Fusion Modeling And Optimization Of The Ethylene Cracking Furnace

The ethylene is one of the most important chemical raw materials, which is the foundation of the petrochemical industry and plays an important role in the national economy. Its production level has become an important symbol to value the development states of petrochemical industrial of a country. The modern large and complex industrial process has not been established the precise mathematical model based on physical chemistry process. It is more and more difficult to control and optimize these complex chemical processes. In recent years, how to effectively use a large number of offline or online data to build an accurate model is a hot field of control theory and engineering, which achieves the process modeling, optimal control and decision-making, maintenance and management.The ethylene cracking furnace production system was selected as a research and application in this article. A systematic study is from the cracking material to end of the cracking cycle. The adaptability of raw pyrolysis model is studied how to be suitable to the operating conditions and different furnaces. The process modeling and optimization method is studied based on the data and mechanistic models according to the ethylene cracking furnace process, in order to seek a new modeling and optimization control method, improve the operation of the device level and the overall economic efficiency of the product. Specific work and research results are as follows:(1) Review the current status and trends of ethylene industrial production, which it has been to large-scale ethylene plant, cracking feedstock diversification, reduce costs, energy consumption et al.The pyrolysis oil is an urgent need to optimize the composition, seek a different oil pyrolysis process model product. Then summarize advantages and disadvantages, the essential difference, mutual relationship between data and model-driven modeling approach. The effective integration of modeling problem-solving process can resolve the problems of process modeling and optimization. Finally, the swarm intelligence optimization algorithm and its improved methods were reviewed, and the current domestic process modeling and optimization control aspects were analysized, and the key elements needed were pointed out for further research fields.(2) Take SL-I cracker as an example, which was developed jointly by Sinopec and Lummus corporation, the mechanism model was established of ethylene cracking process. As to poor adaptability shortcoming of the original kumar molecular reaction dynamics model for naphtha cracking, a sequential quadratic programming chaotic particle swarm optimization (SQPCPSO) was put forward, and the optimization adjustment solution for the reaction selectivity coefficient and the secondary kinetic parameters were proposed. Based on balances of carbon and hydrogen atoms in the reaction, as well as the model and actual yield error, the objective function and adjustment approach were studied. The accuracy of adjusted model is higher than the original kumar model. Different applications of different naphthas verified the effectiveness and practicality of the proposed solution.(3) Similar recognition of different naphtha properties based on fuzzy clustering algorithm is proposed. The model library of different kumar reaction kinetics for different clustering oils was built, at the same time, different types of oil kinetics model library of optimized operation was built also. The fusion modeling method based on data and mechanistic model is proposed according to process operation data, process data and experience knowledge of ethylene cracking furnace process. As to online data modeling accuracy and re-learning problems, a new Moving Window principal component analysis (WMPCA) algorithm based on adaptive soft threshold method to extract principal component, and a WMPCA-RBF modeling approach was proposed online self-learning adjustment process to improve the online modeling accuracy.(4) On above work, the smart optimization operation method for ethylene cracking furnace process was proposed. A multi-group competition Adaptive Particle Swarm Optimization (FCMAPSO) was proposed and used to optimize the smart integration model. Combining with multi oils similarity condition pattern recognition, a different model of intelligent optimization predictive control was built. A software system was developed to optimize the operational condition of practical ethylene cracking furnace, the practical application results verified the effectiveness and practicality of the proposed methods.(5) For multi-objective optimization problem of ethylene cracking furnace, multi-objective optimization strategy of reconciliation comprehensive evaluation method was researched and designed. An adaptive multi-objective particle swarm dynamic evolutionary algorithm was proposed based on hierarchy analysis, real-time dynamic and adaptive evolutionary particle state parameters to improve the distribution and diversity of the target solution. Considering multiple targets and operating constraints, such as the yield, coke thickness and operating cycle, a coordinated and balanced strategy between different targets was achieved according to the selecting preferences between different yields and operating conditions. The proposed approach provides a possible solution for multi-objective optimization of ethylene cracking furnace.