基于计算智能的油田措施规划模型及优化算法研究

The subject is carried out based on the typical problems in the optimization of oilfield development and adjustment. The multi-objective programming model is explored to establish by taking into consideration various factors, and using the computational intelligence theory to solve the related model. First, The micro pore structure types are identified according to the logging information based on the model of the limit process learning machine, and the dynamic and static classification of the single well and the well group is realized by using the fuzzy comprehensive evaluation theory; Secondly, the multi-objective and multi-constraint annual adjustment plan and multi-year measures adjusted stochastic programming model are established. In the model, the different single wells, the construction month, the law of the output decline and the uncertainty of the oil price are considered; Aim at the problem that the shuffled frog leaping algorithm is not high accuracy and easy to fall into local optimum, Combined with the advantages of cloud model theory, reverse learning theory, quantum computation and cultural algorithm, The convergence of the algorithm is analyzed. Modified shuffled frog leaping algorithm is used to optimize the limit process learning machine model and solving programming model, using practical development data validation, the result is verified through actual oil field development data.

课题面向油田开发调整规划优化中的典型问题开展研究，探索建立考虑各种因素的多目标多约束规划模型，并利用计算智能理论对相关模型进行求解。首先，采用极限过程学习机模型，根据测井信息进行微观孔隙结构类型的识别，利用模糊综合评判理论对单井及井组进行动静态分类完成措施井优选；其次，充分考虑不同单井的措施潜力、措施的施工月份、措施实施后的产量递减规律及油价的不确定性等方面，建立多目标多约束的年度措施调整确定规划模型和多年度措施调整随机规划模型；最后，针对混洗蛙跳算法求解精度不高和易陷入局部最优的特点，结合云模型理论、反向学习理论、量子计算和其他智能算法的优势对其进行改进，并对算法的收敛性进行分析，采用改进的蛙跳算法优化极限过程学习机模型和求解规划模型，给出最优工作规划方案，利用油田实际开发数据进行验证。

课题面向油田开发调整规划优化中的典型问题开展研究，探索建立考虑各种因素的多目标多约束规划模型，并利用计算智能理论对相关模型进行求解。首先，结合过程神经元网络和极限学习机各自的优势构建了一种极限过程学习机模型，提出基于改进混洗蛙跳算法对模型的网络隐层节点个数及网络权值进行优化。将测井曲线作为模型输入对储层微观孔隙结构类型进行识别，并将研究的学习机模型推广应用到抽油机井故障诊断；其次，利用油、水井的储层微观孔隙结构类型和宏观的各类动静态数据，基于模糊综合评判理论对单井及井组进行静态分类，并提出了一种基于弗雷歇距离算法的井组开发效果对标算法，建立了井组注入用量和采出程度关系对标曲线，快速优选出待实施措施的井集合；再次，充分考虑不同油、水井的措施潜力、措施的施工月份、产量递减规律等方面建立了油田措施规划单年度多目标优化模型、油田措施规划多年度单目标优化模型、油田措施规划多年度多目标优化模型3种油田措施规划优化模型；最后，针对混洗蛙跳算法求解精度不高和易陷入局部最优的特点，提出了离散混洗蛙跳算法、进化策略自主选择混洗蛙跳算法、文化混洗蛙跳算法、元胞混洗蛙跳算法和基于灰狼优化的改进混洗蛙跳算法5种混洗蛙跳改进算法，并将其应用到油田措施规划优化模型求解及极限过程学习机模型学习。将研究成果编制成软件，利用实际数据优化油田措施规划方案，从智能优化结果来看，安排结果符合油田开发实际，很好的提高了综合调整方案的经济效益，且所研究的理论成果具有很大的推广应用价值。