Recipe Optimizing And Short-Term Scheduling For Oil Blending

Oil blending is a key unit in refinery. This unit includes many kinds of components and products, hundreds of oil storage tanks, different order demands and rigid quality test, relates to a large number of complicated physics and chemical reactions, and it is a large scale scheduling problem. It is very complicated to analyze, optimize and simulate such a large system, especially mathematical model is the emphasis. Therefore, mathematical model become an important research direction in refinery scheduling.Firstly, this thesis synthesizes domestic and international research fruits and current situation of oil blending, and according to the control think of modularization and hiberarchy, presents a overall solution for oil blending. Secondly, the existing recipe optimization models of blending processes are not accurate and fast enough, so this thesis puts forward a new linear programming model for recipe optimization. In order to preserve the model's linearity, an iterative procedure is proposed to effectively deal with non-linear oil properties, which can quickly provide optimum recipes for oil blending. Lastly, this thesis presents a mixed-integer linear programming formulation that addresses the simultaneous optimization of the recipe optimizing and short-term scheduling problem. Depending on the difference of time domain, a discrete-time model and a continuous-time model are set up respectively. Both of them are also dealt with an iterative procedure. People can base on the size and characteristics of oil blending problem, choose one of them and get a flexible and effective solution. The new method is illustrated by solving real world problems in refinery with very low computational requirements.