Improved OA Algorithm And Its Application In The Production Scheduling Of Process Industries

The typical process industries include nonferrous metallurgy, petrochemical and other mainstay industries. The mathematical optimization model for production scheduling in process industries is usually a MINLP model, which contains discrete variables, continuous variables and nonlinear functions. The research on the algorithms for solving MINLP model has always been focus and difficulty. As one of the popular MINLP methods, OA algorithm need to be further improved.This thesis focuses on the improvement strategies of OA algorithm for solving MINLP model, the application on Benchmark problems and three real production scheduling problems of process industries. Numerical experiments show the distinct efficiency of the improvement strategies of OA algorithm.1) Build a flexible platform for realizing OA algorithm. Select the appropriate optimization modules for solving sub-problem, which are integrated into the unified platform for realizing the improved strategies of OA algorithm.2) Present the improved strategy of adding multiple-cut simultaneously. Further we also propose the following accelerated strategies:faster judgment conditions about whether there are feasible solutions for NLP sub-problem and designing high-quality initial solution based on problem characteristics. Numerical experiments on classic periodic scheduling models show that the iteration numbers are significantly reduced, and better solutions could be gotten for some examples.3) Apply the improved OA algorithm into the scheduling problem of real ethylene production and integrated scheduling optimization of well production and maintenance. These two models are both pseudo-convex MINLP, whose characteristics are large-scale problems and long solving time consuming. Compared with DICOPT++, a standard package of OA algorithm, the improved OA algorithm could get the optimal solution with less iteration. 4) Formulate the scheduling problem of aluminum electrolytic and continuous casting production into one MINLP model, and solve the model using the presented improved OA algorithm. The scheduling result could effectively reduce the waiting time of the liquid aluminum and energy loss. The decision support system based on the presented model and the improved OA algorithm is developed, which include several functions such as data management, optimization module and manual adjustment.The improved strategies of OA algorithm significantly improve the solving efficiency for the actual production scheduling problem of process industries, which are significant for solving actual scale MINLP problem.