Abstract

Appraisal is a key step in consenting to develop an asset, or abandoning it, and is pursued after successful drilling of an exploration well in a potential field. During the appraisal process the drainage area and original hydrocarbons in place, as well as ultimate recovery (EUR) from the field are estimated which are often based on minimum set of information gathered during the exploration phase. This lack of data, along with uncertainties surrounding the appraisal data, introduces high degrees of variations in pre- and post- sanction EURs (EUR). These estimates, however, are revisited each time new data becomes available and as a result, the EUR from a field, along with several other factors, is subject to change over the field lifespan. Identifying the key drivers in accurate pre-sanction estimation of ultimate recovery and reducing post sanction EUR variance, helps in resource allocation and sustainable field development.

A major hurdle faced in subsurface characterization of assets is the degree of dependency between attributes and, the often non-linear behavior of these attributes. One way of overcoming these limitations is regression analysis; however, even in a high accuracy fit, regression coefficients by themselves are not necessarily good measures for ranking attributes, and elimination of lower ranked attributes would result in a new ranking of the remaining attributes. In the present study, several data mining techniques are applied on a dataset of 152 deep and ultra-deep water (D&UDW) fields in the Gulf of Mexico (GoM) to determine which of the 77 well-, reservoir- and field-scale attributes best capture the EUR variance for different fluid types in D&UDW fields in the GoM.

Unlike the conventional regression approaches, the present study offers a robust and stable ranking of attributes with high accuracy fit, where low to none contributing (poorly-predictive) attributes can be safely removed without changing the overall ranking of higher attributes. This ensures that a high ranked attribute is indeed a major contributor to accurate estimation of the ultimate recovery from a field, and therefore is worth the investment for capturing its value; on the other hand, a low ranked attribute, in all likelihood, is a redundant attribute and should not be collected; this would in turn free up resources that can be allocated to acquisition of high(er) ranking attributes.

Results of this study identify attributes that are strong overall drivers in over/under - estimation of reserves in pre- and post- sanction stages. We have also ranked the key attributes to reliable EUR estimations, which should be acquired prior to commitment to sanction. In addition, a set of attributes that have been consistently ranked as poor predictors are identified, which can be safely eliminated from data acquisition without affecting appraisal accuracy. Since the database tested was substantial covering all D&UDW fields in GoM, the identified key drivers have broad coverage and application.

This content is only available via PDF.
You do not currently have access to this content.