Birol Demiral

Genetic algorithm for estimating multiphase flow functions from unsteady-state displacement experiments

Abstract Relative permeability and capillary pressure are the primary flow parameters required to model multiphase flow in porous media. Frequently, these properties are estimated on the basis of unsteady state laboratory displacement experiments. Interpretation of the flood process to obtain relative permeability data is performed by one of two means: application of frontal advance theory or direct computer simulation. Application of frontal advance theory requires a number of experimental restrictions such that the pressure drop across the core is sufficiently large that capillary effects, particularly at the outlet end of the core, are negligible. A parameter estimation technique overcomes significant limitations of the classic calculation procedure. In this approach, functional representations or point values are chosen for the relative permeability curves. Adjustable parameters are then picked to minimize a least squares objective function. Previous applications of this approach have used Gauss–Newtons method with or without Marquardts modification. More recently, a simulated annealing method was also utilized. In this study we propose an interpretation method using recently developed genetic algorithms. The advantage and convenience of a genetic algorithm is that the method converges in all situations to a global optimum unlike Gauss–Newton methods, and it is as fast as the simulated annealing method. The performance of the algorithm is demonstrated with data from hypothetical and laboratory coreflood-displacement experiments where a computerized tomography scanner is utilized. It has been determined that the performance of the algorithm depends on the probabilities of crossover and mutation, and the proper usage of the fitness function.

Near Critical Gas Condensate Relative Permeability of Carbonates

Typical gas condensate fields contain a gas/liquid system during depletion. Such systems are difficult to model experimentally because they exhibit near-miscible behavior at high pressure and temperature. One way to simplify labora- tory experimentation is to use a binary retrograde condensate fluid and to adjust temperature to control miscibility. A se- ries of relative permeability test were conducted on a moderate-permeability carbonate core using methanol/n-hexane at near miscible conditions in the presence of immobile water. Potassium carbonate was added to the water to prevent misci- bility with methanol. The experiments used a pseudo-steady-state technique under conditions similar to the near well re- gion of a carbonate gas-condensate reservoir. The flow of gas and condensate at different force ratios was investigated. Relative permeabilities were obtained by matching historical production and pressure data using a coreflood simulator. It was observed that relative permeability depended on fluid composition and flow rate as well as initial condensate and wa- ter saturations. As the wetting phase (condensate) flow rate increased or interfacial tension decreased, relative permeabil- ity versus wetting phase saturation curves shifted towards lower wetting phase saturations. It was found that a simple three-parameter mathematical model that depends on a new dimensionless number called condensate number successfully modeled the gas-condensate relative permeability data. The developed model resulted in a good agreement with published gas-condensate relative permeability data as well as end point relative permeabilities and saturations. The relative perme- ability behavior as a function of IFT highly resembles the one observed in sandstones.

Pollution of an Aquifer by Produced Oil Field Water

Brine is produced from reservoirs as a waste material from crude oil and gas after processing. Waste water may be discharged at the surface or reinjected underground. When it is reinjected, it may be mixed with an underground fresh water source for several reasons. From this point of view, forecasting the pollutant concentrations by knowing the historical data at several locations on a field has great importance when planning the necessary precautions for environmental safety. Aquifer-M in Turkey, having the properties of potable water, is contaminated by oil field water that is injected for disposal purposes. A numerical model is used to determine the extent of pollution due to the injection of saline produced water into aquifer-M. Eight observation wells are drilled to take water and core samples in order to identify both rock and fluid properties of aquifer-M. Water samples taken from different intervals of aquifer-M are analyzed to determine the flow paths for the pollutant movement. The results are interpreted with the help of core property data obtained by computerized tomography (CT) analysis and routine core analysis. By using drilling records, log data, and CT analysis results, two subunits in aquifer-M with different lithological properties are identified. All data are used in a ground water pollution model. Aquifer-M has fresh water with salinity of 5-10 ppm, and the chlorine concentration of injected waste water is approximately 3410 ppm. Since there is a significant difference between these concentrations, the chlorine ion is selected as the indicative ion for locating the pollution front. The model study indicated that the contaminated water has propagated 18.7 km from the site of injection.Brine is produced from reservoirs as a waste material from crude oil and gas after processing. Waste water may be discharged at the surface or reinjected underground. When it is reinjected, it may be mixed with an underground fresh water source for several reasons. From this point of view, forecasting the pollutant concentrations by knowing the historical data at several locations on a field has great importance when planning the necessary precautions for environmental safety. Aquifer-M in Turkey, having the properties of potable water, is contaminated by oil field water that is injected for disposal purposes. A numerical model is used to determine the extent of pollution due to the injection of saline produced water into aquifer-M. Eight observation wells are drilled to take water and core samples in order to identify both rock and fluid properties of aquifer-M. Water samples taken from different intervals of aquifer-M are analyzed to determine the flow paths for the pollutant movement. The results are i...

