Study Of Distillation Column's Mechanism-NN Hybrid Model

Tension of energy source demand has forced distillation to use advanced process control techniques to allow operation in regions. Control strategies based on process model can be used to solve it. Success of these algorithms, such as MPC, depends a great deal on the model used to represent the process. So, modeling of distillation is more difficult due to its time-variant, nonlinearity, strong coupling characteristics.In recent years, advancement of neural network (NN) techniques gives an impulse to the development of nonlinear modeling. Combination of process knowledge with intelligent modeling technologies to model complex system has been a focus in the domain of distillation process. The dissertation establishes distillation column's mechanism-NN hybrid model, based on the mechanism model and NN compensation model. Mechanism model and hybrid model were simulated with MATLAB.At first, history of distillation techniques was briefly introduced in this paper. Then, mechanism modeling of distillation column was chiefly introduced. On the one hand, the parameters of alcohol distillation column were detailedly calculated, and four equilibrium equations of the distillation process were built. On the other hand, mechanism modeling of distillation column was established by analyzing hypothesis condition of simplified mechanism model. And the mechanism model was simulated in MATLAB. RBF compensation model was established in order to compensate stable error of mechanism model. After introducing briefly hybrid modeling theory, combination of mechanism model with NN model was established and simulated in MATLAB. At the same time, disturbance performance of hybrid model was analyzed with simulation. The result of simulation shows that hybrid model have a good antidisarrangement performance and error of hybrid model was limited within±1°C. Finally, research work of this dissertation was summarized and future work was prospected.