Modeling Multilaterals

Transcription

1 Modeling Multilaterals Over the past decade the Middle East s oil and gas companies have come to rely on advanced drilling and completion techniques (such as horizontal wells, multilateral wells and intelligent completions) to develop new fields and to maximize recovery from existing assets. The need to simulate these advanced wells has called for a radical rethink in the well models found in today s reservoir simulation packages. Paul Fjerstad and Jonathan Holmes look at these new models that must be able to determine local flow conditions (flow rate and pressure for each fluid) at every point in the well, allow for pressure losses along the wellbore and across any flow devices that may be present in the completion.

2 Over the past decade, the Middle East s oil and gas companies have come to rely on advanced drilling and completion techniques (such as horizontal wells, multilateral wells and intelligent completions) to develop new fields and to maximize recovery from existing assets. The need to simulate these advanced wells has called for a radical rethink in the well models found in today s reservoir simulation packages. These new models must be able to determine local flow conditions (flow rate and pressure for each fluid) at every point in the well, and allow for pressure losses along the wellbore and across any flow devices in the completion. Oil industry experts began to develop computer simulations of hydrocarbon reservoirs in the late 1940s and the simple simulation methods they developed started to appear in the early 1950s. These were little more than solutions of differential equations for fluid flow in a homogeneous reservoir with simple geometry. The next step was computer modeling of more realistic reservoir architectures using blocks or cells. By the 1960s, improved algorithms and increasing computer power had combined to solve equations faster and more accurately than ever before. The greater speed of processing encouraged geoscientists to increase the size and complexity of their models in the quest for greater realism and accuracy. Modern modeling and simulation programs can handle large numbers of gridblocks, and model much more complicated geology than the early systems. The models they produce can be used to match historical production data (history matching) and so improve the static geological model on which the simulation is based. The key problem in simulation is advancing the state of the reservoir through time, taking into account the affects of factors such as oil production or phase changes in the formation. In a typical computer model, reservoir properties are stored in matrices, with a combination of properties being used to define each block in the grid. A change in one block will affect the properties of its neighbors and the model must represent these changes step-by-step through the productive life of the reservoir. The main variables in the reservoir are gas, oil and water saturation and pressure. A simulator solves the equations that describe how these variables change examining the conditions within every gridblock and stepping forward in time to define the conditions at that new time. The fundamental principles of simulation have not changed over the years, but increased reliance on nonvertical wells and smart well technology have made the simulation process much more complicated. The multilateral challenge Dramatic improvements in drilling technology over the past thirty years have encouraged more and more field operators to develop their fields with multilateral wells. Many Middle East field operators drill multiple horizontal wells or drainholes to improve the production rates for their reservoirs. Multilateral wells have been used for exploration, infill development or reentry into existing wells. They can help to increase the effective producing length when horizontal penetration is restricted, or to minimize partial-penetration (vertical skin) effects in relatively thick formations. Many people consider multilaterals to be technically demanding and complex solutions, but in fact their complexity is dictated purely by field economics and the aims of the asset team. Some reservoirs call for a sophisticated solution involving a horizontal extended-reach well with multiple lateral and sublateral branches (Figure 5.1), but in many cases a relatively simple arrangement can deliver significant production benefits. Whatever the complexity of the design, all multilaterals share a common goal improving production while saving time and money. Multilateral wells improve vertical and areal sweep by enhancing productivity, or injectivity, or both in Figure 5.1: Multilateral wells may be relatively simple or very complex designs, but all share a common purpose improving production while saving time and money 48

