| As an important part of the national economy,the sustainable development of the petrochemical industry depends on efficient and environmentally friendly ethylene production.Therefore,modeling of the ethylene industry has always been a research hotspot in academia and engineering applications.In recent years,the industrial practice related to the integrated production of refining and ethylene has emerged,and it has become a solution to effectively solve the problem of excess oil refining capacity and shortage of ethylene raw materials.The use of superstructure modeling technology to consider the interaction between the upstream and downstream of the entire ethylene process,multi-equipment and multiple plants,and optimize the structure and parameters of the ethylene industry simultaneously can further improve the optimization space of the ethylene industry,and it is necessary to conduct in-depth research.This subject is based on the process graph(P-graph,Process graph)superstructure algorithm,improves related algorithms based on domain knowledge,uses process data to study the modeling of the ethylene industry at all levels,and is used to solve the current optimized operation of the actual ethylene industry and Bottleneck identification and other issues.The main content and results are as follows:1.Guided by data-driven and domain knowledge mixed modeling ideas,the modeling method of ethylene cracking furnace was studied.First,the data-driven modeling method supplemented by mechanism guidance was studied.Using the characteristics of the cracking furnace mechanism model iteration mode similar to the structure of the long-term and short-term memory network,a relevant algorithm based on the long-and short-term memory network was selected to establish the ethylene cracking furnace model.At the same time,when the model is generalized,the cross-iterative technology is used to solve the problem of short and long practice memory network used to solve the problem of lack of generalized data when modeling the cracking furnace,and high-precision modeling is achieved.Secondly,study the modeling method which is mainly based on the mechanism model and supplemented by the data drive.Through analysis,it is found that the molecular model(Kumar model)cannot simulate the co-pyrolysis phenomenon is the main reason for its modeling error.Based on this,a radical reaction network was introduced to enhance some molecular reactions,and a K-R compound model for ethylene cracking with wider raw material adaptability was constructed.Furthermore,data driving such as raw material properties is used to control the scale of the introduced free radical network and optimize the tuning model parameters.Experiments show that,compared with the molecular model(Kumar model),the proposed KR model has better raw material adaptability.For some raw materials that are not applicable to the Kumar model,the KR model can successfully predict its yield(the yield error of key substances from 25%has dropped below 1%).2.Based on the related algorithm of process diagram,the super-structure model of ethylene whole process under the background of refining and chemical integration was established.Based on domain knowledge,the PECMA(P-graph based ethylene cracking modeling and analysis)method was proposed to solve the common difficult problems of large-scale chemical process superstructure modeling such as equipment screening and merger in the modeling process,and circular logistics processing.The optimization goal optimizes the logistics and equipment load in the superstructure with the constraints of the entire ethylene flow and equipment load as the decision variables,and the capacity of the integrated refining unit to provide light raw materials as constraints.The results of applying the proposed superstructure modeling method to actual cases show that the profitability of the process can be increased by 21.46%,and the total carbon emission of the process can be reduced by 40.27%,which verifies the effectiveness of the proposed method.At the same time,the optimal proportion of light raw materials can be obtained from the optimization results,which provides operational guidance for the efficient and low-emission operation of the entire ethylene process.3.In order to solve the problem that the process state variables cannot be expressed when the P-graph method is used to model the raw material and production scheduling problems between multiple ethylene plants,a SPBP(Scheduling Programing based on P-graph)technology based on virtual materials and virtual operation nodes is proposed.Based on genetic algorithm,it supplements the data of partial raw material cracking yield required for modeling,and establishes a superstructure model of raw material and production scheduling among multiple factories.This is the first application of P-graph method in chemical process scheduling modeling.In the case test of inter-factory scheduling based on two actual factory data,the proposed SPBP algorithm obtained the same optimal solution results as the mixed integer linear programming method,which verified the correctness of the proposed method.In addition,the proposed method also obtains a sub-optimal solution set,and a high-quality sub-optimal solution set can provide better flexibility for decision-making and analysis.4.The reasons why the P-graph method generates too many suboptimal solutions and the quality of the suboptimal solution sets are poor during the solution and analysis of the superstructure model corresponding to the large-scale chemical process are analyzed.CEPA-P-graph method to produce high-quality sub-optimal solution sets.The proposed method can remove the inefficient suboptimal solution while keeping the superstructure efficient suboptimal solution to improve the quality of the suboptimal solution set.In the test cases,the total number of suboptimal solutions dropped by 74%,and the efficient suboptimal solutions represented by the optimal solutions did not change due to the deletion of the structure,which verified that the proposed method was correct and effective.5.Taking a thermal cracking ethylene production plant in East China as an actual case,the relevant data of the plant was collected.Based on the production data and equipment design data before the refining-ethylene integration of the plant,a series of methods and algorithms were used to establish a superstructure model,and the types and supply of raw materials available for the ethylene plant after the refining-ethylene integration The data as a working condition constraint optimizes the key logistics and equipment load in the entire process.In the modeling process,the established cracking furnace model is used to supplement the complete cracking product yield data of the relevant raw materials that are necessary for the superstructure modeling but are difficult to obtain in the actual modeling process;based on the PECMA superstructure modeling method,the whole process superstructure model of ethylene is built;then the bottleneck analysis of the process is analyzed based on the suboptimal solution of the superstructure,and the correspondence between the function value of the superstructure optimal and suboptimal solutions and the structure is used to accurately locate the constraints The key bottleneck equipment for process economic efficiency improvement and emission reduction,and obtained the following conclusions:when the light hydrocarbon ratio(LHR,Light hydrocarbon ratio)of the raw material is low,the processing capacity of the ethylene rectification tower is insufficient to effectively separate the crude ethylene produced by the cracking furnace,thereby making it the current bottleneck equipment for ethylene plants;when the light hydrocarbon When it is relatively high,the load of the C4 rectification tower is too low or may not be able to operate normally,making it a bottleneck restricting the efficiency of the ethylene plant.In addition,when the ratio of light hydrocarbons is greater than 18.98%,the insufficient treatment capacity of the methane tower causes part of the methane to be sent to the flare as exhaust gas,making it a bottleneck that restricts the adverse impact of the process on the atmospheric environment.Finally,with the case of material scheduling between two actual factories,it is further verified that the CEPA-P-graph method can produce an efficient sub-optimal solution set that meets the needs of actual industrial applications.This series of conclusions will provide effective guidance for improving the economic and environmental benefits of the process by optimizing the operation and renovation and expansion of the device. |
PDF Full Text Request