Research On The Soft Sensing And Synthetic Optimizing Control Of Distillation Process In Petroleum Refinery

With the rapid development of national economy, the amount of petroleum and its products needed in China is increasing greatly. The industries of petroleum refining are developing rapidly, however the environment is polluted and the energy is also consumed. Continuous distillation process is a very popular unit in the petroleum refining enterprise. The crude oil is separated by the continuous distillation process with the energy supplied. The continuous distillation process is the equipment which consumes the most energy in the petroleum refining enterprise. Its consuming energy is about 1/3 of all energy consumed in the factory. Nowadays, the national petroleum refining enterprise is almost all controlled under the distributed control systems. However, as the restrain of measurement and process control strategy, the performance of the distillation process control is always bad and the energy wasted in the distillation process is huge.The efficiency strategy is taking the advanced process control and optimal control technology to increase the products and decrease the energy wasting. At the same time, the distributed control system supplies the enough software and hardware environments to apply those technologies. The study and application of the advanced process control and optimal control technology to distillation process is meaningful to the petroleum refining enterprise.We take the continuous distillation process as the research object and analysis the theory of separation deeply. The theory of dynamic mechanical model, soft sensor, inferential control and synthetic optimizing control for the distillation processes are studied in this paper. The main contribution and achievements of this dissertation are stated as follows:(1) The dynamic simulation system of the distillation process is very important to the study of advanced process control and optimal control. Firstly, we build the steady model of the distillation column by analysing the theory of separation. Secondly, in order to develop the dynamic simulation system of the gas fractionation equipments, the propane distillation column is chosen as a case and a rigorous equilibrium dynamic model is developed. Based on this model, the distillation process system is also simulated and analyzed. To improve the computing efficiency, a novel algorithm based on the support vector machine regression to predict the vapor-liquid equilibrium is proposed. In order to solve the differential-algebraic equations, the results of the steady-state mechanical model are used as the initialization value of the dynamic model. The simulation results show that the value of the dynamic model is consistent with the actual production value and the computing efficiency is improved. The new dynamic model would be applied for the dynamic simulation system developing and the advanced process control strategy designing.(2) The composition of distillation products is a very important quality value in refineries, but unfortunately few hardware sensors are available on-line to distillation compositions. In this dissertation, a novel method using sensitivity matrix analysis and kernel ridge regression to implement on-line soft sensing of distillation compositions is proposed. In the approach, the sensitivity matrix analysis is presented to select the most suitable secondary variables to be used as the soft sensor's inputs. The kernel ridge regression is used to build the composition soft sensor. Application to a simulated distillation column demonstrates the effectiveness of the method.(3) The design of distillation process and the soft sensing are always separated. As the secondary variables to the soft sensor are fixed in the petroleum refining enterprise, the application of the soft sensor to the distillation process will invest high costs and the generation of the soft sensor is always bad. In this dissertation, a novel integrated soft sensing method is proposed. Firstly, the secondary variables are selected based on the dynamic simulation model of the distillation column and the performance of the modeling is also discussed. The simulation results show that the temperature tray near the input tray is sensitive to the distillate compositions. The optimal secondary variables selection and kernel ridge regression method will improve the performance of the soft sensor. Based on the simulation results, the product soft sensing of an atmospheric column is discussed. Through collecting the value of secondary variables and the primary variable, the dry point soft sensor of aviation kerosene is build. The experimental results show that the soft sensor has a good performance of generation and the absolute error is less than 3℃which will match the requirement of the petroleum refining enterprise.(4) As taking the general control strategy such as indirectly temperature control to control the distillation concentration, the control performance is always bad. A novel cascade inferential control strategy based on kernel ridge regression soft sensor is proposed. Firstly, the ability of inferential control based on linear model is analyzed. Secondly, the ability of the indirectly temperature control, the inferential control based on soft sensor and the novel cascade inferential control are simulated based on the dynamic simulation system of the distillation column. The simulation results show that the novel inferential control is efficiency. Based on those results, we design an inferential control strategy for the atmospheric column.(5) The theory of distillation is analyzed deeply, in order to improve the products output and decrease the energy wasting, the object functions and constraint conditions of the gas distillation process are built and a synthetic optimizing control strategy is proposed. A two-layer optimization structure is proposed to solve the optimal problem. Outer layer takes the PSO optimal algorithm; inner layer uses the steady simulation of the mechanical model. The simulation results show that the energy wasting is decreasing by using the synthetic optimizing control strategy. The distillation process will change its optimal set point according to the enterprise needed and the profit is maximal. The optimal objects of energy and product efficiency are achieved.