Performance Parameters Prediction Research Of Gas Membrane Separation Process Based On Soft Sensor

On-line detection of the key performance parameters (i.e., permeate fast gas concentration, permeate gas flow and exhaust fast gas concentration) are important research content in optimal control of gas membrane separation process. However, they are always cannot or difficult to measure in real time, still measured by artificial analysis on a lot of sample. The measure accuracy is hard to guarantee for the auxiliary variables are time-varying and the influence of environment. When there is a big error appeared, it cannot provide an effective guidance for its process optimization and real-time control.In allusion to the problems mentioned above, the paper, combined with the hot research spot of optimal control study in current complex industrial process and intelligent detection, used PCA to analysis the different influence of auxiliary variables to the primary variables in gas membrane separation process, to simplify its process modeling. Then, based on soft measurement intelligent modeling and field data, made a study on whether intelligent modeling in different technological process and predicting the performance parameters of gas membrane separation process are applicable. The study content includes the prediction research of performance parameters in one-stage refinery gas hydrogen recovery membrane separation process based on PCA-RBFNN and PCA-LSSVM, the prediction research of performance parameters in two-stage refinery gas hydrogen recovery membrane separation process based on PCA-LSSVM and in two-stage CO2separating from natural gas membrane separation process based on BPNN and PSO-BPNN, the on-line prediction research of the performance parameters in three-stage refinery gas hydrogen recovery membrane separation process based on LSSVM and in two-stage CO2separating from natural gas membrane separation process based on LSSVM.Based on MATLAB software platform and field data, the paper simulated the offline/online models that had designed above and made prediction study on the key performance parameters of the one-stage, the two-stage and the multistage process in gas membrane separation process. Simulation results show that the designed intelligent measurement models combined with gas membrane separation process can effectively predict its important performance parameters. This study provides a good foundation for the real-time control and optimization control of gas membrane separation process.