机器学习预测油气产量现状

doi:10.13809/j.cnki.cn32-1825/te.2021.04.018

摘要/Abstract

摘要：

机器学习是一种通用的数据驱动分析方法,也是一个重要的油气大数据分析利用手段。油气勘探开发作为具有悠久历史和庞大数据基础的重要领域,具有很大的数据挖掘潜力。利用油气田大数据分析技术可以帮助决策者进行投资分析、风险评估、生产优化,带来巨大的经济效益。机器学习方法早已被研究人员尝试应用于油气领域相关研究,随着机器学习算法的发展,许多应用场景被不断提出,但针对具体场景的通用方案仍在探索中。笔者从最基本原理着手介绍了机器学习的建模过程,梳理了用于油气田大数据分析的3类主要机器学习方法的发展历史,结合油气田大数据的特点,讨论了油气田大数据分析利用的核心内容、目标及优势,分析了机器学习在油气领域的主要应用场景,总结了目前典型油气产量预测中存在的问题及对策。

关键词: 油气田大数据, 数据驱动模型, 产量预测, 机器学习, 智能油田

Abstract:

The machine learning is not only an important tool for oil and gas big data analysis, but also a general data-driven analysis method. As an important field with a long history and a large data base, oil and gas exploration and development has a great potential for data mining. The use of big data analysis technology for oil and gas field can help decision makers to conduct investment analysis, risk assessment and production optimization, which brings significant economic benefits. The machine learning method has been tried by the researchers applying to the researches on oil and gas. Nowadays, many application scenarios have been proposed with the development of machine learning algorithms, but general solutions for specific scenario are still divided. So that, we introduces the procedure of a machine learning modeling upon the most basic principles, and summarizes the development history of the main three kinds of machine learning methods that can be applied to oil and gas big data analysis. And then based on the characteristics of oil and gas field big data, the core contents, goals and advantages of oil and gas field big data analysis and utilization are discussed, the main application scenarios of machine learning in oil and gas field are analyzed, and the existing problems and countermeasures in typical oil and gas production prediction are summarized.

Key words: big data of oil and gas field, data-drive model, production forecast, machine learning, intelligent oilfield

中图分类号:

TE328

黄家宸,张金川. 机器学习预测油气产量现状[J]. 油气藏评价与开发, 2021, 11(4): 613-620.

HUANG Jiachen,ZHANG Jinchuan. Overview of oil and gas production forecasting by machine learning[J]. Petroleum Reservoir Evaluation and Development, 2021, 11(4): 613-620.

图/表 3

图1

表1

油气田大数据的特点及其分析利用的核心内容"

数据特点	互联网大数据	油气田大数据	油气田大数据分析利用核心内容
4V特征（数据量Volume、数据类型Variety、数据价值Value、产生速度Velocity）	数据量大;类型繁多;价值密度低;产生快、失效快	数据量大;类型繁多;价值密度高;产生快、失效慢	尽快尽早地挖掘数据之间的关系,如：地质和开发条件与产量的关系
数据产生基础	被动无意识产生,需要数据挖掘才能找到数据的价值	主动设计产生,所有测量都与油气生产有关	根据应用场景和所选分析方法,进一步进行原始数据采集
数据之间的相关性	事先未知,需要进行统计分析,统计意义通常不明确	知道定性关系,但有时难以准确定量地描述	挖掘内在机理与观测数据的本质关系而非统计关系
处理技术	使用通用的大数据处理方法,这些机器学习算法通常是基于特定问题而被发明的	带专业约束关系的处理。需要建立适合油气田大数据分析的改进过的机器学习新模型	结合专业知识选择合适的数据驱动模型,使用数学模型对数据驱动模型进行约束和优化
数据结构化特征	可用数据较为完整;对于图像、自然语言等非结构化数据,有较为成熟的预处理算法和相应机器学习模型	历史数据通常不完整,特别是数据关系缺少;一些主导产量的信息难以结构化,如：构造特征、沉积环境等	如何进行数据预处理（包括数据筛选、特征工程等）。数据预处理工作量最大,并且对最终分析效果影响最大
数据预测的准确性	信息通常完备,特征中包含预测目标的全部信息,数据量足够大时弱学习器等价于强学习器,预测准确性高	信息通常不完备,很多重要信息不能被获取或特征化,数据噪声较大,预测效果存在理论上限,预测准确性低	在样本和可选特征有限、数据随机性成分较多的情况下,选择最适合预测场景模型,尽量提高预测准确性和稳定性

