Developing and Applying Fluid Properties for the Wara Formation, Greater Burgan Field, Kuwait: A Supergiant Field With Variable API Oil Gravity

The Greater Burgan field located in southeastern Kuwait was discovered in 1938 but production from the field did not begin until 1946. The "Super Giant" field covers a surface area of about 320 square miles. Production from the Greater Burgan Field has been primarily from the middle Cretaceous clastic reservoirs of the Wara and Burgan (third & fourth sand) formations which are separated by the carbonate Maddud formation as shown in Fig. 1. The Burgan formation is the major producer in Greater Burgan with the Wara formation having produced approximately 10% of the field's total production.

The Greater Burgan field is comprised of four distinctive structural culminations. These structures are identified as the Ahmadi, Magwa, Burgan and NE Burgan areas. Recent work has recognized that each of these areas are characterized by different PVT properties which are mostly likely attributed to the original oil migration and fill-up history of the Greater Burgan field.

Past studies conducted by Kuwait Oil Company (KOC) in 1960 and 1995 did not incorporate API gravity data from non-PVT sources. API gravity Vs depth data available from core data, and initial production tests were added to the data sets used in previous studies. This more comprehensive data set was normalized and then curve fit with exponential equations. Four distinctive correlation trends for the Ahmadi, Magwa, NE Burgan and Burgan areas were developed from this work.

The aim of this paper is to highlight the variability of PVT properties that can occur both vertically and areally in large fields like Greater Burgan and their impact on simulation results and other analyses.

Past KOC PVT studies did an adequate job in capturing the API Vs depth correlations for the lighter oil ranges in each of the structural areas. However, PVT properties for the heavier oil located in the reservoirs flank positions were not captured. This was mainly due to the lack of PVT analyses for the heavy oils and the limited production from wells in downdip locations.

KOC's 1960 Study was completed using Wara API samples which where divided into six areas Main Burgan, N.E.Burgan, Burgan Graben, Magwa, Magwa Graben & Ahmadi. This study did not include samples from wells down flank that produce heavier (<24° API) oil. The 1960 correlation developed used 24°API as it's low gravity end member. The observed exponential curves are shown in Fig.2.

KOC's 1995 Study examined 123 PVT analyses from 103 wells. The reservoir fluid properties for the individual regions were then computed with a variable gravity option based on a linear interpolation of the properties "end member" samples for each PVT region as follows:

1. The Wara sand and 3^rd Sand from the Burgan Field.

2. The Wara sand from the Ahmadi and Magwa Fields.

3. The 3^rd sand from the Ahmadi and Magwa Fields.

4. The 4^th sand from the Burgan Field.

This study identified "end member" wells from which linear correlations of API gravity Vs Depth were constructed. Fig (3) shows the resulting correlations developed for the Greater Burgan field. As shown in Fig. 3 the Ahmadi and Magwa areas were represented with a single linear correlation while the Burgan area was represented by two linear intersecting functions. The 1995 study did not recognize any heavy oil less than 24°API gravity. The two area defined correlations were used to initialize a 1995 full field simulation model for the Wara formation.

Correctly characterizing the fluid properties of a reservoir model is very important to developing a successful flow simulation. In the previous two KOC Wara simulation studies the heavier oil gravities (<24°) were not represented in the reservoir model.

The failure to adequately model the API variability with depth will have a large effect on the reservoirs oil mobility down flank, formation volume factor (Bo), and solution GOR. These variables can greatly impact the results of a simulation especially when dealing with reservoirs as large as those in the Greater Burgan field. To be able to accurately capture the initial fluid properties in and around the OWC in each area a more comprehensive evaluation of PVT properties was needed. It is important that the variation of API properties are not only identified but also captured in the initialization of the simulation fluid properties.

Normalization of PVT data to the OWC and Gas cap in each area is not normally done. For the Greater Burgan field was found that normalizing this data to the distance between the original gas cap and the oil-water contact helped to:

• Identify the need for different API regions.

• Allowed us to merge data sets in the Ahmadi and Magwa areas to improve data density over the full range of oil gravities.

• Generate better curve fit correlations.

The data was normalized using the following equation:

Normalized Data = Height above OWC / (Gas Cap Elevation - OWC Elevation)

The original oil-water-contact (OWC) elevation for all areas of the Greater Burgan field is 4471' s.s. Table (1) shows the GOC depth assumed for all Burgan field areas.

The normalized data was used to develop PVT correlations for all Burgan field areas. Based on review of the normalized data the following 3 regions were identified to have unique PVT properties:

1. Ahmadi and Magwa Areas

2. NE Burgan Area

3. Burgan Area

Fig.4 shows the normalized depth Vs API gravity for Burgan area. Two exponential curves were used to best fit the distribution of the data. The lower curve was used for API gravities less than 30°API and the upper curve was used for the higher API Burgan data. A comparison of the new curves to the 1995 and 1960 correlation, highlights the improved correlation fit to the full range of data. The improve fit can be explained by the following reasons:

• The 1995 analysis did not include non-PVT API data.

• The 1995 and 1960 correlations used 24°API oil as the low gravity end member.

• The 1995 correlation did not honor the curvilinear nature of the API data observed in the 1960 and most recent PVT analyses.

