Research On Hybrid Modelling Algorithm Of Kappa Number Prediction In Displacement Digester System

Black liquor in the pulping process is one of the major pollution with the characteristics of high viscosity, high organic content and high chroma. It’s very harmful to the environment if the black liquor is discharged directly. One of the effective ways to solve this problem is recycling use. Displacement disgester system (DDS) sets a lot of black liquor storage tanks to reuse the black liquor at different time, so DDS is energy saving and environment protection pulping technology. However, due to the difficulty of sampling and lacking of reliable, economical and practical Kappa number measuring instrument, it’s difficult for DDS to control the pulp quality. Pulp quality is mostly controled by the experience of skilled operator or H factor, and the effect is not ideal. Putting forward practical kappa number measuremnet strategy is a research hotspot of paper industry. This thesis is based on the displacement cooking DCS project of a Sichuan Pulp Paper enterprise, with funding support of ISTCP (Cooperation research on the key technology and equipment of energy saving and environmental protection displacement digester). The main research contents of this thesis include the displacement disgester process and mechanism analyzing, secondary variables selecting, soft measurement modeling and the kappa number prediction system designing, and the details are presented as following.Ⅰ The selection of modeling vaviables for the displacement disgester kappa number modelingBased on the analysis on the difference and connection between traditional batch cooking and displacement disgester, the influential factors of kappa number prediction are determined. At the same time, the H factor, black liquor amount, chip weight and other 11 paramenters were selected as the kappa number modeling varibles, which ensured the accuracy and stability of the kappa number model, and simplified the model structure.Ⅱ Building of the displacement disgester kappa number hybrid model.Recursive PLS method and RBF neural network were respectively used to built displacement cooking kappa number model. Acording to the comparison of the simulation results, the predictions of the two models were both bad. Because of the defects of the two models, a hybrid kappa number model was proposed based on recursive PLS model and RBF neural networks model,which was realized by weighting average of the two model predicted results. The simulation results showed that the prediction accuracy and relative error of hybrid model was improved significantly, it was superior to using either recursive PLS model or RBF neural networks model.Ⅲ Implementation of the displacement disgester kappa number online prediction system.An implementation of displacement digester kappa number soft measurement system was given by using Siemens S7-400 controller hardware, and PCS7 software development platform. The system used the OPC typical client/server mode, which realized real-time online soft measurement of the kappa number through the OPC protocol.In this paper, the hybrid model was built by weighting average of the two model predicted results, so the prediction accuracy of the kappa number was improved significantly. The program combined existing DCS system, and the implementations of displacement disgester kappa number online prediction system were based on Matlab and OPC protocol and Industrial Ethernet communication. Engineering application results showed that the program was effective.