Figure and Table Captions 

	Figure 1. Deep Gulf of Mexico index map showing lower slope fields with middle Miocene to Paleocene turbidite reservoirs.
	Figure 2. Schematic cross section showing location of emerging new Miocene-Paleocene plays to other Gulf of Mexico plays.
	Figure 3. Reservoir model well log examples of massive sheet sand (Tahiti [left]) and interbedded sand and shale (Great White [right]) reservoirs.
	Figure 4. Reservoir model type logs with depofacies interpretation and reservoir properties upscaled to geocellular dimensions.
	Figure 5. Dynamic model setup with injector and producing well.
	Figure 6. Porosity-permeability cross-plots.
	Figure 7. Cumulative probability function for primary and water flood cases.
	Figure 8. Key uncertain variables impacting water flood performance.
	Table 1. Uncertainty parameters.

 

Introduction and Problem Statement 

Many deep-water Gulf of Mexico (GoM) discoveries of the past five years are in water depths greater than 4000 feet and in older Tertiary reservoirs of middle Miocene to Paleocene age. Middle Miocene fields and discoveries include Atlantis, Tahiti, Neptune, K-2, Thunder Horse, and Shenzi. Eocene-Paleocene fields and discoveries include Great White, Trident, St Malo, Jack, and Cascade. Structural styles of these lower slope fields include compressional anticlines, turtle structures and sub-salt three-way dip closures against salt faces (Figures 1 and 2). Some of these reservoirs are highly compartmentalized by faulting. In this setting, rock compaction may be less important as a production drive mechanism, and aquifer support (possibly augmented by water flooding) assumes more significance. Porosity and permeability decrease is related to greater burial depth and compaction as well as temperature-related cementation.

 

Much of the production experience in the deep-water Gulf of Mexico is from upper Miocene through Pleistocene reservoirs. The characteristics observed in these reservoirs and fields are summarized as follows:

• Pay often consists of stacked reservoirs.

• Permeability, porosity, and oil properties are good, resulting in high flow rates.

• Reservoirs are overpressured.

• Rock compaction is a primary drive mechanism.

• Aquifer influx is also often present.

 

The above reservoir characteristics result in high primary recovery factors and only a few developments have included waterflooding ; e.g., Lobster and Petronius.

 

By contrast, older middle Miocene to Paleocene reservoirs are characterized by the following:

• Reservoirs are often at greater subsea depths: 20,000 to 30,000 ft.

• Reservoirs often have high pressure (>15,000 psi) and temperature (>180^oF).

• Turbidite deposition was in coalescing basin floor fans; i.e., sheet sands.

• Seismic imaging of subsalt reservoirs is often poor.

• Reservoirs are consolidated, cemented and have low rock compressibility.

• Increased diagenesis in sands with volcaniclastic components reduces compressibility.

• Paleogene reservoirs have lower porosity and permeability.

• Primary recovery factors are expected to be lower due to lower reservoir properties and less compressibility.

• Water injection may be necessary to increase reservoir recovery.

 

The requirement for water injection to supplement reservoir drive energy, to improve oil rate, and to maintain oil production rates is of primary consideration in development planning for the new, ultra-deep water discoveries. The objective of this study was to quantify the incremental oil recovery potential for a range of the reservoir properties observed in these new middle Miocene through Paleocene discoveries.

 

Probabilistic Modeling 

A parametric simulation study was performed using experimental design to calculate increment oil recovery due to water injection and to identify the influence of parameters on recovery factor. The experimental design workflow is summarized below:

• Define uncertainty parameters and ranges.

• Set-up the experimental design matrix.

• Run the simulation cases defined in the matrix.

• Perform a multivariate regression to develop a linear relationship between recovery factor and uncertainty parameters (called the “proxy” equation).

• Generate an “S-curve” for recovery factor using a proxy equation.

 

A total of eleven uncertainty parameters were used in the parametric study. The parameters and range of uncertainty for each are detailed in Table1. Both static and dynamic parameters were considered.

 

The geologic uncertainty parameters incorporated into the static models include: structural dip, faulting, facies, aquifer size, and reservoir parameters (absolute permeability and heterogeneity). Dynamic uncertainty parameters include: fluid properties, water injection variables (timing and injection rates), and relative permeability variables (residual oil saturation and endpoints). Two static models were constructed based on the stratigraphy and reservoir properties from a thick-bedded middle Miocene reservoir (e.g., Tahiti Field) and a thinner-bedded Paleocene (e.g., Great White Field, Figure 3). Geocellular and dynamic simulation models were built with 200 x 200 ft cells having a thickness of 5 ft. Simple depofacies consisting of sheet, distal sheet, channel and shale were populated, and reservoir properties were distributed in these depofacies. Upscaled depofacies and properties are compared to the wireline logs in Figure 4, and a cross section showing injector and producer well locations in the dynamic model is shown in Figure 5. Permeability distributions were generated for three different Dykstra- Parson’s coefficients; 0.27, 0.6, and 0.8. Porosity-permeability cross-plots are shown in Figure 6. Three different fluids were considered with GOR (API) of 1,800 scf/stb (35° API), 1,100 scf/stb (30°API), and 500 scf/stb (27° API).  

Experimental design matrices were generated for both primary and water flood scenarios, based on the eleven uncertainty parameters. Eighteen primary cases and twenty-seven water flood cases were run. Proxy equations for both primary and water flood oil recovery were generated from the simulation results.  

Cumulative probability functions, “S-curves,” of oil recovery for both primary and water flood were calculated from the proxy equations using Monte-Carlo simulation (Figure 7). P50 oil recovery is 30% for primary and 37% for water flood, yielding incremental recovery of 7% of OOIP. As expected, incremental recovery for water flood is larger when primary recovery is low and lower when primary recovery is high. It is important to focus on incremental oil recovery rather than absolute recovery factor due to the modeling of a single producer-injector well pair.  

The key parameters impacting water flood performance are depofacies and net-to-gross (representing thief zones or limited connectivity), and aquifer size (Figure 8). Secondary parameters impacting water flood performance are permeability, faulting, residual oil saturation, and relative permeability endpoints. The least important parameters are beddip, injection voidage-replacement-ratio, and PVT properties.