深圳大学学报理工版

中国新疆玛湖致密砂砾岩油藏非均质性强,各井地质和施工条件参差不齐,压裂设计方案缺乏针对性,压裂效果的主控因素也有待进一步深化认识.为此,建立了一套基于机器学习的玛湖地区水平井产能预测及压裂设计优化的流程框架.以玛湖地区75口压裂水平井为例,针对储层和工程2类16个影响因素,采用随机森林算法确定了压后产能主控因素,建立了遗传算法优化的逆传播算法神经网络产能预测模型,并基于此对水平井的压裂进行优化设计.结果表明,现场2口井在优选压裂参数方案后,产能分别提高了9.3%和37.3%.建立的产能预测及压裂设计优化方法可为现场施工提供指导.

The Mahu tight glutenite reservoir is highly heterogeneous, and the geology and construction conditions of each well are uneven. The fracturing design scheme lacks pertinence. The main control factors of the fracturing effect need to be further understood. Therefore, we establish a set of machine learning-based process frameworks for horizontal well productivity prediction and fracturing design optimization in the Mahu area. Taking 75 fractured horizontal wells in the Mahu area as an example, based on 16 influencing factors of 2 types of geology and engineering, the random forest algorithm is used to determine the main control factors of post-fracturing productivity, and a error back Propagation(BP)neural network productivity prediction model optimized by genetic algorithm is established. Based on this, the optimal design of fracturing for horizontal wells is carried out. The results show that the productivity of two wells in the field is increased by 9.3% and 37.3% respectively after the fracturing parameter scheme is optimized. The established productivity prediction and fracturing design optimization methods can be applied to other oil fields and provide targeted guidance for on-site construction.

引言
1 特征参数选取及预处理
2 压裂水平井产能预测模型建立
3 单井压裂设计优化
4 结论

图1 箱型图识别平均砂比<br/>Fig.1 Box diagram to identify average sand ratio

图1 箱型图识别平均砂比
Fig.1 Box diagram to identify average sand ratio

图2 遗传算法优化BP神经网络过程<br/>Fig.2 BP neural network process optimized by genetic algorithm

图2 遗传算法优化BP神经网络过程
Fig.2 BP neural network process optimized by genetic algorithm

图3 训练样本与测试样本产量的原始分布情况<br/>Fig.3 The original distribution of the output of training samples and test samples

图3 训练样本与测试样本产量的原始分布情况
Fig.3 The original distribution of the output of training samples and test samples

图4 测试集样本预测结果<br/>Fig.4 Test set sample prediction results

图4 测试集样本预测结果
Fig.4 Test set sample prediction results

表1 井1和井2与邻井特征参数及预测结果<br/>Table 1 Characteristic parameters and prediction results of well 1, well 2 and adjacent wells

表1 井1和井2与邻井特征参数及预测结果
Table 1 Characteristic parameters and prediction results of well 1, well 2 and adjacent wells

表2 井1正交参数表<br/>Table 2 Well 1 orthogonal parameter table

表2 井1正交参数表
Table 2 Well 1 orthogonal parameter table

表3 井2正交参数<br/>Table 3 Well 2 orthogonal parameter

表3 井2正交参数
Table 3 Well 2 orthogonal parameter

图5 不同方案下井1预测产能与投资<br/>Fig.5 Productivity and investment of Well 1 under different plans

图5 不同方案下井1预测产能与投资
Fig.5 Productivity and investment of Well 1 under different plans

图6 不同方案下井2预测产能与投资<br/>Fig.6 Productivity and investment of Well 2 under different plans

图6 不同方案下井2预测产能与投资
Fig.6 Productivity and investment of Well 2 under different plans

