Data Mining Techniques For Optimizing Fast Track Re-engineering of Mature Fields

Abstract

Mature fields have a long production history file, and many other large data sets (static data, pressure, and so on). The challenge is often to select the most promising object (a field in a block, a reservoir in a multilayered field, or an area in a wide field). We present a global approach to solve this difficulty, that uses the concepts of Data Mining. It involves Sampling (gather the data, validate, and select the appropriate level of detail), Exploration (using a large set of graphical tools), Data Management (for production data, extracting production indicators), Modelisation (using for instance the hyperbolic oil production decline equation), and Aid to final decision. All along, graphical tools and multivariate statistical analysis are widely used and must be carefully designed. The whole of the approach requires a good geological and reservoir experience and a Reservoir and Geological Synthesis (RGS). Three field cases illustrate this fast track approach. Introduction Oil and gas companies have to face an ever increasing offer of fields for farm in, that have a long production history, because many fields were started in the 1960's, 80's and because the international market is more widely open. Companies portfolios include fields with a large number of wells — several hundreds or more — and several decades of production history. For these mature fields, extracting the most — or the best — of existing reserves is a must, for both the Companies and the countries with reserves. The remaining potential of these fields may be rather large. In some cases, it is necessary to solve all production-related difficulties that became apparent meanwhile the field was under production : insufficiently known heterogeneities, production mechanisms implemented with an unclear impact on production development, as for instance an insufficient reservoir pressure response to injection. In other cases, fields were developed with classical scenarios of natural depletion or water injection. The implementation of new techniques of EOR, or IOR, can sharply increase remaining reserves. Some papers have already been published about analytical evaluation methods for mature fields, that are not based on reservoir numerical simulations (Ref. 1 to 4). Recently, Albertoni and Lake (Ref. 5) have released a method to investigate well connections, on the basis of production rate fluctuations in five spot waterflood patterns. Multivariate linear regression method were applied. We now present a global methodology, based on our "Welfare" integrated software. Many types of data can be taken into account and subjected to statistical processing. Presentation of the method A noticeable part of today reservoir re-engineering activity is related, either to the selection of most promising zones, if one reservoir is concerned, or to most promising layers, when a multi-reservoir field is investigated. The challenge may even be to perform a fast track selection of a few fields among many. Each field has its own geological characteristics and its own development patterns and presents related difficulties in terms of performance, sectorisation, response to production mechanisms or location of untapped reserves. Petroleum engineers are facing a challenge which is comparable with those encountered in marketing, e.g. in the management of the relationship with the customer. Investing in the relation with a long-time customer is cheaper than acquiring new clients, and it can provide a lot of profit (Ref. 6). The most appropriate approach needs the personalisation of the relationship. So, the challenge is to extract the relevant information, concerning a small, but promising, part of customers in a market. Data Mining techniques are now currently used for this (Ref. 7).

Using Marketing Technology in Reservoir Engineering: The Application of Data Driven Predictive Analysis on a Mature Oil Field

“…DDPA has recently been developed and introduces marketing technology such as data mining (Coste, 2002) and Bayesian networks into the petroleum industry. Marketeers are used to handle huge amounts of data, which have been collected on customer bonus cards and which are used to find statistical correlations.…”

Zonal Allocation and Increased Production Opportunities Using Data Mining in Kern River

“…Neural networks have been thoroughly covered in numerous publications and will not be be repeated here 1,2,3,4,5,6,7,8 .…”