To reliably predict the reservoir performance, an accurate model of the reservoir is necessary. For reservoir simulation purposes, the flow unit model is most practical approach. The flow units are defined according to geological and petrophysical properties that influence the flow of fluids in the reservoir. Identification and prediction of flow units are strongly dependent on the availability of permeability distribution. This need for permeability distribution significantly limits the identification of flow units in reservoirs where permeability measurements are not abundant such as most reservoirs in the Appalachian Basin.
In this study, statistical and artificial intelligence techniques were employed to identify flow units based on limited data obtained from core analysis supplemented by mini-permeameter measurements, geological interpretations, and well log data in a heterogeneous oil reservoir in the Appalachian Basin. An innovative methodology was then developed to predict flow units using only well log data. The distribution of flow units in the reservoir was then predicted based on abundant well log data. Finally, permeability and porosity distributions were predicted based on the distribution of the flow units in the reservoir. This approach led to development of a reliable reservoir model. The accuracy of model was verified by successful simulation of the production performance. The methodology presented in this paper can serve as a new guideline for the characterization of heterogeneous reservoirs.
The Appalachian Basin has numerous abandoned or marginally productive oilfields. Significant amount of oil remains in these reservoirs however further development of these reservoirs, or similar reservoirs in other basins, is hampered by the lack of sufficient data to evaluate their potential and predict their performance. The key parameter for reservoir characterization is the permeability distribution. In reservoirs where permeability measurements are not abundant, permeability must be predicted from well log data. The goal of this study was to develop a methodology for reservoir characterization with limited permeability data. In order to achieve this goal, an oil reservoir in West Virginia was selected for study. The reservoir was discovered in 1895. Between 1895 and 1901 over 500 wells were drilled in this reservoir. By 1910 most of the wells were plugged and abandoned. In 1981, a pilot waterflood was commenced in the field. Based on the pilot recoveries; development proceeded to a full-scale waterflood. Over 140 new wells were drilled as injectors or producers. Secondary recovery of the field has been in progress since 1987. Core and core analyses were available from 6 wells which were drilled during the waterflood evaluation period. All of the wells drilled for water flooding operations had gamma ray and density logs available.
The information available form the core descriptions and core analyses from the six wells were used to define the Flow Units. Statistical and graphical techniques were employed to identify and verify the Flow Units based upon permeability-porosity relationship within each Flow Unit. Our previous investigations (Aminian, et al 2000 and 2001a) have revealed that Artificial Neural Networks (ANNs) to be very useful for predicting permeability using geophysical log data. This study extends this application by developing a methodology to identify Flow Units within the reservoir when permeability is not available. Artificial Neural Networks are then developed and utilized to predict the distribution of flow units, permeability, and porosity using well log data from 125 wells reservoir-wide. The production-injection data were available from several 5-spot patterns. This offered the unique opportunity to verify the reservoir description by history matching with a reservoir simulator.