Research On The Multi-objective Optimal Operation Of The Steam Network In The Refinery

The optimal operation of steam network of refinery has significant impact on safe production and cost control. The steam network includes refinery production equipments and steam-driven power system which are connected and mutually affected through steam and fuel flows. Many researchers conducted researches on the steam network operation optimization, but these researches often perform optimal design of boilers, turbines and steam pipes, which limited to steam-driven power system, without the consideration of integrated optimization of production process. The high energy-cost issue of refinery in our country is stilled unsolved. This thesis is based on national science and technology foundation. In this thesis, a multi-objective operation optimization method of refinery steam network is given by which realizes the integrated optimization of refinery steam-driven power system and production process. This thesis mainly includes the followings:(1) The current research situation of operation optimization in refinery steam network is introduced which includes:First, the description of production process, production equipment and the equipment of steam-driven power system come in detail. Second, the key points of modeling are covered, which include raw material supply, equipment schedules, production cost, multi-period data and so on. Third, Analysis of material flow among refinery equipments and steam-driven power system is given. Based on this analysis, refinery process chart and topological mapping are constructed.(2) During the selection process, the non-dominant NSGAII with elite strategy is without considering density and random initialization of the population, which causes drawbacks on convergence speed and uniformity of solution distribution. Targeted on these two drawbacks, improvements are given:First, not only based on sequence effect, the population density is considered when conducting selection process. To achieve this, the individual adaptation value is is calculated to replace irank. Second, initial values of certain part of individuals are given during the population initialization, which directs the convergence of the population. Third, the test and evaluation of improved algorithm is given through stand test question, which demonstrates the improvement of NSGAII on convergence speed and uniformity of Pareto solution distribution.(3) The single period mathematical modeling with multi-goal is built for steam network. Solution is given by using the improved NSGAII, at the same time, compared with the solution obtained from GAMS. And then, analysis of the solutions and verification of the validity of the model and the improved NSGAII are performed.(4) Based on the single period model, the shift between production schedules and inventory are merged into the model. This extension casts the original problem into MINLP problem with multi-period and multi-goal. To solve this MINLP problem:First, decomposition is conducted to make it into single-goal problem and solve it by GAMS. The solutions obtained from GAMS are used as parts of the initial population of improved NSGAII. Then, improved NSGAII and follow-up analysis are performed which verify the validity of the model and the improved NSGAII.