3 sandstone and tight carbonate reservoirs. The technique has been applied in a broad range of reservoirs, but most published studies focus on its application in single-layer scenarios. However, many of the oilbearing zones in the world s most prolific reservoirs are made up of several different lithological units. In these reservoirs, rock properties can vary considerably through the stratigraphic sequence, mainly as a result of geological factors such as rock type and depositional environment. For example, siltstones or chert units are usually lowpermeability layers, while in conglomerate formations permeability values are generally much higher. Recently, multilateral wells have been put to new uses providing high-quality completions in layered formations where the producing layers have strongly contrasting reservoir properties, or in wells where the possibility of water or gas coning precludes the use of conventional stimulation techniques. One of the toughest decisions facing the reservoir engineer is where to place these multilaterals so that they provide the best possible drainage. Given the costs associated with choosing the wrong trajectory for well placement, accurate prediction of well performance is a crucial part of field-development planning. Geologists and reservoir engineers must ensure that new multilaterals are not affected by borehole breakout. Figure 5.2: This simplified, multisegment well has four lateral branches in different producing formations To achieve this they must plan well orientation relative to the principal stresses in the reservoir formation. It is also important to predict the longterm performance of a new well. For example, if the well cuts across open fractures this may initially boost oil production, but the gains could be short-lived if water coning and early breakthrough follow. In the Middle East normal practice is to define the trajectory of a new multilateral using the static geological model and drill it openhole, without using a well model. Schlumberger is proposing a different approach modeling possible well trajectories before drilling begins and predicting long-term behavior on the basis of the model and new information gained while drilling. The key benefits of this approach are that the asset team can assess the potential benefits of intelligent components and test alternative drilling and completion strategies. This helps to optimize well placement, saves money and improves hydrocarbon recovery. The well model allows the reservoir engineer to design and deploy intelligent completions that can be controlled from surface to optimize production rates, reduce water cut and extend the productive life of the well. The combination of model and intelligent completion helps geoscientists to predict problematic behavior for example where and when water and gas tongues could develop long before it happens. This, in turn, allows the engineer to insure against the problem. With good prediction of breakthrough points, the well can be completed with chokes in the appropriate zones. This predictive capability is not possible when trajectories are designed using static geological models. Predicting performance Layering, branch location and wellbore effects have a direct influence on multilateral well performance. The key problems that asset managers face are how to identify layers that need multilateral branches and what branch lengths will maximize productivity. Effective planning and execution of drilling operations should deliver the right well trajectory, minimize water sumps and maximize cleanup. When asset teams try to predict the productivity of horizontal wells, they generally assume that inflow will be proportional to the properties in a single layer the layer that contains the well. However, this simplistic approach may not stand up to close scrutiny and engineers often seek additional information to make a more accurate assessment. In many fields, data gathered from nearby vertical wells provide valuable insight into the pressure depletion and layer pressure differentials that affect the performance of a horizontal well. Vertical wells also reveal the flow characteristics and inflow performance for producing completions that reservoir engineers use when defining new well trajectories. However, to get a realistic picture of multilateral performance in heterogeneous multiphase reservoirs, engineers must apply numerical simulation techniques. These techniques represent each multilateral well as a series of gridblock completions in the formation (Figures 5.2 and 5.3). This numerical modeling of existing field data helps the reservoir engineer to identify possible completion locations for multilateral wells and to ensure that the multilateral trajectory that is selected and drilled minimizes or eliminates water sumps. 49

