fax 01-972-952-9435. AbstractDew point pressure is a crucial parameter for evaluation of gas condensate reservoirs. Hence, the availability and accuracy of this property has been the main bottleneck for the reservoir simulation and development. Although the dew point pressure can be determined experimentally from collected laboratory samples frequently these measurements are not available. In these cases, fluid reservoir properties are determined with the use of empirical correlations (explicit methods) or they can be determined iteratively using an equation of state (EOS). Theoretical results and practical experience indicate that feed forward network can approximate a wide class of function relationships very well. This property is exploited in modeling petroleum engineering process. In this work, a set of conventional feed forward multilayer neural network have been proposed to predict dew point pressure of gas condensate reservoirs. The accuracy of the method is evaluated by its application for dew point pressure estimation of various reservoir fluids not used in the development of the model. Furthermore, the performance of the model is compared against the performance of other alternatives correlations reported as the most accurate and general correlations for dew point pressure prediction. Results of this comparison show that the proposed method out performs the other alternatives, both in accuracy and generality. The network was developed using experimentally Constant Volume Depletion (CVD) measured condensate sample of south pars reservoir and collected data from literature of 111 gas condensate samples covering a wide rang of gas properties and reservoir temperature.The network has an average absolute error of 2.573%, 3.832% and 2.612% for training, validation and test processes, respectively.

Dew point pressure is a crucial parameter for evaluation of gas condensate reservoirs. Hence, the availability and accuracy of this property has been the main bottleneck for the reservoir simulation and development. Although the dew point pressure can be determined experimentally from collected laboratory samples frequently these measurements are not available. In these cases, fluid reservoir properties are determined with the use of empirical correlations (explicit methods) or they can be determined iteratively using an equation of state (EOS). Theoretical results and practical experience indicate that feed forward network can approximate a wide class of function relationships very well. This property is exploited in modeling petroleum engineering process. In this work, a set of conventional feed forward multilayer neural network have been proposed to predict dew point pressure of gas condensate reservoirs. The accuracy of the method is evaluated by its application for dew point pressure estimation of various reservoir fluids not used in the development of the model. Furthermore, the performance of the model is compared against the performance of other alternatives correlations reported as the most accurate and general correlations for dew point pressure prediction. Results of this comparison show that the proposed method out performs the other alternatives, both in accuracy and generality. The network was developed using experimentally Constant Volume Depletion (CVD) measured condensate sample of south pars reservoir and collected data from literature of 111 gas condensate samples covering a wide rang of gas properties and reservoir temperature. The network has an average absolute error of 2.573%, 3.832% and 2.612% for training, validation and test processes, respectively. 1. Introduction One of the crucial steps in gas reservoir evaluation and simulation is dew point pressure prediction of retrograde samples. Accurate reservoir fluid properties are essential for all reservoir-engineering calculations such as estimation of reserves and forecast and planning of future enhanced oil and gas production. Hence, the availability and accuracy of this property has been the main bottleneck for the gas reservoir simulation and development. Calculations show that even small error in estimation of PVT properties leads to a much higher error in reservoir's model design, operation and control. This small error results in inappropriate reservoir simulation and poor performance of the reservoir's model. In general, there are two classes of estimation methods for calculation of dew point pressure. The first class is determined experimentally from collected laboratory samples frequently these measurements are not available. The second class, fluid reservoir properties are determined with the use of empirical correlations (explicit methods) or they can be determined iteratively using an equation of state (EOS). Empirical correlations are relatively easy to use, but to date, they have not been able to duplicate reliably the temperature behavior of constant-composition fluids. An EOS, while duplicating the behavior of constant-composition fluids, may have convergence problems and must be calibrated to existing and available experimental data. The statistical approach is comparatively a more versatile approach to the problem of PVT parameters prediction. It makes use of the available experimental dew point data (the dependent variable) and develops functional relationships with the experimental fluid characterization and reservoir temperature data (the independent variables). It, however, requires the assumption and satisfaction of multi-normal behavior and linearity, and hence it must be applied with caution.

