlinear, varying parameters and uncertainty, a reduced mechanism model was proposed. On the basis of corrected polymerization mechanism model, time varying coefficient of vinyl chloride polymerization process is estimated and predicted online using modified MHE method. The application result indicates that the online parameter estimation based on MHE method is effective and stable. It also provided a good basis for advanced control and optimization strategies in PVC process.