4 Not equal to the sum of its parts The productivity of a multilateral well is not usually the sum of the predicted production from its branches. In addition to individual branch flow, reservoir engineers have to evaluate the pressure-competition effects of inflow performance and the interference effects of commingled production before making an economic assessment. If these factors are ignored, the production estimates will be wrong, and could even be wildly optimistic. The models that engineers create to assess productivity typically describe conditions within each well by a single, implicit main variable corresponding to the well s flowing bottomhole pressure. Hydrostatic pressure variations over the perforated sections of the well are included using explicitly calculated, local fluid densities. A variant of this type of model uses three implicit, main variables and is necessary for black-oil simulation. The extra variables in the black-oil model correspond to the flowing fractions of water and gas within the wellbore. They allow engineers to model wellbore crossflow between completed zones. The concept can be extended to compositional simulation by including one extra variable for each component. In cases where multilateral wells are not completed correctly, the benefit of additional productivity can be offset by reduced lift performance. Commingling flow from several multilateral branches has some advantages. For example, the combination may allow a larger tubing size and increase the flow rate. However, potential disadvantages include reduced artificial lift potential and nonuniform pressure reductions and reservoir depletion. When a well produces with artificial lift, the multilateral completions will tend to raise the pump or bottom gas-lift mandrel and reduce performance. Crossflow is a particularly serious problem in multilateral wells. If one branch of a multilateral well waters out before the others, this creates an imbalance between the branches and could cause additional production problems. If the multilateral targets are chosen carefully at the design stage, crossflow should be very rare during production. During shut-in periods, however, crossflows are hard to avoid unless special hardware has been installed downhole. The simulation toolbox Many procedures have been tried to deliver the best simulation results in the shortest time but the most common in use today are known as the fully implicit method and the IMPES (implicit pressure, explicit saturation) method. A fully implicit model solves for both pressure and saturation at the end of each simulation timestep, while an IMPES model uses saturation values from the beginning of the timestep to solve for pressure at the end of the step. Each has advantages and disadvantages, and the major commercial simulators support both methods. IMPES is computationally faster because fluid saturations are solved in a single matrix calculation. The simulator then iterates until pressures in the gridblocks have an internally consistent solution. However, this solution may be difficult to find if saturation (which the IMPES method assumes is constant within a timestep) changes rapidly during that step. The IMPES model can cope with this by decreasing the length of the timestep but this solution may require a huge number of timesteps to achieve a stable solution. Figure 5.3: The use of a simulation grid improves the model s description of the well path and helps engineers to avoid water sumps in their horizontal wells Unfortunately even this may not be enough to reach a firm conclusion and in some cases the simulator may be unable to find a stable solution using this method. The fully implicit method is a more stable option with saturation and pressure being obtained simultaneously so any difference between one timestep and the next is less critical than with the IMPES method. This means that timesteps can be larger but it also means that the fully implicit method requires additional processing time to achieve solutions. However, in a reservoir where properties change rapidly the fully implicit method may provide a solution in less time than the IMPES method even though each iteration takes longer to complete. The ECLIPSE 300* reservoir simulation software has a feature called AIM the adaptive implicit method that takes a flexible approach. When the program finds parts of the model where properties change rapidly and the IMPES method may be unable to provide a reasonable result, the simulator adopts an implicit approach in the relevant areas. Users can specify a maximum proportion of the model to be solved implicitly. Experience suggests that if more than 20% of the model falls into the implicit category, results will be achieved more quickly by solving the whole model with an implicit method rather than using the combined approach. 50

5 The right model for the job Fully implicit, strongly coupled well models provide a good representation of most standard wells in a reservoir simulation, but they have two important limitations: The pressure gradient within the wellbore is, at formation level, an approximate hydrostatic head calculation that ignores the effects of friction. A single set of variables is used to represent the entire contents of the wellbore within the formation, so that when crossflows occur, the mixtures that flow back into the formation represent the average contents of the wellbore. This would only be true if the fluids in the wellbore were completely mixed. As horizontal wells and completions became more widespread, reservoir engineers began to worry about the loss of productivity caused by frictional pressure gradients along the wellbore and wondered how they could evaluate this. They then extended the standard, three-variable model to calculate the frictional pressure losses along the completed sections of a well. In this type of model (the wellbore friction model), the three variables are supplemented by an additional main variable for each gridblock completion. These blockrelated variables represent the frictional pressure drop between the completion and the well s bottomhole reference point. The friction calculation is based on a homogeneous flow model. The hydrostatic head within the wellbore is calculated in the same way as in the three-variable model, using explicit values of the local, fluid-mixture density and assuming homogeneous flow. The wellbore friction model has been used to simulate horizontal wells in many projects. It was easily adapted to handle multilateral wells, but the flow situation in multilaterals underlines the main shortcoming of the model the handling of crossflow. The model has just two variables to represent the fluid contents of the entire well, so fluid flowing back into the formation will do so in proportions that reflect the average contents of the wellbore, irrespective of the location of the injecting completions within the well. This treatment is clearly inappropriate in most circumstances, for example where crossflow is confined to one particular branch of a multilateral well. Engineers can model crossflow much more effectively by dividing the wellbore into segments, each of which has its own set of strongly coupled main variables. This multisegment approach can also include more sophisticated models to calculate the wellbore pressure gradients from local flowing conditions. The multisegment method is readily adapted to cope with multilateral wells, preserving the correct upstreaming of fluid flows in a network of well branches. At the same time, it allows engineers to model the next generation of smart wells that contain downhole flow-control devices. This is possible because each segment can be configured to represent the effects of specific devices. A closer look at segments The multisegment well model provides a detailed description of flow in the wellbore. Although specifically designed for horizontal and multilateral wells, it can also provide a very detailed analysis of fluid flow in standard vertical and deviated wells. The well-model equations are solved fully implicitly and simultaneously with the reservoir equations. This provides stability for the model and helps to ensure that operating targets are met exactly. The detailed description of flow conditions within the well is obtained by dividing the wellbore and any lateral branch into a number of onedimensional segments (Figure 5.4). Each segment has its own set of four independent variables to describe local fluid pressure, flow rate and flowing fractions of water and gas. Each segment may have completions in one or more reservoir gridblock, or none if there are no perforations at that location. The four variables within each segment are evaluated by solving material-balance equations for oil, water and gas, and a pressuredrop equation that takes into account the local hydrostatic-, friction- and acceleration-pressure gradients. The pressure drop can be calculated using one of two multiphase flow models built into ECLIPSE MSW* software: a homogeneous-flow model where all the phases flow with the same velocity; or a drift-flux model that allows slip between the phases. This allows rapid calculation and provides continuous results across a wide range of flowing conditions. A third option involves precalculated tables for determining pressure drop across a segment. This allows the reservoir engineer to use more complex and computationally expensive multiphase flow models if their results are first translated into Figure 5.4: This small part of a simulation grid indicates how the nodes for each grid cell in the multilateral branches are connected to each other and to the vertical well 51

