Paper Number: 100295-PA
DOI: 10.2118/100295-PA
Journal: SPE Reservoir Evaluation & Engineering
Volume: Volume 11, Number 6
Date: December  2008
Pages:pp. 991-999
Reducing Reservoir Prediction Uncertainty by Updating a Stochastic Model Using Seismic History Matching
利用地震史匹配校正储层随机模型以降低储层预测误差
Summary
摘要
We have developed a method in which spatial and dynamic information offered by time-lapse, or 4D, seismic surveys is used in history matching of reservoir simulations. Improved predictions of both recovery and areal sweep are then obtained by reducing uncertainty. Flow simulations are converted to predictions of seismic-impedance attributes using a petroelastic transform and suitable rescaling. The resulting misfit between the model and observed data is combined with an equivalent measure for well data, and these are used to constrain simulations by iteratively updating the model. Updated-model probabilities can then be used to analyze uncertainty.
我们发明了一种方法：利用时滞、4D和震测量提供的空间和动态信息进行储层模拟的历史匹配。通过降低误差我们最终能够更精确地预测采收率和面积驱油效果。通过岩石弹性变形和适当的比例转换把塑性变形模拟结果转换为地震阻抗属性的预测。模型和实测数据的结果不吻合再结合油井资料的等价测度，利用这些限制模型以反复校正模型。于是校正模型概率可以用来分析误差。
The method has been applied to the Schiehallion UK Continental Shelf (UKCS) reservoir. We first found a good match to production and seismic data in the field. From the updated probability distribution of the parameters, we then took the best models from the history-matching process and made predictions to determine the most likely outcomes.  这种方法已经应用与英国Schiehallion的大陆架储层之上。我们首先在油田内找到生产和地震数据匹配的较好的地区。利用更新过的参数概率分布，我们采用历史匹配过程中表现最好的模型再预测出最可能的结果。
We have found that the 4D data reduces uncertainty in predictions of the areal sweep and the pressure distribution. The seismic response is strongest at the injector wells but also helps in the interwell regions. Conventional history matching often struggles to constrain parameters in these regions because of the inherent nonuniqueness of the problem. The uncertainty of permeability- and fault-transmissibility multipliers was also determined in those areas.
我们发现4D数据可以降低在预测面积驱油和压力分布中的误差。地震响应在注水井中反应最强烈，而在井间地区也有一些响应。常规的历史匹配往往由于这种问题固有的不唯一性而尽量限制井间区域的参数。我们最终同样测定了该区渗透率和断层传导率乘数的误差。
附上前言：
Introduction
Reservoir management maybe improved if the present state of the field is known and if changes may bepredicted. The former requires information about current fluid-sweep andpressure changes, while the latter requires accurate reservoir description anda predictive tool, such as a simulation model. With this information, importantdecisions can then be made including facility maintenance and welloptimization, but more importantly, unswept areas can be identified and newwells drilled.
如果知道油田当前的状态并且能够预测油田状态的可能的改变情况将会有利于油藏管理工作的提高。前者需要知道当前流体区域和压力变化信息，而后者需要知道精确储层描述和一种预测工具，如一个模拟模型。有了这些信息，才能做诸如设备维护、油井优选以及更重要的确定流体未波及区和新井开钻等重要决定。
Conventionally, simulationmodels have been used to estimate the possible reservoir state and predictbehavior. The modeling commonly begins with the geologist who creates a numberof static geomodels, often constrained to log and core data from wells inaddition to preproduction 2D- or 3D-seismic data. The models may be upscaled bythe geologist and then modified by an engineer so that they match static anddynamic well data, including fluid-production rates and pressures [e.g.,Ertekin et al. (2001)]. Conventionally, this history-matching approach involveschanging model parameters manually, though automated methods are increasinglybeing used. Nonuniqueness of the models can hamper the process because ofmissing information, particularly regarding changes in the fluid distributionsand pressures between wells.
模拟模型以前常被用来估计储层可能的状态和预测储层行为。通常是地质师先创建几个静态地质模型，这些模型常常只限于利用油井的测井和岩心数据加之生产前的2D或3D数据。这些模型可能高度的取决于地质师，而后工程师利用流体生产速率和压力等动态油井数据配合静态数据修改模型。虽然现在也使用自动修改模型参数的方法，但这种历史匹配方法常常是人工修改模型参数。由于缺少信息，尤其是井间流体分布和压力信息，模型的不唯一性会强烈影响历史匹配方法的准确性。
4D-seismic surveys canprovide this information and reduce the nonuniqueness. Repeated 3D-seismic surveyscan be compared to identify changes in fluid saturation and/or pressures andthis is performed qualitatively, and almost routinely, in a number of North Seaand Gulf of Mexico fields [e.g., variouspapers in Parker et al. (2003)]. The goal for many geoscientists and engineersis to integrate this data so that it may be used quantitatively to constrainsimulation models and improve predictive capability [e.g., Gosselin et al.(2003) and Mezghani et al. (2004)].
4D地震测量能够提供这些信息而降低不唯一性。重复3D地震测量可以用来对比流体饱和度与压力的变化，在北海和墨西哥的一些油田，这种操作是定性的也是常规性的。地球科学家和工程师的目标是综合利用这些数据，进而定性地限制模拟模型从而提高预测能力。
To obtain such improvements, we have developed anautomated-history-matching method that includes 4D seismic data along withproduction data, on the basis of an integrated workflow (Fig. 1). The methoduses a quasiglobal stochastic method for choosing new models on the basis ofcalculated misfits between observed and predicted data. We have applied ourmethod to the Schiehallion UKCS turbidite reservoir, in which 4D seismic datahave shown great promise (Chapin et al. 2000; Parr et al. 2000; Saxby et al.2001). We updated the operator’s model and obtained an improved match toseismic data while retaining the good match to production data already present.Finally, the uncertainty of the parameters and predicted behavior was analyzed.
为了达到这个目的，我们发明了一种在综合工作流的基础上利用4D地震数据和生产数据的自动历史模拟方法。这种方法利用准全局随机方法选取新模型，这些模型都是在计算实测和预测数据不吻合度的基础上得到的。我们已经将该方法应用于UKCS Schiehallion 的浊积岩油田中，在这里4D地震数据显示出了较高的可信度。我们校正了操作者的模型，并且获得了在与现存生产数据较好匹配，同时又与地震数据较匹配的改进结果。最终分析了参数误差和预计行为。
[ 此贴被hougq在2009-04-28 17:43重新编辑 ]
