Evaluating production strategies for liquids-rich shale wells

The performance of gas condensate reservoirs is difficult to predict without conducting detailed reservoir modeling to enable the reservoir and completion engineers to compare predicted performance from various wellbore and completion configurations. A better understanding of reservoir conditions and hydraulic fracture properties and their interaction under production conditions can be used to improve well and completion designs and to help maximize hydrocarbon recovery in liquids-rich shale developments. Significant challenges that impede accurate computer modeling of these reservoirs include the following:

• Adequately gridding horizontal wells with multiple fractures;
• Availability of adequate pressure, volume, and temperature (PVT) data for liquids-rich environments;
• The selection of an optimum stimulation treatment design; and
• The low effective reservoir permeability.

For the purpose of this study, reservoir information from a typical liquids-rich well in the Eagle Ford formation was used to evaluate a multitude of reservoir conditions, completion options, and production control to help understand how production can be optimized in a complex liquids-rich environment.

FIGURE 1. The workflow is detailed for the decision-under-uncertainty approach to identify desirable completion and production strategies. (Images courtesy of Halliburton)

Fracture, reservoir parameters considered

To have a better appreciation of well completion practices in relation to reservoir dynamics through a stochastic reservoir modeling approach, several parameters (Table 1) were varied.

A homogeneous reservoir model was created that simulated average properties from the liquids-rich section of the reservoir. The sample was 2,286 m by 1,219 m by 79 m (7,500 ft by 4,000 ft by 260 ft), which was modeled with a grid of 48 by 17 by 7, along with local grid refinement to model the fractures, bringing the total number of cells for the 54 fracture cases to 101,910. The lateral length of the well was kept constant at 1,676 m (5,500 ft). Cases with 36, 54, and 72 fractures were simulated.

For this model, the well was fractured in 18 stages, each with four perforation clusters. The number of fractures equaled the number of fractures per stage multiplied by the number of stages. For example, the 36-fracture case would correspond to an example where only two clusters developed fractures in each stage. Even though the fractures in the same well can be quite different, a simplified model with identical fractures was used for analysis. Table 1 lists the data for the base case.

Setting of stochastic cases

The completion and production decisions that can yield the best returns for cumulative oil production at the end of a 5,000-day simulation cannot be collected directly because major sources of uncertainty influence the production behavior of the well. Therefore, an optimization-under-uncertainty approach, as described in Figure 1, was used to maximize the mean value of the cumulative oil produced by an optimizer. The mean value is obtained by sampling the uncertainty parameters multiple times for each set of decision parameters chosen by the optimizer from the 96 possible combinations. The solution space was exhaustively searched by the optimizer by running all 576 possible simulation cases.

TABLE 1. The parameters varied and their values used in the base case simulation are shown.

Results, interpretation

The results were used to determine the most important factors influencing cumulative oil and gas production after simulation. The factors on cumulative oil production, in decreasing order of effect and using data for all of the 576 runs, are illustrated in Figure 2. This chart also shows that these factors influence cumulative oil and gas production differently or to different degrees.

Fluid type is most highly correlated with cumulative oil production but not as much with cumulative gas production. This result is not unexpected because the liquid yield does not affect gas production as much as oil production since, with its high mobility, gas is much more easily produced than oil whether the reservoir fluid is a lean or a rich condensate. Oil production, by contrast, is dependent on a favorable fluid type being available in the reservoir. All other factors correlate in the same direction with oil and gas production but by different degrees.

Fracture length is the second most important factor for oil production and the most important factor for gas production. A completion that maximizes fracture length is favorable to both oil and gas production. Matrix permeability has a high correlation with cumulative production. Similarly, an increased number of fractures – or decreasing fracture spacing – is positively correlated with both oil and gas production. Fracture conductivity is positively correlated with gas production. Higher cumulative production is obtained for lower flowing bottomhole pressure. The conductivity endurance of sand and ceramic proppants has a small effect on cumulative oil and gas production considering all simulations together.

Improved stimulation effectiveness and better reservoir quality also tend to result in a more rapid decline in reservoir pressure and in the accumulation of more condensates in the reservoir.

FIGURE 2. A tornado chart shows all the simulations taken together.

Within a specific reservoir environment characterized by a specific matrix permeability and fluid type, the decision parameters appear to follow a hierarchical pattern when examining cumulative production or other time series-based behavior of the wells. The ranking of factors in order of their impact depends on the specific reservoir environment (Figure 3).

Reservoir fluid type is a primary variable of production and should be closely considered to appreciate production performance characteristics in liquids-rich reservoirs. Special efforts should be made to collect reliable reservoir fluid samples in the early part of production from a gas-condensate well. Matrix permeability is another uncertainty that accounts for a large percentage of variability in production (of both oil and gas), and techniques to measure ultra-low permeability should be improved and employed.

The best cases – those with the highest mean cumulative oil production – all have longer and more conductive fractures, and the opposite is true for the worst cases. However, the best cases also carry higher risk due to higher variability (standard deviation). Since the matrix permeability is not typically known – because of the difficulty associated with making laboratory measurements and petrophysical estimation in shales and the PVT data that are either not collected or suffer from high uncertainty due to separator sampling – the best combinations can still yield poor cumulative oil production. On a risk-reward basis, decisions can be made to complete and produce a well that gives the highest mean production – reward – at an acceptable standard deviation – risk.

FIGURE 3. The trellis plot shows cumulative production of oil (top) and gas (bottom) arranged with CGR (1=30, 2=75, and 3=150 bbl/MMscf) as columns and matrix permeability (20 nD and 200 nD) as rows.

Observations

The methodology adopted in this study is a framework to weigh the quantitative impact of decision and uncertainty variables and their combinations on the desired performance from wells. Picking winning combinations of completion and production decisions also requires defining the tolerance level for variability in well productivity, which arises from uncertainty in matrix permeability and CGR. Since a large percentage of variability in well productivity is accounted for by the variability of these two factors, reducing their uncertainty by collecting more data and developing newer measurement techniques could lead to decision-making with less risk and more reward.

Acknowledgment

This article is based on SPE Paper 166177, which was presented at the Society of Petroleum Engineers Annual Technical Conference and Exhibition in New Orleans in October 2013.