| Ethylene, one of the most important monomers produced worldwide, is the essential of the entire petrochemical industry. The ethylene cracking furnace is the most important equipment in the production of ethylene. The level of operation and technique of the ethylene cracking furnace can directly affect the economic benefit of the whole ethylene company. Recently, the demand of propylene has increased dramatically and its increase ratio has exceeded that of ethylene, which has broken the supply-demand structure whose centre is only ethylene yield. So the yield of propylene has become the second objective to be maximized apart from ethylene yield. Thus, the accurate simulation of ethylene cracking furnaces and the multi-objective optimization for both ethylene and propylene are very necessary.At present, the lack of the effective method to estimate the selectivities of the first-order reaction (z) of the naphtha pyrolysis process has cumbered the application of simulation and optimization in the industrial production of ethylene. This thesis first proposed a regression method considering the second-order reactions for estimating z. The method can be used for obtaining the z of different kinds of naphtha based on the product yields from plant analytical laboratory, and then a standard database of z containing some standard samples can be established. For estimating z of the new naphtha by using the established database of z, a fuzzy matching method was proposed here. It has been shown that the yields calculated based on the estimated z are very close to the experimental ones, which demonstrates the effectiveness of the proposed method.On the basis of the accelerating simulation strategy for the ethylene cracking furnace, the optimization model of periodic operation of the ethylene cracking furnace was developed. An advanced simulated annealing method proposed in this thesis(ASA), genetic algorithms(GA) and the sequential quadratic programming (SQP) were used to solve the optimization problem, respectively. A comparison of the above three methods shows that SQP is much more efficient, which can demonstrate that the gradient method SQP is still more effective to settle the complicated constrained optimal nonlinear problem.After the analysis of the existing multi-objective optimization algorithms, a new parallel hybrid multi-objective optimization algorithm was proposed for the first time. The algorithm which using parallel NSGA-II (the non-dominated sorting genetic algorithm) module and SQP module combines the advantage of multi-objective evolutionary algorithms and the linear weighted method and improves the calculation efficiency evidently. The new algorithm is more efficient to settle the complicated nonlinear constrained problems and can be easily applied to other multi-objective optimization problems in industry.Based on the demand of the industrial production, the multi-objective optimization model of ethylene cracking furnace was established, in which there are two objectives: the yield of ethylene and the yield of propylene. Then the proposed new hybrid method and NSGA-II were used to settle this problem and obtained the Pareto front respectively. A comparison of these two methods also demonstrates that the new method is more efficient to converge and captures more spectrum of the Pareto front than NSGA-II. The Pareto solutions provided multiple optimized operation alternatives for the ethylene plant.An Ethylene Pyrolysis Simulation and Optimization System (EPSOS) were developed based on the above-mentioned research. EPSOS has been used in the ethylene plant of PetroChina Lanzhou Petrochemical Company. The long-term industrial experimental data shows that the software not only can forecast the products yields of different naphtha, but also can provide the optimized operation strategy for increasing ethylene and propylene yields simultaneously. The economic benefits obtained by implementing EPSOS in the ethylene plant were affirmed sufficiently.
|
PDF Full Text Request