蒋文超
JIANG Wenchao

• 1:国家能源陆相砂岩老油田持续开采研发中心
• 2:中国石油大庆油田有限责任公司勘探开发研究院

摘要(Abstract)：

针对大庆油田油井压裂效果预测方法不确定性大、精度低的问题，利用机器学习与模型融合技术建立了油井压裂效果预测模型。采用统计学方法对SN区块油井压裂效果与地质、工程、生产参数的相关性进行了分析，借助基于LightGBM模型的封装法分析了油井压裂效果影响因素的重要程度，并进行了特征选择；采用支持向量机、神经网络、随机森林和LightGBM4种算法对油井压裂效果进行了预测。在此基础上，利用算术平均、加权平均、堆叠3种方法对4个算法进行融合，得到了精度更高的预测模型，并应用该预测模型对SN区块水力压裂方案进行了设计与优化。结果表明：神经网络建立的模型比其他3种算法精度更高，模型决定系数R~2为0.603；融合后的模型预测精度高于每个算法的精度，其中堆叠法效果最好，其R2高达0.899。该数据驱动建模技术具有较强的易用性和可移植性，对压裂方案设计具有一定指导意义。
Aiming at the problem of much uncertainty and low accuracy in prediction of production well hydraulic fracturing in Daqing Oilfield, machine learning and model ensemble are used to establish a prediction model of production well hydraulic fracturing for Block SN. Statistical methods are used to analyze correlation of production well hydraulic fracturing with geological parameters, engineering parameters and production parameters. Wrapper method based on LightGBM is used to identify the factors influencing production well hydraulic fracturing effect, and features selection is carried out. Prediction of production well hydraulic fracturing is performed by using support vector machine, neural network, random forest and LightGBM. On this basis, arithmetical mean, weighting mean and stacking methods are used to ensemble the 4 algorithms to obtain prediction model with higher accuracy which is used in designing and optimization of hydraulic fracturing plan for Block SN. The results show that the model established by neural network has higher accuracy than other 3 algorithms, with determination coefficient R~2=0.603. The ensemble model has higher prediction accuracy than each model, among which stacking method is most accurate with R~2=0.899. This data-driven modeling technique has high usability and portability providing guidance for fracturing design.

关键词(KeyWords)： 大庆油田;水力压裂;统计分析;机器学习;模型融合
Daqing Oilfield;hydraulic fracturing;statistical analysis;machine learning;model ensemble

Abstract:

Keywords:

