This paper presents a description of the process used to evaluate data from the Comodoro Rivadavia and Mina El Carmen zones in the El Tordillo Field in Argentina. The evaluation presented is holistic in nature and was performed by a team of experts. This evaluation consists of field-data quantification, integration, statistical and visual analysis, and development of a predictive artificial neural network (ANN) model capable of identifying sands with commercial hydrocarbon potential. The ANN model was used to identify patterns and trends related to the geology, reservoir, well-drilling issues, and swab-test production result. The purpose of this analysis was to identify sand attributes that are indicators of the sand's productive fluid type and capability to produce. This process was used to resolve difficult petrophysical interpretation issues associated with a complex sandstone reservoir system.

The derived ANN models were subjected to a blind test. These models were accurate 86% of the time at predicting oil/water ratios. The models were used to rapidly evaluate 149 sands in two new wells and have proven useful in the evaluation of new-well oil-production potential.


El Tordillo Field. The El Tordillo field was discovered in 1932. It covers an area of approximately 45.2 square miles (117 square kilometers). The field has three main productive formations: El Trebol, Comodoro Rivadavia, and Mina El Carmen. As of June 2002, the field had produced approximately 34 million m[3] oil, and 4.32 Bcf gas from these three formations.

As of June 2002, a total of 1,089 wells had been drilled in the El Tordillo field. Tecpetrol drilled 245 wells. Currently, 546 wells are in production and 1,226 workovers have been or are now being performed field-wide. The field also has 135 wells injecting 32 700 m[3]/d water to help increase oil production.

Challenges. The Comodoro Rivadavia and Mina El Carmen formations comprise a complex succession of sandstones and shales deposited during the Cretaceous Age. Gross thickness is in excess of 3,640 ft (1110 m) in the El Tordillo field. The formations can be divided into seven production markers. Deposited under different depositional environments, the sandstones of each marker have characteristic petrophysical attributes, but in general, exhibit diverse reservoir-rock qualities. The porosities and permeabilities vary greatly throughout the field.

Years of exploitation of the complex and layered reservoirs of the El Tordillo field have resulted in a number of undrained, low-permeability, or severely damaged zones with significant hydrocarbon reserves. The conventional petrophysical analysis using a conventional log alone is not sufficient to define the zone candidate to perforate (salinity of the formation water varies from sand to sand). To produce these candidate zones in a profitable way, identification of sands with production enhancement potential are required. In addition to these zones, near-depleted zones require extension of production life.

Objective of Analysis. The objective of the ANN effort was to analyze mud logging, pressures, geological correlations, oil gravity, swab tests, and electric logs to identify prospective zones. Based on this data we built multivariable analysis ANN models to determine those parameters that will identify, with a high degree of certainty, promising candidates for simple or multiple fracture treatments.

As part of this study, two ANN models have been developed for the purpose of identifying promising sand candidates. Each of these models is trained to predict a specific aspect of an actual swab-test result. The swab-test fluid-inflow (STFI) model uses drilling parameters and sand characteristics to predict whether a sand has the potential to produce fluid inflow into the wellbore. The swab-test water-cut model (STWC) uses geologic information about the sand, along with log-derived characteristics to predict an expected percent water fraction that will be produced from a sand. The accuracy of these models was 67% and 86% respectively, in the blind test performed.

This content is only available via PDF.
You can access this article if you purchase or spend a download.