A Methodology for Identifying and Quantifying Uncertainties (Seismic to First Phase Field Development) in Reservoir Simulation: Case Study of Two Fields in Saudi Arabia

Moriwawon, Babatunde (Saudi Aramco) | Al-Turki, Ali (Saudi Aramco) | Al-Shawaf, Ali (Saudi Aramco) | Radhey, Bansal (Saudi Aramco)

OnePetro 

Abstract

Understanding and management of subsurface uncertainties has become increasingly important for oil and gas companies to optimize reserve portfolios, make better field development decisions and improve day-to-day technical operations such as well planning. In this paper, the authors developed models, workflows and methodologies for multiple development scenarios that span the range of uncertainties for all stages of reservoir description and simulation for two fields (Field A and Field B) in Saudi Arabia. These processes are based on responses to uncertainty parameters in order to capture and quantify uncertainties. Uncertainty quantification for the two reservoirs was performed using experimental design (ED) approach taking into account seismic interpretation, geological uncertainty, and dynamic data. ED is an intelligent way of sampling parameter space by selecting combination of predefined variables to minimize the number of “experiments” to characterize the behaviour of a system, and to limit the number of models to be run. As a result, robust and responsive reservoir models were constructed to predict the field performance as well as to evaluate various future development strategies alongside better risk-assessments.

Introduction

Understanding and management of subsurface uncertainties has become increasingly important for oil and gas companies to optimize reserve portfolios, make better field development decisions and improve day-today technical operations such as well planning. Uncertainty refers to a parameter that has an unknown value. In other words, uncertainty is a lack of certainty. It is a state of having limited knowledge where it is impossible to exactly describe the existing state or value of specific parameter.

In reservoir studies, uncertainties exist at every step: geophysical study followed by structural modelling, geological modelling and dynamic flow simulations (Adepoju et al. 2009, Rivera et al. 2007). Geophysical uncertainties are related to acquisition & processing, interpretation (migration, picking, time-to-depth inversion and seismic-to-well tie), velocity law, etc. Geological uncertainties are coming from sedimentary concept, geological scheme, rock properties (heterogeneity, spatial variability, extension and orientation of sedimentary bodies, petrophysical characteristics, and phase contacts), etc. Number of uncertainties is related to the dynamic flow simulations that include fault transmissibility, extension of barriers, abs. permeability, relative permeability, ratio Kv/Kh, viscosity, PVT, Well PI etc. These uncertainties affect the ability to understand the reservoir behaviour and also make reliable production forecasts and risk-free decisions (Adepoju et al. 2009, Dejean and Blanc 1999, Rivera et al. 2007).

Therefore, it is crucial to develop models for multiple development scenarios that span the range of uncertainties for all steps of reservoir description and simulation. The method that is able to do that is a process based response to uncertainties. Uncertainty quantification for Field A and Field B was done using the experimental design approach.  Experimental Design is a “smart way” of sampling parameter space by selecting combinations of variables. It is an established method to minimize the number of “experiments” to characterize the behaviour of a system described by many parameters. The traditional statistical method requires modelling of all possible subsurface realizations and development scenarios. It is impractical due to the large number of models to be considered. For instance, 19 parameter case easily requires 3^19= 1,162,261,467 models. Experimental Design (ED) offers a technique to limit the number of models to be run by ‘intelligent selection’ of models that will represent the full uncertainty space (‘sampling’). Detailed workflow is presented below in Figure 1.