基金项目(Foundation): 国家科技重大专项“大庆长垣特高含水油田提高采收率示范工程”（2016ZX05054）

作者(Author): 蒋文超
JIANG Wenchao

参考文献(References)：

• [1]卫秀芬，刚晗.大庆油田压裂工艺技术创新发展与前景展望[J].石油规划设计，2012,23(5):1-6.WEI Xiufen, GANG Han. Review and outlook of the fracturing technique in Daqing Oilfield[J]. Petroleum Planning&Engineering,2012,23(5):1-6.
• [2]李贺.大庆长垣油田特高含水后期水驱控水提效试验区开发效果[J].大庆石油地质与开发，2021,40(4):94-100.LI He. Evaluation on development effects of water cut control and efficiency-improvement test areas at the late stage of ultra-high water cut in Daqing Placanticline Oilfield[J]. Petroleum Geology&Oilfield Development in Daqing, 2021, 40(4):94-100.
• [3]马玉娟.大庆长垣油田三类油层压裂驱油提高采收率技术及其应用[J].大庆石油地质与开发，2021,40(2):103-109.MA Yujuan. Application of fracturing-flooding EOR technique in TypeⅢoil reservoirs in Daqing Placanticline Oilfield[J].Petroleum Geology&Oilfield Development in Daqing, 2021, 40(2):103-109.
• [4]邓刚.压裂驱油试验邻井压窜风险预警方案及其应用[J].大庆石油地质与开发，2022,41(1):91-96.DENG Gang.Early warning scheme of pressure channeling risk in adjacent wells in fracturing displacement test and its application[J].Petroleum Geology&Oilfield Development in Daqing,2022,41(1):91-96.
• [5]王云超.三元复合驱工业应用跟踪评价及措施调整方法[J].大庆石油地质与开发，2020,39(1):107-113.WANG Yunchao.Tracking evaluation and measure adjusting method of ASP flooding industrial application[J].Petroleum Geology&Oilfield Development in Daqing,2020,39(1):107-113.
• [6]曲世元，姜汉桥，李俊键，等.复杂小断块油藏水驱辅助注气吞吐提高采收率研究[J].特种油气藏，2021,28(4):116-122.QU Shiyuan, JIANG Hanqiao, LI Junjian, et al. Study on oil recovery enhancement by gas-injection stimulation assisted by water flooding in complex reservoir with small fault block[J].Special Oil&Gas Reservoirs,2021,28(4):116-122.
• [7]王洪亮，穆龙新，时付更，等.基于循环神经网络的油田特高含水期产量预测方法[J].石油勘探与开发，2020,47(5):1009-1015.WANG Hongliang,MU Longxin,SHI Fugeng,et al.Production prediction at ultra-high water cut stage via recurrent neural network[J]. Petroleum Exploration and Development, 2020, 47(5):1009-1015.
• [8]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.JI Lei,LI Juhua,XIAO Jialin.Application of random forest algorithm in the multistage fracturing stimulation of shale gas field[J]. Petroleum Geology&Oilfield Development in Daqing,2020,39(6):168-174.
• [9]唐军，彭超.涪陵页岩气藏水平井初始产能预测方法[J].大庆石油地质与开发，2020,39(6):160-167.TANG Jun,PENG Chao. Predicting method of the initial productivity for the horizontal well in Fuling shale gas reservoirs[J].Petroleum Geology&Oilfield Development in Daqing,2020,39(6):160-167.
• [10]唐佰强，刘招君，孟庆涛，等.松辽盆地东南隆起区上白垩统青山口组油页岩有机碳含量预测及效果评价[J].大庆石油地质与开发，2021,40(6):124-132.TANG Baiqiang, LIU Zhaojun, MENG Qingtao, et al. Prediction and effect evaluation of organic carbon content of oil shale in Upper Cretaceous Qingshankou Formation in Southeast Uplift of Songliao Basin[J].Petroleum Geology&Oilfield Development in Daqing,2021,40(6):124-132.
• [11]匡立春，刘合，任义丽，等.人工智能在石油勘探开发领域的应用现状与发展趋势[J].石油勘探与开发，2021,48(1):1-11.KUANG Lichun, LIU He, REN Yili, et al. Application and development trend of artificial intelligence in petroleum exploration and development[J]. Petroleum Exploration and Development,2021,48(1):1-11.
• [12]程翊珊，李治平，许龙飞，等.预测油层无机积垢的BP神经网络方法[J].大庆石油地质与开发，2021,40(3):84-93.CHENG Yishan,LI Zhiping,XU Longfei,et al.BP neural network method for predicting the inorganic scaling in the reservoir[J]. Petroleum Geology&Oilfield Development in Daqing,2021,40(3):84-93.
• [13]张瑞，贾虎.基于多变量时间序列及向量自回归机器学习模型的水驱油藏产量预测方法[J].石油勘探与开发，2021,48(1):175-184.ZHANG Rui,JIA Hu. Production performance forecasting method based on multivariate time series and vector autoregressive machine learning model for waterflooding reservoirs[J]. Petroleum Exploration and Development,2021,48(1):175-184.