Determination of Two and Three Phase Relative Permeabilty Values by Using a Pore Network Model

In this study, a Pore Network (PN) tool was developed in order to obtain two and three phase relative permeability values. It has been found that the observed results agree with published data. It has also been found that as contact angle increased, irreducible water saturation decreased, and oil relative permeability increased for two phase oil-water imbibition simulations. During the three phase displacement simulation by pore network simulation it was observed that water and gas relative permeabilities are only dependent on their own saturations, but oil relative permeability is dependent on its initial and own saturations in addition to water and gas saturations. It was also observed that as contact angle increased, the three phase region enlarged, and as relative permeability of oil increased, the oil isoperms became linear from concave towards the 100% oil apex.

WAX DEPOSITION IN HORIZONTAL WELLS

ABSTRACT In this research, wax deposition in horizontal well bores was studied. In a horizontal well, the main mechanism for wax appearance is the pressure change rather than the temperature change. Six different thermodynamic models were improved for horizontal wells by adding a pressure function, and a computer program was developed to simulate the precipitation phenomena in horizontal wells. Solid-liquid equilibrium constants for each pseudo-component were calculated. Critical pressures and overall solid mole fractions of the crude oil at different error ranges are determined. In addition to these, the changes in productivity index due to wax deposition for steady state and pseudo-steady state conditions were examined by the addition of a “wax factor”. It was observed that pressure change was slightly effective on wax appearance in horizontal wells.

Determination of three-phase relative permeabilty values by using an artificial neural network model

In this study, an artificial neural network (ANN) tool, which uses the data obtained from a pore network (PN) model, was developed in order to obtain three-phase relative permeability values. During the development of this ANN tool, four different stages were implemented in which ANN structures were changed in order to find the best architecture that would predict the oil isoperms correctly. By using the data obtained from the PN model, training was implemented and the prediction power of that tool was tested. When the data obtained from PN and ANN tools were compared, it has been found that irrelevant variables affected the ANN model negatively as decreasing its ability to learn perfectly. Finally, it has been observed that trends of the isoperms were effectively predicted and the overall quality of predictions was improved by changing the ANN structure.

Reservoir Description and Development of a Mature Oil Field

The Mishovdag oil field is located in the southwest of Baku, Azarbaijan. The sandstone reservoirs consisting of five middle Pliocene age Horizons I, II, III, IV, and XII provide 40% of total oil production from the Sirvan oil field region. The reservoir trap is an anticline, and its size is approximately 15 ́ 5 km. Since its discovery in 1956, 516 wells had been drilled and 198 of them are still producing from successive layers of sandstone formations. This study was conducted to describe Horizon I of Block-9, prepare input data for a modeling study, and suggest development scenarios for this block. From this point of view, it was aimed to properly describe the reservoir properties with the use of core and, mainly, well log data. In this respect, these data set were evaluated to define the reservoir. According to field reports, seven producing layers were present in Horizon I of Block9. From the results of further analysis on well logs, it was recognized that the reported seven layers were not continuous within Block-9 so, for modeling studies, these sandstone layers could be grouped under three main sand layers, namely, S1, S2, and S3, that were separated by two clay zones. The results of the modeling study showed that oil production was mainly from level S3 and level S1 was less swept by water injection. The oil saturation distribution at three levels at the end of 39 years of production indicated that there was still recoverable oil in levels S1 and S2. No free gas could be observed in any of the levels because the pressure maintenance provided by water injection caused free gas to redissolve in oil.The Mishovdag oil field is located in the southwest of Baku, Azarbaijan. The sandstone reservoirs consisting of five middle Pliocene age Horizons I, II, III, IV, and XII provide 40% of total oil production from the Sirvan oil field region. The reservoir trap is an anticline, and its size is approximately 15 × 5 km. Since its discovery in 1956, 516 wells had been drilled and 198 of them are still producing from successive layers of sandstone formations. This study was conducted to describe Horizon I of Block-9, prepare input data for a modeling study and suggest development scenarios for this block. From this point of view, it was aimed to properly describe the reservoir properties with the use of core and, mainly, well log data. In this respect, these data set were evaluated to define the reservoir. According to field reports, seven producing layers were present in Horizon I of Block-9. From the results of further analysis on well logs, it was recognized that the reported seven layers were not continuous within Block-9 so, for modeling studies, these sandstone layers could be grouped under three main sand layers, namely, S1, S2, and S3, that were separated by two clay zones. The results of the modeling study showed that oil production was mainly from level S3 and level Sl was less swept by water injection. The oil saturation distribution at three levels at the end of 39 years of production indicated that there was still recoverable oil in levels S1 and S2. No free gas could be observed in any of the levels because the pressure maintenance provided by water injection caused free gas to redissolve in oil.