基于中东地区某陆上碳酸盐岩油藏8年的开发数据以及超过37口井的试井和测井数据,经数据收集和准备、模型建立、模型训练和验证、模型应用4个主要步骤,开发了该油藏的自顶向下模型（TDM）,并利用该模型进行了产量预测和敏感性分析。该TDM包含5个相互连接的数据驱动人工神经网络模型,每个神经网络对一个关键的动态参数进行建模,一个模型的输出是下一个模型的输入。该TDM历史拟合效果较好,在时间和空间上均得到了验证,TDM应用于新数据时具有泛化能力并且可以准确预测3个月内的油藏动态。使用经过历史拟合和验证的TDM进行产量预测,结果表明在给定的操作条件下,随着时间的延续,该油藏产油量下降而产水量增加;通过改变水气交替注入的注入量、注入周期预测产量,结果表明,该油田提高注入量并不一定会使产油量增加,不同注入方案下注入周期为3个月的产油量比注入周期为6个月时更高。TDM为优化水气交替注入参数提供了一种快速而可靠的办法,同时能够优化加密井的位置及其深度。图11参10
YOUSEF Alklih Mohamad
,
KAVOUSI Ghahfarokhi Payam
,
ALNUAIMI Marwan
,
ALATRACH Yara
. 预测数据分析法在中东某老油田提高采收率中的应用[J]. 石油勘探与开发, 2020
, 47(2)
: 366
-371
.
DOI: 10.11698/PED.2020.02.15
Top-Down Modeling (TDM) was developed through four main steps of data gathering and preparation, model build-up, model training and validation, and model prediction, based on more than 8 years of development and production/injection data and well tests and log data from more than 37 wells in a carbonate reservoir of onshore Middle-East. The model was used for production prediction and sensitivity analysis. The TDM involves 5 inter-connected data-driven models, and the output of one model is input for the next model. The developed TDM history matched the blind dataset with a high accuracy, it was validated spatially and applied on a temporal blind test, the results show that the developed TDM is capable of generalization when applied to new dataset and can accurately predict reservoir performance for 3 months in future. Production forecasting by the validated history matched TDM model suggest that the water production increases while oil production decreases under the given operation condition. The injection analysis of the history matched model is also examined by varying injection amounts and injection period for water and gas (WAG) process. Results reveal that higher injection volume does not necessarily translate to higher oil production in this field. Moreover, we show that a WAG process with 3 months period would result in higher oil production and lower water production and gas production than a 6 months process. The developed TDM provides a fast and robust alternative to WAG parameters, and optimizes infill well location and its corresponding true vertical depth (TVD).
[1] ASKARI F R, KHAMEHCHI E.A novel approach to assist history matching using artificial intelligence[J]. Chemical Engineering Communications, 2015, 202(4): 513-519.
[2] MOHAGHEGH S D.Reservoir simulation and modeling based on artificial intelligence and data mining (AI&DM)[J]. Journal of Natural Gas Science and Engineering, 2011, 3(6): 697-705.
[3] MOHAGHEGH S D, GRUIC O, ZARGARI S, et al.Top-down, intelligent reservoir modelling of oil and gas producing shale reservoirs: Case studies[J]. International Journal of Oil Gas and Coal Technology, 2012, 5(1): 3-28.
[4] NAGEH M, EL E, MAHMOUD A, et al.Application of using fuzzy logic as an artificial intelligence technique in the screening criteria of the EOR technologies[R]. SPE 175883, 2015.
[5] LEE J Y, SHIN H J, LIM J S.Selection and evaluation of enhanced oil recovery method using artificial neural network[J]. Geosystem Engineering, 2011, 14(4): 157-164.
[6] BELLO O, HOLZMANN J, YAQOOB T, et al.Application of artificial intelligence methods in drilling system design and operations: A review of the state of the art[J]. Journal of Artificial Intelligence and Soft Computing Research, 2015, 5(2): 121-139.
[7] KHAZAENI Y, MOHAGHEGH S D.Intelligent time-successive production modeling[R]. SPE 132643, 2010.
[8] ZARGARI S, MOHAGHEGH S D.Field development strategies for Bakken shale formation[R]. SPE 139032, 2010.
[9] SHAHAB D M, OGNJEN S G, SAEED Z, et al.Modeling, history matching, forecasting and analysis of shale reservoirs performance using artificial intelligence[R]. SPE 143875, 2011.
[10] HAYKIN S S.Neural networks and learning machines[M]. Upper Saddle River, NJ: Pearson education, 2008.