• [14]但松林，刘尚奇，罗艳艳，等.基于BP神经网络预测高含水层对SAGD开发效果的影响[J].大庆石油地质与开发，2019,38(2):73-80.DAN Songlin, LIU Shangqi, LUO Yanyan, et al. Predicted SAGD development effects by BP neural network for the high-watercut reservoir[J].Petroleum Geology&Oilfield Development in Daqing,2019,38(2):73-80.
• [15]贾德利，刘合，张吉群，等.大数据驱动下的老油田精细注水优化方法[J].石油勘探与开发，2020,47(3):629-636.JIA Deli,LIU He,ZHANG Jiqun,et al.Data-driven optimization for fine water injection in a mature oil field[J].Petroleum Exploration and Development,2020,47(3):629-636.
• [16] SHELLEY B, HARRIS P C. Data mining identifies production drivers in a complex high-temperature gas reservoir[J]. Spe Production&Operations,2009,24(1):74-80.
• [17] BOB S,STAN S.The use of artificial neural networks in completion stimulation and design[J]. Computers and Geosciences,2000,26(8):941-951.
• [18] WANG S H,CHEN S N.Insights to fracture stimulation design in unconventional reservoirs based on machine learning modeling[J].Journal of Petroleum Science and Engineering,2018,174:682-695.
• [19]马俊修，石胜男，陈进，等.基于机器学习的玛湖地区水平井压裂设计优化[J].深圳大学学报（理工版），2021,38(6):621-627.MA Junxiu, SHI Shengnan, CHEN Jin, et al. Optimization of horizontal well fracturing design in Mahu area based on machine learning[J]. Journal of Shenzhen University Science and Engineering,2021,38(6):621-627.
• [20]李雪晨，马新仿，肖凤朝，等.基于模糊综合评判的致密油储层压裂选井组合方法[J].大庆石油地质与开发，2022,41(2):147-156.LI Xuechen,MA Xinfang,XIAO Fengchao,et al.Combination method of fracturing well selection in tight oil reservoir based on fuzzy comprehensive evaluation[J]. Petroleum Geology&Oilfield Development in Daqing,2022,41(2):147-156.
• [21]檀朝东，贺甲元，周彤，等.基于PCA-BNN的页岩气压裂施工参数优化[J].西南石油大学学报（自然科学版），2020,42(6):56-62.TAN Chaodong,HE Jiayuan,ZHOU Tong,et al.A study on the optimization of fracturing operation parameters based on PCA-BNN[J].Journal of Southwest Petroleum University(Science&Technology Edition),2020,42(6):56-62.
• [22]李晨阳. ZT区块页岩气井产能影响因素分析及预测[D].西安：西安石油大学，2020.LI Chenyang.Analysis of affecting factors and productivity prediction of shale gas well in ZT Block[D].Xi’an:Xi’an Shiyou University,2020.
• [23]王永卓，王瑞，代旭，等.松辽盆地古龙页岩油水平井箱体开发设计方法[J].大庆石油地质与开发，2021,40(5):157-169.WANG Yongzhuo, WANG Rui, DAI Xu, et al. Compartment development design method of horizontal well for Gulong shale oil in Songliao Basin[J]. Petroleum Geology&Oilfield Development in Daqing,2021,40(5):157-169.
• [24]罗红文，李海涛，安树杰，等.致密气藏压裂水平井温度剖面影响因素分析[J].特种油气藏，2021,28(4):150-157.LUO Hongwen,LI Haitao,AN Shujie,et al.Analysis of influencing factors of temperature profile of fractured horizontal well in tight gas reservoirs[J]. Special Oil&Gas Reservoirs, 2021,28(4):150-157.
• [25]周济民，张海晨，王沫然.基于物理经验模型约束的机器学习方法在页岩油产量预测中的应用[J].应用数学和力学，2021,42(9):881-890.ZHOU Jimin,ZHANG Haichen,WANG Moran.Machine learning with physical empirical model constraints for prediction of shale oil production[J].Applied Mathematics and Mechanics,2021,42(9):881-890.
• [26] KE G L MENG Q,FINLEY T,et al. LightGBM:A highly efficient gradient boosting decision tree[C].New York:31st Conference on Neural Information Processing Systems,2017.
• [27] CORTES C,VAPNIK V.Support-vector networks[J].Machine Learning,1995,20(3):273-297.
• [28] LECUN Y, BENGIO Y, HINTON G. Deep learning[J].Nature,2015,521:436-444.
• [29] WOLPERT D H,MACREADY W G.No free lunch theorems for optimization[J]. IEEE Transactions on Evolutionary Computation,1997,1(1):67-82.
• [30] BREIMAN L.Random forests[J].Machine Learning,2001,45(1):5-32.
• [31] PEDREGOSA F, VAROQUAUX G, GRAMFORT A, et al.Scikit-learn:Machine learning in python[J]. Journal of Machine Learning Research,2011,12(10):2825-2830.
• [32] CHOLLET F. Keras[EB/OL].[2021-09-03]. https：//github.com/keras-team/keras.