Value of Information from History Matching - How Much Information is Enough?

Abstract

Summary With the rapid increase in computing power over the past several decades, automatic or semi-automatic approaches to history matching (HM) have become viable replacements for the traditional manual HM approach. HM approaches now include robust and efficient numerical algorithms with the ability to account for geological and petrophysical uncertainties. Downhole rate and pressure data are commonly collected for the purpose of uncertainty reduction through the HM process. Although the cost required to collect such data, and conduct the HM, is significant, few companies conduct an a priori analysis of the information value from the data. Although some studies have demonstrated the post-hoc value of HM data, few have demonstrated its a priori value; i.e., the assessment required to determine whether it is worthwhile investing in gathering the data and conducting the HM. In this paper, we illustrate and discuss an a priori analysis on information valuation, known as the Value-of-Information (VOI) analysis. The VOI from HM is assessed for future production data with the goal of informing the decision-maker of the potential value of investing in downhole measuring devices and HM procedures. We present the scientific basis for VOI analysis followed by an example of its implementation for an improved-oil-recovery (IOR) case. In the example, we use our proposed workflow of assessing VOI from HM to calculate the VOI from different types of production data and compare their values to distinguish between constructive and wasteful information gathering. The contributions of this paper are three-fold. Firstly, we provide a consistent definition of VOI from production data and HM, and discuss the details of the calculations. Secondly, we propose a workflow of assessing VOI from HM. Thirdly, we present an IOR example using our proposed workflow involving the use of Ensemble Kalman Filter (EnKF) combined with Robust Optimization (RO) to calculate the VOI. Finally, we identify and discuss the possible causes for the limited use of VOI methods in HM contexts and suggest ways to increase the use of this powerful analysis tool.