表1

表2

参考文献 39

[1]	李航. 统计学习方法[M]. 北京: 清华大学出版社, 2012.
	LI Hang. Statistical learning method[M]. Beijing: Tsinghua University Press, 2012.
[2]	周志华. 机器学习[M]. 北京: 清华大学出版社, 2016.
	ZHOU Zhihua. Machine learning[M]. Beijing: Tsinghua University Press, 2016.
[3]	GOODFELLOW I, BENGIO Y, COURVILLE A. Deep learning[M]. Boston: MIT Press, 2016.
[4]	FISHER R A. The use of multiple measurements in taxonomic problems[J]. Annals of Eugenics, 1936, 7(2):179-188. doi: 10.1111/j.1469-1809.1936.tb02137.x
[5]	BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1):5-32. doi: 10.1023/A:1010933404324
[6]	CORTES C, VAPNIK V. Support-vector networks[J]. Machine Learning, 1995, 20(3):273-297.
[7]	ROSENBLATT F. The perceptron: A probabilistic model for information storage and organization in the brain[J]. Psychological Review, 1958, 65(6):386. pmid: 13602029
[8]	RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323(6088):533-536. doi: 10.1038/323533a0
[9]	CUN Y L. Generalization and network design strategies[J]. Connectionism in Perspective, 1989, 19:143-155.
[10]	ELMAN J L. Distributed representations, simple recurrent networks, and grammatical structure[J]. Machine Learning, 1991, 7(2):195-225.
[11]	GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[J]. Advances in neural information processing systems, 2014, 27.
[12]	FREUND Y. Boosting a weak learning algorithm by majority[J]. Information and Computation, 1995, 121(2):256-285. doi: 10.1006/inco.1995.1136
[13]	WARD J H. Hierarchical grouping to optimize an objective function[J]. Journal of the American Statistical Association, 1963, 58:236-244. doi: 10.1080/01621459.1963.10500845
[14]	MACQUEEN J. Some methods for classification and analysis of multivariate observations[J]. Proceedings of the fifth Berkeley symposium on mathematical statistics and probability, 1967, 1(14):281-297.
[15]	DEMPSTER A P, LAIRD N M, RUBIN D B. Maximum likelihood from incomplete data via the EM algorithm[J]. Journal of the Royal Statistical Society: Series B(Methodological), 1977, 39(1):1-22.
[16]	CHENG Y. Mean shift, mode seeking, and clustering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1995, 17(8):790-799. doi: 10.1109/34.400568
[17]	SHI J, MALIK J. Normalized cuts and image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(8):888-905. doi: 10.1109/34.868688
[18]	MIKOLOV T, KARAFIÁT M, BURGET L, et al. Recurrent neural network based language model [C]// Eleventh annual conference of the international speech communication association, Chiba, 2010.
[19]	GREGOR K, DANIHELKA I, GRAVES A, et al. Draw: A recurrent neural network for image generation [C]// International Conference on Machine Learning, PMLR, Lille, 2015.
[20]	DOYA K. Bifurcations of recurrent neural networks in gradient descent learning[J]. IEEE Transactions on Neural Networks, 1993, 1(75):218.
[21]	SCHUSTER M, PALIWAL K K. Bidirectional recurrent neural networks[J]. IEEE Transactions on Signal Processing, 1997, 45(11):2673-2681. doi: 10.1109/78.650093
[22]	GRAVES A, FERNÁNDEZ S, SCHMIDHUBER J. Multi-dimensional recurrent neural networks [C]// International conference on artificial neural networks. Springer, Berlin, 2007.
[23]	KALCHBRENNER N, DANIHELKA I, GRAVES A. Grid long short-term memory [C]// International Conference on Learning Representations, London, 2016.
