新型线性先行GA算法在油品调和中的应用  

Application of a New Linear First GA Algorithm in Oil Blending

作　　者：李琦[1] 李响[1] 邵诚[1] 

机构地区：[1]大连理工大学辽宁省工业装备先进控制系统重点实验室,辽宁大连116024

出　　处：《控制工程》2012年第1期94-98,共5页Control Engineering of China

基　　金：国家973资助项目(2007CB714006);中央高校基本科研资助项目(893360)

摘　　要：油品调和是生产成品油的关键环节,调和配方决定了利润的最终大小。配方优化模型中物料平衡约束、产品产量约束属于线性约束,但质量指标约束中存在的非线性约束,使整个求解问题成为非线性约束求解问题。针对炼油厂油品调和复杂非线性约束配方优化问题,提出了一种新的线性先行GA算法。通过将线性约束与非线性约束拆分开,先求解凸多面体的全部顶点和极方向,解出满足线性约束的所有可行解所在的可行区域,进而用GA算法在此可行域内,通过优化凸组合的系数,从而达到搜索整个线性解区间的目的,完成非线性优化模型的求解。仿真结果表明,新算法大大缩小了GA算法的搜索空间和需要处理的约束,能快速的获得理想的调和配方,使利润最大并保证对质量卡边的要求。Oil blending is a key step of producing refined oil, the blending recipe determined profits in refineries. Mass balance and production volume constrains are linear constrains. But there are non-linear constrains in quality index constrain. This made the recipe optimization become a non-linear problem. It is difficult to obtain satisfying optimum solution by traditional methods. According to this problem, a new linear first GA algorithm was presented. The algorithm proposed a new idea dividing the constraints into liner and non- liner constraints. First, found out all the extreme points and direction of polyhedron, these extreme points and direction construct the feasible region of the liner constrains. And then deal with the non-liner constrains with GA algorithm in the feasible region. By optimi- zing convex combination coefficients to achieve searching all linear feasible region. The simulation results present that the search space and constraints of GA can be reduced greatly through the new algorithm. The ideal blending recipes can be found quickly, and get the maximum profit with a little margin of quality index.

关 键 词：线性先行 非线性约束 凸多面体 遗传算法 油品调和 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]