The Pains and Gains of Experimental Design and Response Surface Applications in Reservoir Simulation Studies

Abstract

With the dearth of easy oil in the industry, the importance of consistency in quantifying uncertainties and assessing their impact on investment decisions have become very crucial in management decisions. This has seen the stocks of both experimental design and response surface techniques in the E&P industry rise significantly as an alternative to the more traditional uncertainty analysis. Whilst there are papers describing experimental design workflows and the different methods of generating response surface models for reservoir simulation studies, there is also a growing need to share practical examples of the lessons learned in constructing experimental designs and using response surface models to interrogate the experimental design outcomes. After extensively applying these concepts for over 18 months in identifying the major sub-surface uncertainties, explaining observed production performance and in prescribing additional development options for fifteen reservoirs situated in four different fields that are at different stages of maturation, it has been possible to capture many useful lessons. These lessons will both strengthen the benefits and ease the pains of applying these concepts in reservoir simulation studies. Introduction The ability of a simulation model to satisfactorily explain the past reservoir performance underpins its reliability to predict the future reservoir performance. Unfortunately, simulation models are not unique and this undermines the credibility of forecasted results. To mitigate this impact on business decisions, the reservoir engineer conducts several equally probable simulations to capture the range of uncertainties. Left unchecked the number of simulation runs can easily become unmanageable. Experimental design, ED, and the associated response surface methodologies, RSM, offer a cost-effective and efficient way to assess the impact of uncertainties on business decisions. The method also helps to identify the major parameters that have the most influencing impact on the business decision. Since the first introduction of experimental design to the oil industry in the early 90's (Damsleth, Egeland, Larsen) reservoir engineers have developed and successfully applied several experimental design workflows to various reservoir engineering studies (Friedmann, White, Amudo, Graf, Salhi, …). A typical workflow features the following steps:Uncertainty FramingScreening parametersConstraining uncertainty rangesRisk Analysis Following a brief introduction of the reservoirs, the outline of this paper mirrors the above typical steps in an experimental design workflow. The paper presents the lessons and experiences distilled from the application of ED and RSM concepts to 15 detailed reservoir studies and at different points in their project maturity spectrum. Though this paper does not purport to have all the answers, it attempts to address the painful challenges and highlight the benefits of using the technology.

Advanced ML and AI Approaches for Proxy-based Optimization of CO2-Enhanced Oil Recovery in Heterogeneous Clastic Reservoirs

“…Furthermore, DoE helps an individual to obtain the most-likely scenario that achieves the optimal response through a process (Lazic , 2006). The DoE has been proven as an efficient tool in conducting uncertainty framing, screening parameters, Bayesian updating of parameters, and risk analysis (Amudo et al , 2009).…”

Guidelines for Uncertainty Assessment using Reservoir Simulation Models for Green- and Brown-field Situations

“…Vogel (1956), Chu (1990), Damsleth et al (1992), Landa and Guyaguler (2003), Friedmann et al (2003), King et al (2005), Al Salhi et al (2005), Amudo et al (2009), Adepoju et al (2009), Rai et al (2009), Zubarev (2009, Izgec et al (2010), Wolff (2010), Bang (2013), Sonde et al (2013) and Wei (2013) represent some early and recent examples of use of ED methods in the petroleum industry to tackle various types of problems.…”

Experimental Design Methodology for Reserves Quantification Based on Soft Computing Modelling

“…Over the past decade EDA methods have been used extensively to develop surrogate models for 3D reservoir modelling (Van Elk et al 2000;White et al 2001;Smith et al 2004;Cheong and Gupta 2005;Amudo 2009;Cebastiant and Osbon 2011). In practice when working with 3D reservoir models to generate surrogate models for the output response only key uncertainties are considered as major impacts.…”