[1] 任宜伟,楼宣庆,段宝江,等.工程参数对 L 区煤层气直井产量影响的定量研究[J].石油钻采工艺,2016, 38(4):487- 493.
[2] WANG Shuhua, CHEN Shengnan. Integrated well placement and fracture design optimization for multi-well pad development in tight oil reservoirs[J].Computational Geosciences,2019,23(3):471- 493.
[3] ANIFOWOSE F, LABADIN J, ABDULRAHEEM A.A hybrid of functional networks and support vector machine models for the prediction of petroleum reservoir properties[C]// The 11th International Conference on Hybrid Intelligent Systems(HIS). Melacca, Malaysia: IEEE,2011:85-90.
[4] 叶俊华, 郭娇娇, 金伟东, 等. 灰色关联法在吉006断块压后主控因素分析中的应用[J]. 中国石油石化, 2016(22):111-113.
[5] 张杰.数据挖掘技术在油井压裂施工决策中的应用[D].北京:中国科学院,2007.
[6] 陶亮,郭建春,李凌铎,等.致密油藏体积压裂水平井产能评价新方法[J].特种油气藏,2019, 26(3): 89-93,127.
[7] JIU Bo, HUANG Wenhui, LI Yuan, et al. Influence of clay minerals and cementation on pore throat of tight sandstone gas reservoir in the eastern Ordos Basin, China[J]. Journal of Natural Gas Science and Engineering, 2021, 87: 103762.
[8] ORAK N H, REEDER M, PEKNEY N J. Identifying and quantifying source contributions of air quality contaminants during unconventional shale gas extraction[J].Atmospheric Chemistry and Physics,2021,21(6):4729- 4739.
[9] BALDONI E, PHILIPPIDIS G, SPEKREIJSE J, et al. Getting your hands dirty: a data digging exercise to unearth the EU's bio-based chemical sector[J]. Renewable and Sustainable Energy Reviews, 2021, 143: 110895.
[10] 李政宏,刘永福,张立强,等.数据挖掘方法在测井岩性识别中的应用[J].断块油气田,2019, 26(6):713-718.
[11] 雷蕾, 吴乃君, 刘鹏, 等. 灵敏度分析: 分类器中的缺失数据[C]// 第11届海峡两岸信息管理发展策略研讨会论文集. 武汉: 华中科技大学出版社,2005:153-157.
[12] TRONG D D, VIET T Q, KHOA V D, et al. A two-dimensional sideways problem with random discrete data[J]. Computers and Mathematics with Applications, 2021, 86: 16-32.
[13] LEBANOV L, GHIASVAND A, PAULL B. Data handling and data analysis in metabolomic studies of essential oils using GC-MS[J]. Journal of Chromatography A, 2021,1640:461896.
[14] 樊毅龙,马先林,连建文.水力压裂缺失数据填补方法研究比较[J].石油化工应用,2020, 39(11):48-55.
[15] 宋宣毅,刘月田,马晶,等.基于灰狼算法优化的支持向量机产能预测[J].岩性油气藏,2020, 32(2):134-140.
[16] 纪磊,李菊花,肖佳林.随机森林算法在页岩气田多段压裂改造中的应用[J].大庆石油地质与开发,2020, 39(6):168-174.
[17] 林霞,刘宗尚,高宇,等.基于机器学习的产油量主控因素分析[J].信息系统工程,2019(12):94-97, 99.
[18] ZHAO Ruidong, SHI Junfeng, ZHANG Xishun, et al.Research and application of the big data analysis platform of oil and gas production[C]// International Petroleum Technology Conference. Beijing: Society of Petroleum Engineers, 2019: 1-9.
[19] YU Yinshan, SHAO Mingzhen, JIANG Lingjie, et al. Quantitative analysis of multiple components based on support vector machine(SVM)[J]. Optik, 2021, 237: 166759.
[20] BU Fankang, HE Jun, LI Haorun, et al. Interval-valued intuitionistic fuzzy MADM method based on TOPSIS and grey correlation analysis[J]. Mathematical Biosciences and Engineering, 2020, 17(5): 5584-5603.
[21] ALEXANDRINO P S L, GOMES G F, CUNHA Jr. S S. A robust optimization for damage detection using multiobjective genetic algorithm, neural network and fuzzy decision making[J]. Inverse Problems in Science and Engineering, 2020, 28(1): 21- 46.
[22] 孙兆亮. 吉木萨尔致密油藏水平井压裂裂缝优化设计研究[D]. 北京:中国石油大学(北京), 2018.

备注

引言

1 特征参数选取及预处理

2 压裂水平井产能预测模型建立

3 单井压裂设计优化

4 结论

期刊信息

备注

引言

1 特征参数选取及预处理

2 压裂水平井产能预测模型建立

3 单井压裂设计优化

4 结 论

期刊信息

4 结论