Oil blending is the most important step of producing refined oil, This step includes many kinds of components and products, many oil storage tanks, different order demands and rigid quality test, relates to a large number of complicated physics and chemical reactions. It is very complicated and crucial to analyze, optimize and simulate such a large system. Therefore, mathematical model become an important research direction in refinery scheduling. And resolving blending receipt is key and difficult.Firstly, this thesis synthesizes domestic and international research fruits and current situation of oil blending, and compares current oil blending method. Establish a new gasoline blending receipt model because of current oil blending receipt model is not exact and has too much solution process. This thesis is based on accumulated penalty function method and optimizing arithmetic operators of genetic algorithm, improves inequality constrained nonlinear programming genetic algorithm, and use this method to solve oil blending receipt function. By comparing simulation results with the existing advanced methods results to verify the validity and wide adaptability of this method. This improved method can raise refinery profits efficiently. Then bases on Dalian Petrochemical practical situation to simulate oil blending with tank- residual oil, improving this method have wide applicability.Finally, this thesis researches the online pipeline oil blending method by using NIR as detector and introduces the acquiring and effect online detector value, online rectifying could be implement by using this method.
|