Algorithm Research On Mixed Integer Quadratic Programming Of Oil Refining Scheduling Model

Quadratic programming problem, as a special type of optimization problem, has a very important significance in operations research now. In real life, Quadratic programming has been used in numerous applications, such as economy, finance, image processing, chemical, engineering control, molecular biology, environmental engineering and other fields, so in study of either local optimization or global optimization problems, quadratic programming has always been a important subject of study. Refinery mixed integer quadratic programming scheduling model is a kind of large-scale optimization problem, the research of solution algorithm to solve the model has theoretical significance and practical value.In this paper, we start from dynamic planning process of the oil refining industry, to analyze a type of the quadratic programming-positive semidefinite quadratic programming problems, which will be analyzed and discussed from the results. As we know, a problem can be transformed into an equivalent convex problem, so we just need to use the algorithm which can find a local minimum solution to get the final global optimal solution. Therefore, we propose a hybrid algorithm which is proposed based on improved interior point method and branch and bound algorithm. First, combine the logarithm penalty function method and quasi-Newton algorithm, to convert the problem into the unconstrained form, then use Wolf-Powell line search to determine the step size to avoid small search step. Then analyze the convergence of the algorithm, to let the algorithm terminated after a finite step by setting a reasonable stopping criterion. After the relaxation problem is solved with the modified Newton interior point method, According to the characteristics of the original function, this paper developed a special branch of strategies to improve branch and bound method to make an efficient access to feasible solutions to satisfy the integer constraints. Use own characteristics based on convex programming, we can get the optimal solution of the original problem. Finally, we use this hybrid algorithm to solve the refinery mixed integer quadratic programming scheduling model, and compared with the existing similar algorithm to show the effectiveness of the proposed algorithm.