[24]	BENGIO Y, SIMARD P, FRASCONI P. Learning long-term dependencies with gradient descent is difficult[J]. IEEE Transactions on Neural Networks, 1994, 2(5):157-166.
[25]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780. doi: 10.1162/neco.1997.9.8.1735
[26]	GERS F A, SCHMIDHUBER J, CUMMINS F. Learning to forget: Continual prediction with LSTM[J]. Neural Computation, 2000, 12(10):2451-2471. doi: 10.1162/089976600300015015
[27]	NAIR V, HINTON G E. Rectified linear units improve restricted boltzmann machines [C]// ICML, Haifa, 2010.
[28]	PASCANU R, MIKOLOV T, BENGIO Y. On the difficulty of training recurrent neural networks [C]// International conference on machine learning, PMLR, Atlanta, 2013.
[29]	SRIVASTAVA N. Improving neural networks with dropout[D]. Toronto: University of Toronto, 2013, 182:566.
[30]	LOLON E, HAMIDIEH K, WEIJERS L, et al. Evaluating the relationship between well parameters and production using multivariate statistical models: A middle bakken and three forks case history [C]// SPE hydraulic fracturing technology conference, Woodlands, 2016.
[31]	CLAR F H, MONACO A. Data-driven approach to optimize stimulation design in Eagle Ford formation [C]//Unconventional Resources Technology Conference, Denver, 2019.
[32]	MONTGOMERY J B, O’SULLIVAN F M. Spatial variability of tight oil well productivity and the impact of technology[J]. Applied Energy, 2017, 195:344-355. doi: 10.1016/j.apenergy.2017.03.038
[33]	ZHOU Q, KLEIT A, WANG J, et al. Evaluating gas production performances in marcellus using data mining technologies [C]//Unconventional Resources Technology Conference. Denver, 2014.
[34]	CLARKSON C R, JENSEN J L, PEDERSEN P K, et al. Innovative methods for flow-unit and pore-structure analyses in a tight siltstone and shale gas reservoir[J]. AAPG bulletin, 2012, 96(2):355-374. doi: 10.1306/05181110171
[35]	SCHUETTER J, MISHRA S, ZHONG M, et al. Data analytics for production optimization in unconventional reservoirs [C]// Unconventional Resources Technology Conference, San Antonio, 2015.
[36]	陈云天. 基于机器学习的测井曲线补全与生成研究[D]. 北京:北京大学, 2020.
	CHEN Yuntian. Research on well log completion and generation based on machine learning[D]. Beijing: Peking University, 2020.
[37]	SAGHEER A, KOTB M. Time series forecasting of petroleum production using deep LSTM recurrent networks[J]. Neurocomputing, 2019, 323:203-213. doi: 10.1016/j.neucom.2018.09.082
[38]	周于皓, 刘慧卿, 祁鹏, 等. 基于循环神经网络的缝洞型油藏油井产量预测[J]. 计算物理, 2018, 35(6):668-674.
	ZHOU Yuhao, LIU Huiqing, QI Peng, et al. Forecast of oil production in fractured-vuggy reservoir by using recurrent neural networks[J]. Chinese Journal of Computational Physics, 2018, 35(6):668-674.
[39]	谷建伟, 周梅, 李志涛, 等. 基于数据挖掘的长短期记忆网络模型油井产量预测方法[J]. 特种油气藏, 2019, 26(2):77-81.
	GU Jianwei, ZHOU Mei, LI Zhitao, et al. Oil well production forecast with long - short term memory network model based on data mining[J]. Special Oil and Gas Reservoirs, 2019, 26(2):77-81.

模型	优势	劣势
传统解析模型	遵从实际机理,结果易于解释,更容易用于指导开发设计; 解析模型对计算机算力要求低,数值模拟对已知信息使用充分; 用于新的目标区时不需要先验的认识	模型应用条件苛刻,实际应用中必须先获得模型中所有参数; 无法处理一些非结构化的数据,如：井的层位信息等
数据驱动模型	普遍适用性高,稳定性好,需要后期人工调整的部分较少; 基于模糊逻辑,对专业知识依赖少; 可以使用通用的数据特征处理方法,容易处理非结构化数据	基于统计学原理,不易解释机理; 需要很多先验的训练样本才能得到可靠的模型,并且模型不能推广到其他研究区
数据驱动建模工作要点	充分利用已知信息,结合专业知识进行数据特征工程; 根据已知数据的特点,建立对目标区最适用的、通用的预测模型,最终实现动态生产监测。样本中对其标签产生影响的、但未作为机器学习或统计学建模特征的若干属性要尽量一致,排除非确定性信息的影响