The same procedures were applied for the Magwa, Ahmadi and Northeast Burgan areas. Based on initial review of normalized data sets we determined that the Magwa and Ahmadi data behaved as a single correlation set. Because the Magwa data set contained more heavy oil samples combining the data allowed us to generate a better correlation of the heavy oil regions in both areas. As shown in Fig. 5 the new correlation shows values down to 10° API gravity at the OWC while previous correlations calculated values no lower than 24° API near the OWC.

Fig.6 shows the 3 transforms that were generated for the Ahmadi/Magwa, N.E.Burgan, and Burgan areas and input into the simulation model. Notice that the Burgan curve has lower API values for the upper region of data than the MG/AH area. This is expected since the MG/AH areas have original gas caps. We believe that the Magwa /Ahmadi structures were first filled to spill point followed then by migration of hydrocarbons to the Burgan structure. The gravity segregation in Magwa /Ahmadi and NE Burgan before hydrocarbon migration of the oils at spill point to the Burgan area is the reason for heavier oil in this area even though it is structurally higher (see Fig. 7).

A comparison of the curves with data shows a fair degree of scatter with an excellent match. Data points that did not fit the curves were not investigated. It is thought that most of the scatter is a function of faulting, measurement precision, and production history.

Using the exponential equations developed from the normalized correlations, depth Vs API gravity charts were generated for Burgan, Magwa, Ahmadi and N.E.Burgan areas separately. A comparison between 1995 correlation and the new correlation curves is shown in Fig.8. Note that the MG/AH curve has higher API gravity than the Burgan curve.

The 1995 study used functional forms of existing correlations (Standing and Beal & Beggs Robinson equations) with optimized coefficients to compute the reservoir properties from just a few variables. The desired properties from this work are the bubble point pressure, gas-oil ratio, formation volume factor, oil and gas compressibility, and oil and gas viscosity. The key variables identified for correlation purpose were the differential liberation gas-oil ratio, Rs, the residual oil gravity and the reservoir temperature. The gas-oil ratio from differentia! liberation is a function of the residual oil gravity as shown in Fig.9. By using the new API correlation with the 1995 equations, we could generate a spreadsheet to calculate the initial API gravity if depth and reservoir pressure were given. It also can calculate the viscosity, bubble point pressure, Rs and gas-oil ratio if API gravity, depth and reservoir pressure are identified.

Initial Rs estimations are determined by the Rs Vs API correlation see Fig.9 (This correlation was completed using the initial Rs and API data). A reasonable linear correlation was obtained for Burgan and Magwa/ Ahmadi areas. The Rs correlation line for Magwa & Ahmadi goes to higher API value than the Burgan line because Magwa and Ahmadi arear have more gas due to their associated gas caps.

The API gravity change has a great impact on the simulation model initialization and OOIP estimates since the API gravity effects viscosity and formation volume factor of the oil. Changes in viscosity effect key input parameters for the simulation model such as mobility ratio. This will apply a significant change in determination of the oil mobility near OWC and gas caps. For example, the mobility ratio for 0.8 of a relative oil permeability and a relative water permeability of 0.2 with oil and water viscosities of 6.6 and 0.55 cp respectively is 1/3. If the oil viscosity changes to 2.2 cp and every thing else stayed the same, the mobility ratio will be 1.0. Mobility ratio is probably the most important factor involved in determining sweep-out pattern efficiency. For instance, a 1/3 mobility ratio for five-spot well spacing will give a sweep-out pattern efficiency of 55% while 1.0 of mobility ratio for the same spot well spacing will apply 65% of the sweep-out pattern efficiency.

This API correlation was applied in build-up analysis and wellbore flow correlations. Because of the changes in viscosity resulting from the API correlation permeability estimates were changed by a factor of three. This can be shown by an example of a build-up test for BG-XX as shown in Fig. (10). Using a viscosity value of 0.8 cp will give a permeability estimate of ≈200 md while 2.2 cp of viscosity will show ≈ 600 md as the permeability estimate.

• The observed oil gravity data with depth has an exponential curve shape which is asymptotic near the gas/oil and water/oil contacts. This relationship was observed in each area of the Greater Burgan field. The entire field shares a common water/oil contact with each area having a unique gas/oil contact. An interesting observation is that the Burgan dome, which has the most structural relief, does not have a gas cap although it was originally at or very near the bubble point pressure. By normalizing the depth we were able to improve our depth correlation and show that the Magwa & Ahmadi data could be combined into a single normalized equation. Specific gravity changes were found to not be reservoir specific but are a function of normalized depth (see Fig.11).

• This variation of oil gravity with depth has long been known and was first published in 1960 for the Burgan field. This was prior to additional testing of oil near the oil water contact and thus early correlations did not see the rapid decrease in oil quality near the oil water contact. It is thought that this is a function of the filling history of the Burgan field and our correlations show that at the spill point of N.E.Burgan the oil gravity is the same as the oil gravity at the crest of the Burgan dome (37° API). By contrast the Magwa and Ahmadi oil gravity at the crest is 39° API gravity.

• Integrating different API sources improved the reliability of the API data correlation.

• Normalizing variable API data can improve PVT correlations.

This technique may be applicable on other large oil reservoirs with variable API.

Special thanks to the Kuwait Oil Ministry and the Kuwait Oil Company for permission to present this paper. KOC General Superintendent, Mr. Ali Al-Kandiri, and Superintendents, Ms. Badria Abdul-Raheem, and Mr. Mohammed Redha, for their support of the Wara Team project.