Modeling And Solving For Production Scheduling Optimization Of A Refinery Plant

Production scheduling is important for the refineries because it concerns many sections such as inventory, supply chain, blending, equipment processing etc. In order to gain the maximum profits, it is necessary to make a good scheduling. This thesis presents four production scheduling models for a refinery by using mathematical programming, and different solving algorithms respectively. The main works of this thesis are as follows:A production scheduling model of a single goal linear programming for a refinery plant is presented. There are five processing units in the refinery plant which includes two atmospheric vacuum distillation units, a catalytic cracker, a solvent oil equipment and a gas separation unit. The objective function of the optimization is to maximize the profit of the refinery in a period and the constraints are mass balance, equipment processing capacity quality demand, tank capacity, energy consumption constraint and inventory constraint.Because of the nonlinearity of gasoline octane number in oil blending, a model of single goal nonlinear programming is addressed according to the quadratic nonlinearity oil blending model by Twu-Coon method. A hybrid genetic algorithm (GA), which makes use of the position displacement strategy of the particle swarm optimizer (PSO) as a mutation operation, is applied to solve the nonlinear programming model. The model is validated by the result researched.Considering the uncertain price in the oil market, a production scheduling model of fuzzy linear programming is presented based on the theory of fuzzy programming. The fuzzy programming is transformed into a deterministic multi-objective programming by taking three typical numbers which are minimum values, most probable values and maximum values within the price range instead of the fuzzy number, and then solved by using minimax algorithm. On account of the time-varying factors, a dynamic programming model with multistage production scheduling is addressed with the profits as objective function and mass balance, equipment processing capacity quality demand, tank capacity, energy consumption and inventory as constraints. The decision variables are the production of the refinery product in multistage and the state variables are inventory. Because of the inequality constraints, ordinary method for solving dynamic programming is unsuitable. In this thesis, the total income expression is expanded into the sum of profits in each period, and then the dynamic programming is transformed into nonlinear programming with inequality constraints which is solved by using hybrid genetic algorithm. At last, a calculating example is given to verify the validity of the model.