(669f) Reservoir Uncertainty Reduction and Quantification: An Optimization Based Consensus Method Using Multiple Static Measures

It is well known that uncertainty exists in the oil reservoirs. The reservoir thickness and spatial distribution of shale, oil saturation, porosity and permeability are uncertain (not exactly known) parameters that must be quantified to support decision making in reservoir management. Flow simulation is an important step in reservoir development and operations. It takes reservoir geological property data as input to obtain the reservoir performance. Simulation results are further used to assist decision making in well placement and well production control. However, flow simulation is a computationally intensive task due to complex geological heterogeneities and numerical thermal modeling. For this reason, generally only a small number of flow simulations can be performed. Thus, the choice of reservoir uncertainty realizations becomes crucial in reservoir management.
Randomly selecting a small number of realizations tends to inaccurately represent the overall reservoir uncertainty. To avoid this problem, ranking methods have been widely used in the past to select realizations [1-3]. However, ranking methods highly depend on the specific ranking measure employed and may not correctly capture the reservoir uncertainty distribution. To improve the quality of uncertainty reduction and quantification, in this work we proposed a new consensus approach that uses multiple static measures to select reservoir uncertainty realizations. The proposed method quantifies the relationship between discrete distributions of realizations using probabilistic distance, which is calculated based on the geological property information and multiple static measures. A novel mixed integer optimization based technique is developed to select a representative subset of the realizations to ensure that a few realizations have close statistical characteristics in terms of response as the entire set of realizations. Those selected realizations can then be further used to perform the flow simulation and to optimize the well placement and well production. In this work, effectiveness of the proposed method is demonstrated through case studies in reservoir production optimization problem.

References
1. Deutsch, C.V., Srinivasan S. Improved reservoir management through ranking stochastic reservoir models. SPE/DOE Tenth Symposium on Improved Oil Recovery. 1996, 35411.
2. Derakhshan S.V., Deutsch C.V. Direct simulation of P10, P50 and P90 reservoir models. Canadian International Petroleum Conference / SPE Gas Technology Symposium Joint Conference, 2008, 188.
3. Fenik D.R., Nouri A., Deutsch C.V. Criteria for ranking realizations in the investigation of SAGD reservoir performance. Canadian International Petroleum Conference. 2009, 191.