Studies On Simulation And Optimization Of Complex Separation Processes

A typical chemical engineering process consists of pre-treatment, reaction, and separation sections. The separation section plays an important role to obtain pure products. In this thesis, the modeling, simulation, and optimization of commercial complex separation processes with features of large scale, high purity, thermal-coupling, and varying loads are studied.As it is well known, the thermodynamic model plays an important role to the modeling accuracy. For a high purity separation system, it is difficult to derive an accurate thermodynamic model via experiments under traditional conditions. Also because of the nonlinear feature in the high purity region, it is not acceptable by using extrapolation. The method to directly use engineering data for the thermodynamic parameter estimation is proposed in this thesis. The study shows that at a high purity region the parameter estimation of the thermodynamic model is hardly influenced by the process model uncertainty, which is different from the normal case and makes the estimation feasible. Applications to real industrial systems are also presented.Because the separation section is normally the last step of a chemical engineering plant, its feeds are frequently affected by previous sections and product demands, which means the feed condition may often deviate from the designed specifications. If the process operation is still kept at the designed condition, it may result in a non-optimal operation when the feeds deviate. Therefore, an observer for feed states is essential for the optimal operation for a separation system. A first principle model based method to predict the feed variation by using the column temperature profile is proposed in this thesis. Theoretical analysis is given in agreement with simulation results under different cases.The multi-component separation is often implemented by a distillation column sequence, in which a number of distillation columns are used in serial. This results in a large scale system for the simulation and optimization tasks. The existence of thermal-coupling feature among equipments further makes the process model complicated with tight constraints. Newton method and its variants are normally used for the fast process simulation and optimization. However, a good initial is always required for a successful convergence. The lack of the good initial guess often makes the simulation and optimization to fail in many cases. In this thesis, a homotopy-based backtracking method (HBM) is proposed to deal with this problem. Applications show that this method has obviously better performance in process simulation and optimization with wide range operations.The cryogenic air separation unit (ASU) is a process that provides large-volume and high-purity products of oxygen, nitrogen, and argon. This process consists of almost all the features of a complex separation system under study in this thesis, including large-scale, high-purity, thermo-coupling, and load variation. As a typical complex separation process, the modeling, simulations and optimizations of ASUs are conducted. Two typical flowsheets, the internally compressed and the externally compressed cryogenic air separation plants, are discussed. Besides the thermodynamic parameter estimation, variables setting and data reconciliation are also discussed. The HBM has been successfully applied to deal with the key problems of the RTO associated with the automatic load control (ALC) system. With an objective function of the optimal the energy cost and constraints for product quality specifications, the RTO provides reference values to act as a feedforward of the MPC, which greatly reduces the transition time of the automatic load control. Further analysis shows that this method also helps to locate the physically feasible boundary among various requirements of load change.