Date of Graduation

2004

Document Type

Thesis

Degree Type

MS

Committee Chair

Jaime Toro

Abstract

While the advantages of Artificial Neural Networks (ANN) for reservoir characterization are widely known, this project expands such benefits by providing a method by which an ANN can be trained prior to its application to real data. First, a geological and 3D seismic interpretation of the lower Pennsylvanian Atoka-Morrow sequence in the Buffalo Valley Field, New Mexico was executed. Then, to design the ANN and to test its predictions, I generated well-logs and synthetic seismic models for seismic attributes extraction. In order to bridge the vertical resolution gap between well-logs (high resolution) and seismic data (low resolution), VSP data was added as an intermediate step to train the ANN. This results in more effective predictions of density and sonic velocity values of the interval of interest. A synthetic model based on well-log correlation, seismic interpretation, and regional information provided the data set for the ANN training. The synthetic modeling also guided the 3D seismic interpretation of the Atoka and Morrow Formations. The 3D seismic interpretation and well log characterization justified the subdivision of the Morrow Formation into three subunits. The lower part is characterized by heterogeneous and erratic sands embedded in shale. The middle part is generally a series of stacked lenses of fining upward of sands and interbedded shales. The upper part is constituted by a laterally variable section of sands with carbonate content interbedded with shale. The lower and middle Morrow are interpreted as having been deposited in a fluvial environment (in a meandering system flowing with a north-south trend). This fluvial system seems to have evolved primarily in a location that followed a “paleovalley” where the Barnett shale was present. This location also corresponds to where the Chester carbonates were structurally low. Additionally, the trend of the interpreted boundary of the fluvial system matches the trend of high cumulative gas production in the Buffalo Valley Field. Then, this meandering system gradually transformed into the deltaic environment of the upper Morrow. Thus, the Morrow represents a drop in the sea level with the transition from a marine environment of the Chester Formation into a fluvial to deltaic system. The Atoka Formation was divided into two sub-units. The lower unit is a package of sands with some carbonate content and interbedded shales that coarsen upward. This subunit was interpreted as having been deposited in a deltaic to shallow marine environment. The Upper Atoka is composed of sands, carbonates, and interbedded shales. This unit displays a coarsening upward parasequence, which at times is followed by a fining upward parasequence. This is interpreted as being a deltaic sequence that progressively became marine (with sands and carbonates deposited parallel to the paleo coastline), which was later terminated by a maximum transgressive event, producing the regional shale that defines the top of the Atoka and the beginning of the Strawn Formation. The 3D seismic data shows faulting associated to the Marathon-Ouachita Orogeny. Locally, the major positive structure nearby, the Pedernal Uplift with a N-S, slightly eastern structural trend which seemed to have control the structural setting of the Buffalo Valley Field. The major faults consist of two interpreted high-angle reverse faults dipping to the west with a preferential strike direction of between N5E and N15E. They are located near the east and west edges of the study area. These faults seem to have affected deep pre-Devonian or even older rocks in the section. There is not enough evidence to argue that these faults resulted from wrenching. A reverse character is more evident, while lateral motion is not easy to identify, except for the flower structure s, or pop ups, in the north part of the area, which would suggest some lateral motion. The timing of these reverse faults, dipping primarily to the west, is inferred to have started some time before the lower Pennsylvanian and could have continued or been reactivated later affecting up to Wolfcampian strata (Early Permian). Syndepositional faulting related to the western and eastern faults occurred during deposition of the Atoka and Morrow Formations, resulting in a thicker section in the thrown block This project offers a good example of interaction of synthetic modeling and actual interpretation of the data to improve the design, training, and validation of ANN when applied to reservoir characterization.

Download

Share

COinS