用于炼油厂原油库存调度的混合模拟退火算法

Application of Hybrid Simulated Annealing Algorithm of Crude Inventory Management in Refinery

现有的数学规划法在解决原油库存调度优化问题时存在着组合爆炸的问题,是阻碍调度优化实用化的主要原因。由于实践中往往只要求快速地获得一个较好解,因此作为启发式算法之一的模拟退火法,在解决调度问题的实用化方面具有很大的优越性。但由于模拟退火法较适于处理无约束的整数规划问题,而在原油库存调度优化模型中却存在着大量的实数约束,所以在其中直接应用模拟退火法比较困难。该文将模拟退火法与线性规划法相结合,以前者调动后者,后者为前者提供可行解判据,构成了一种优化混合算法。在将混合算法应用于原油库存调度问题时,该文采用了特定的编码方式,使各控制变量在随机变化时尽量满足相关的约束条件,从而避免了许多无效解的产生。实例计算结果表明,同传统的混合整数线性规划方法相比,这种混合算法可以快速地给出优化解,其优化值与全局最优值差别不大,表明混合算法可以更好地解决实际原油调度问题。

Crude scheduling problems are mainly optimized by mathematical programming at present,but there exists combinatorial explosion phenomenon within this method for large scale scheduling problems,which makes this method infeasible for real-life applications.In practice it is,however,always only required to get a better solution in a relatively short period while scheduling.As one of heuristic methods,simulated annealing （SA）is more applicable to meet this requirement.But it is difficult to apply SA to scheduling problems directly because SA is initially designed for integer programming while there are lots of real variable constraints in crude scheduling model.Based on this knowledge,a hybrid algorithm is proposed in this paper combining SA with linear programming （LP）,in which the latter is responsible for feasibility checking for these real variable constraints under the control of the former.Then this hybrid algorithm is used in crude scheduling,and special coding procedures are devised in order to keep intermediate solutions feasible as much as possible.The final calculation result shows that,compared with mixed integer linear programming （MILP） ,the hybrid algorithm can promptly provide a good solution which has almost the same optimum as MILP,and is more applicable to deal with real-life crude scheduling problems consequently.