No abstract

The Third Sand Upper (3SU) is one of the three sub-reservoirs in the Third Sand of the Greater Burgan field, the world's largest sandstone oil field. Initial oil production begun in 1948 and 3SU field development has not been aggressive due to its poor reservoir quality and productivity. After 60 years of primary production, only 7.5% recovery has been achieved. Infill drilling was identified as a key development strategy in 3SU. In 2008, a simulation study was initiated to investigate infill drilling potential and its impact on production and recovery. We opted for a sector model mainly due to practicality and time constraint. The 780,800 cells sector geological model was scaled-up to 421,632 cells for flow simulation. Due to the sand-to-sand contact with the lower Burgan sands, it is imperative to include these reservoirs in the model to achieve proper energy balance. Accordingly, four pseudo layers were added to the simulation model to allow fluid migration from the lower reservoirs. The 3SU sector simulation model has 100m × 100m areal cells and individual layers with 4–6 feet thickness. Overall, the sector model has 30 times refinement compared to previous 3SU models (Ambastha et al, 2006). The history match has been carried out for 37 3SU historical wells with 60 years of production history. Detailed study of interactions among field permeability distribution, aquifer strength, fluid migration and fault transmissibility specifications on simulation results was key in developing meaningful history match. Water cut match was less than satisfactory for wells located in the dome area due to modeling deficiency introduced by the pseudo layers. Three infill drilling spacing scenarios were set up to evaluate prediction performance of 800-meter, 400-meter and 200-meter well spacing. Results of the 50-year prediction runs indicated that an incremental recovery of 11% can be achieved by reducing the current well spacing of 800-meter to 400-meter. Introduction Greater Burgan field, which is located in southeastern Kuwait, covers a surface area of about 320 square miles and has been ranked as the largest clastic oil field in the world. The four main reservoir units comprising the Greater Burgan Field complex are Wara, Mauddud, Burgan Third Sand (3S) and Burgan Fourth Sand (4S). The massive 3S is further subdivided into Third Sand Upper (3SU), Third Sand Middle (3SM) and Third Sand Lower (3SL). The 3SU reservoir is sandwiched by a tight Mauddud formation above and a permeable 3SM sand below. Figure 1 shows the corss-section of the major reservoir-horizons in the Greater Burgan field. 3SU reservoir communication occurs mainly through sand-to-sand contact with 3SM but extensive faulting also allows communication between Wara, Mauddud, 3S and 4S reservoirs. The Greater Burgan Field is separated into three producing areas, Burgan, Magwa and Ahmadi. No structural, geologic or reservoir features distinguish these areas, although PVT differences are assigned for areas north and south of the Graben fault. Figure 2 shows the areal view delineating these 3 areas. Initial 3SU production begun in early 1948. Despite of its significant STOOIP, 3SU has not been a dominant producer due to its inferior productivity. Overshadowed by the prolific 3SM reservoir, 3SU development has not been the priority and its potential was not fully assessed. In 2007, Kuwait Oil Company (KOC) has started revitalization of several low priority reservoirs in order to achieve the corporate production growth by 2020. In 3SU reservoir, two new wells were drilled in 2008 and 2009 to evaluate the performance of infill drilling. At the same time, a 3SU sector model was built to investigate the incremental recovery of infill drilling. This simulation effort was carried out by the KOC Greater Burgan Studies team with consulting assistance from Schlumberger.

The need to develop new tools that allow reservoir engineers to optimize reservoir performance is becoming more demanding by the day. One of the most challenging and influential problems facing reservoir engineers is well placement optimization. The North Kuwait field (NKF) consists of six fields containing four naturally fractured carbonate formations. The reservoirs are composed of relatively tight limestone and dolomite embedded with anhydrate and shale. The fields are divided into isolated compartments based on fault zones and supported by a combination of different fluid compositions, initial pressures, and estimated free-water levels. Due to natural complexity, tightness, and high drilling costs of wells in the NKF, it is very important to identify the sweet spots and the optimum well locations. This paper presents two intelligent methods that use dynamic numerical simulation model results and static reservoir properties to identify zones with a high-production potential: reservoir opportunity index (ROI) and simulation opportunity index (SOI). The Petrel* E&P software platform was chosen as the integrated platform to implement the workflow. The fit-for-purpose time dependent 2D maps generated by the Petrel platform facilitated the decision-making process used for locating new wells in the dominant flow system and provided immense support for field-development plans. The difference between the two methods is insignificant because of reservoir tightness, limited interference, and natural uncertainty on compartmentalization. At this stage, pressure is not a key parameter. As a result, unlike brown fields, less weight was given to simulated pressure, and SOI was used to select the well locations. The results of this study show that implementing these workflows and obtaining the resulting maps significantly improve the selection process to identify the most productive areas and layers in a field. Also, the optimum numbers of wells using this method obtained in less time and with fewer resources are compared with results using traditional industry approaches.