Boruta-Optuna-XGBoost融合模型的聚驱油田产量智能预测方法研究

扫码查看

原文链接

国家科技期刊平台
NETL
NSTL
万方数据

中文摘要：在聚驱油田开采时,存在向油层中加注水溶性高分子聚合物的过程,此过程包含大量参数,对油田产量具有较大影响.针对传统油田产量预测方法存在的人工计算量大、准确率低的问题.提出一种Boruta-Optuna-XGBoost融合模型对聚驱油田产量进行预测,解决传统方法存在的问题.通过Boruta特征筛选方法进行聚驱油田特征筛选,降低特征冗余,提高特征相关性,避免模型过拟合;使用Optuna超参数优化算法对XGBoost进行自适应超参数评价,提高模型精度;使用最优超参的XGBoost算法对聚驱油田产量进行回归预测,通过算法建立油田注入参数和油田月产量之间的逻辑关系模型,对聚驱油田的月产量进行预测.所提方法应用在大庆油田的实际有效数据的准确率达95％,证明了方法的有效性,能够对油田的生产效益、资源配置和可持续发展产生影响,也为数字化聚驱油田智能产量预测发展提供了新思路.

外文标题：Model intelligent prediction of polymer flooding oilfield production based on Boruta-Optuna-XGBoost

外文摘要：Accurate prediction of polymer flooding oilfield production plays an important role in the development planning of oilfields.In the oil recovery process of polymer flooding oil fields, the process of injecting water-soluble polymer into the oil layer requires many parameters and exerts a significant impact on the oilfield production.To address the issues of heavy computations and low accuracy in traditional oilfield production prediction methods, this paper proposes a Boruta-Optuna-XGBoost fusion model to predict the production of polymer flooding oil fields.First, feature redundancy is reduced, feature correlation is improved, and model over-fitting is prevented by the Boruta feature selection method for polymer flooding oilfield feature selection.Second, the Optuna hyperparameter optimization algorithm is employed to evaluate XGBoost adaptive hyperparameters and improve model accuracy.Finally, the optimal hyperparameter XGBoost algorithm is employed to regress and predict the production of polymer flooding oilfields.A logical relationship model between oilfield injection parameters and monthly production is established through the algorithm to predict the monthly production of polymer flooding oilfields.The method is applied to the collected data from Daqing Oilfield, with an accuracy rate of 95％, demonstrating its effectiveness and improving the production efficiency, resource allocation, and sustainable development of the oilfield.It also provides a new way for the intelligent production prediction in digital polymer flooding oilfields.

外文关键词：

polymer flooding oil fieldsproduction forecastfeature screeninghyperparametric evaluationXGBoost

作者：

田枫、曹凯光、赵玲、张孟阳、刘芳、苏若禹、常丽娟

展开 >

作者单位：

东北石油大学计算机与信息技术学院,黑龙江大庆 163318

关键词：

聚驱油田产量预测特征筛选超参评价 XGBoost

基金：

黑龙江省自然科学基金黑龙江省高等教育教学改革项目黑龙江省省属本科高等学校基本科研业务费项目黑龙江省哲学社会科学研究规划年度项目

项目编号：

LH2021F004SJGY202101492022TSTD-0322EDE389

出版年：

2024

DOI：

10.3969/j.issn.1674-8425(z).2024.04.021

重庆理工大学学报

重庆理工大学

重庆理工大学学报

CSTPCD北大核心

影响因子：0.567

ISSN：1674-8425

年,卷(期)：2024.38(7)

参考文献量24