Index Prediction And Operation Optimization Of Ethylene Distillation Column Based On SCN

Distillation process is the most extensive mass transfer unit operation process in the petrochemical industry and chemical industry,and it is also one of the most energy-intensive unit operation processes in the petrochemical field.In this thesis,the main research object is the ethylene distillation column,which is the operating equipment that can separate cracking gas to obtain ethylene products.The process improvement and operation optimization on this device can fully improve the production efficiency,and can effectively help enterprises to improve profits and economic benefits.The distillation process is relatively complex and has many interference factors.The characteristics are summarized as follows.First of all,the distillation column has many column plates,which has strong nonlinear characteristics and is a high-order object type which leads to a relatively slow response to the control input.Secondly,the distillation process is influenced by a variety of factors,which is due to the fact that the device includes many control loop,and the control loop is not independent each other.In the actual production process,the production index is mainly measured by the industrial chromatograph or the manual measurement,which causes a lot of cost rise.At present,the research on the ethylene yield problem mainly focuses on the establishment of a single model,which does not cover everything and has poor generalization ability.The change of the external conditions will cause a drop in the prediction accuracy.Overall,the research object in this thesis is the ethylene distillation column.In consideration of the actual production situation in the enterprise,the model construction and the operation optimization are investigated.The main aspects of work are as follows.（1）Firstly,the subject background and the current research status at home and abroad in this field are investigated.Then,the working principle,process flow and control scheme of the ethylene distillation column are briefly introduced.Finally,the factors affecting the ethylene yield of the distillation process are explained.（2）After briefly introducing the technological mechanism and process flow of the ethylene distillation process,the technological parameters affecting the ethylene yield are determined,such as the steam flow rate at the bottom of the column,the return flow rate at the top of the column,the temperature at the top and the kettle of the column,the feed flow rate,the sensitive plate temperature and the pressure at the top of the column,etc.To solve the problem of model failure caused by the change of working conditions,a multi-model modeling method based on fuzzy C-means clustering algorithm（FCM）is proposed to divide working conditions.In this paper,refrigerant temperature and unit load are used as working condition variables to classify.In this thesis,the refrigerant temperature and device load are used as the operating condition variables to conduct the clustering division.When screening the main influencing parameters of ethylene yield,the dimension reduction was carried out according to principal component analysis（PCA）.Finally,the input variables of X₁,X₂,X₃,X₄were determined as the model,and the output variables were ethylene yield.A multi-operation model of ethylene yield under different operation conditions was established by using random configuration neural network（SCN）.The simulation experiment and analysis show that the proposed modelling method is more effective and accurate.（3）Through in-depth analysis and consideration of the technological requirements of the ethylene distillation process,the optimization model with the objective of maximizing the ethylene yield is established,which is in line with the company’s operation and actual production conditions.The optimization strategy is proposed to solve the above optimization model by using GA and SQP.In addition,the actual ethylene distillation process will be affected by the inevitable external conditions or production process,such as the environment,surrounding noise,cost constraints and a series of influences.Therefore,it is impossible to establish a accurate process model that matches the actual ethylene distillation process completely,that is,the model has certain uncertainties（parameters,structure or process disturbances）.In this thesis,the hierarchical structure of the production decision is studied deeply and the specific tasks in different layers are analyzed.An adaptive real-time optimization strategy is proposed to solve the model mismatch in the ethylene distillation process,namely,an adaptive real-time optimization method based on the modified terms,with the core of real-time optimization layer（RTO）,which is used to compensate the influence of model mismatch on the optimization results of the original optimization problem.The effectiveness of the proposed method is verified by simulation experiments,and the optimization strategy has strong practical significance and value,which lays an important foundation for the successfully implementation of the whole process optimization control of the ethylene distillation process.