Parameter Estimation And Real-time Optimization Of Load Change Process In Air Separation Unit

The Large-scale air separation unit(ASU)uses cryogenic distillation to provide massive high-purity gases products for steel,chemical and other industries.The ASU consumes a lot of energy,which can make up to 75%of the product cost.With the demand periodicity of downstream gas,ASU needs to change production load widely and smoothly on the premise of ensuring the quality of high-purity gas,otherwise it will lead to a sharp increase in oxygen vent rate and energy consumption.The ASU has a complicated process with high degree nonlinearity.Therefore,manual load change operations can be difficult,so it is important to realize a model-based real-time optimization that is able to change the load automatically.In order to solve the problems above,this study mainly focuses on the rigorous modeling,parameter estimation and large-scale load change real-time optimization of the ASU.The main contributions are as follows:1.Equation-oriented rigorous modeling of the ASU.An equation-oriented steady-state rigorous model of the ASU is constructed,including thermodynamic properties calculation and MESH equations.In order to reduce the difference between the actual system and the ideal model,the segmented tray efficiency parameters are introduced to the ideal model.Then,the improved model is verified by being compared with the simulation results of the ideal model.This model becomes the basis of parameter estimation and real-time optimization.2.Parameter estimation and parameter estimability analysis.Based on the steady-state rigorous model,the parameter estimation of the ASU is constructed,which is used to estimate the binary interaction factor and the tray efficiency in the model.In order to solve the inestimability of some parameters caused by limited measurement points and measurement noise,a systematic parameter estimability analysis method is proposed to filter out the subset of parameters for future estimation.The method includes hierarchical clustering analysis based on sensitivity vectors,and post-optimization analysis based on confidence intervals.Then the above method is applied to the ASU and validated by large measurement noises.The result illustrates that the method can effectively select out the estimated parameters and improve the accuracy of the model.3.Load change real-time optimization under model mismatch.Firstly,based on the rigorous model of ASU,the load change operation optimization of ASU is constructed.In order to solve the degradation of the optimation caused by model mismatch in the large-scale load change process,a real-time optimization method is designed,which consists of model update based on the parameter estimation and load change operation optimization.And the initial value strategy of iterative calculation to enhance convergence is designed.Finally,the method is applied to real-time optimization simulation of large-scale load change ASU.The result shows that in the case of large measurement noises,the real-time optimization algorithm can achieve stable load change operation and converge to the optimal operating point while achieving satisfied product quality and output.