| Oilfield water injection network system is a complex network system which is consistedof water injection stationsã€water distributionã€water injection network and water injectionwells, and its topology optimization problem involves large-scale hybrid optimization designproblems which include discrete optimization〠nonlinear optimization〠multi-objectivefunction optimization problems and so on, currently the oilfield water injection networksystem is still under the artificial planning and designing stage, since there is no moreefficient algorithms to solve these problems and difficult to obtain better results.This article is based on bi-level programming theory exploring the oilfield star waterinjection network topology optimization problem; the main work to be carried out includesthree aspects as follows:1ã€On the foundation of deeply analyzing the oilfield star water injection network’stopology features, describing the problems based on bi-level programming theory. In thebi-level programming theory, the upper level is layout optimization problems which includethe site optimization of water distributionã€diameter optimization and so forth, and the lowerlevel is the water consumption needs of water injection wells which is under the control ofwaterflood development indicators system, that is to say it is access optimization betweenwater injection wells and water injection stations which belong to combinatorial optimizationproblem of discrete variables.2ã€Based on the bi-level programming theory, we can respectively using pipe networksystem’s minimum investment or minimum annual cost of the discounted value as the upperlevel’s objective function to build the mathematical model of oilfield star water injectionnetwork’s topology optimization, meanwhile in the lower level we can also use the minimumwater injection pipeline head loss as the objective function to fix the membership betweenwater injection stations and water injection wells.3ã€According to the structure characteristics of the bi-level programming model, we canuse the electromagnetic-like mechanism algorithm to solve the model which has theadvantage of simple structureã€few parametersã€fast convergence rate. But toward thealgorithm’s problem of premature convergence and initial value dependent, we adopt themethods of control population diversity and simplifying the total force calculation formula toimprove the optimization performance of algorithm, furthermore the validity and reliability of the algorithm was also shown by the example calculation. |