6 tabular form. Obtaining the productionrelated pressure drop by interpolating the table is rapid and efficient. Tables also provide a very effective way to represent pressure losses that occur in some flow control devices such as chokes, because more accurate models of these devices would require additional computing time. ECLIPSE MSW allows simulation engineers to change choke settings at any time during a simulator run by switching to a different table. The ability to control the settings of subsurface devices allows the simulation engineer to predict the changes that will result from modifications to production strategy for the well. Heel Low pressure Toe Figure 5.5: This pressure distribution plot around a horizontal wellbore assumes no friction in the wellbore. The assumption produces an even and symmetrical pressure drop along the length of the well Reaching more realistic results Multilateral well performance in multiphase and heterogeneous reservoirs can be evaluated using numerical simulations. Including the wellbore friction component of flow greatly enhances the description of the pressure profile around the wellbore (Figures 5.5 and 5.6). Using a more realistic wellbore flow indicates that pressure is being reduced faster at the heel end of the horizontal well than at the toe end. A lower pressure zone at the heel of the well will increase oil production from the area around the heel. If the frictional pressure drops around the wellbore were ignored, the model might produce an unrealistic drainage pattern. This could, in turn, lead to an overoptimistic prediction of production rates and hydrocarbon recovery. When engineers include the wellbore friction data they produce a realistic picture of skewed pressure distribution towards the heel of the well (Figure 5.7). Figure 5.8 shows the pressure distribution along the horizontal wellbore when the frictional pressure drops caused by the roughness of the pipe or by openhole completions are included. The roughness of the drilled section associated with openhole completions causes very high frictional pressure drops along the wellbore. High pressure Low pressure Heel Low pressure Toe Figure 5.7: A vertical section of pressure distribution along the wellbore shows the low-pressure zone skewed towards the heel of the well. This is caused by wellbore friction Figure 5.6: This pressure distribution plot around a horizontal wellbore was produced with friction effects taken into account. It clearly indicates that pressure depletion is most significant at the heel end of the well 52

7 Near Wellbore Modeling The NWM* Near Wellbore Modeling tool helps reservoir engineers to introduce detailed near-wellbore models or sector models into existing full-field simulation models. It supports common workflows such as simulation modeling while drilling, and using simulation to optimize completion intervals in vertical, highly deviated, horizontal or multilateral wells. These workflows rely on the use of a simulator to predict the dynamic behavior of well configurations and completions. To make them effective, engineers have to generate a reduced near-wellbore model or a sector model in the area of interest within a larger or full-field simulation model. New technology in the NWM tool provides better modeling of wellbore fluids and the near-well reservoir region. Automation of engineering tasks makes the NWM tool easy to use. A key feature is the ability to simulate the near-wellbore model independently of the full-field simulation model while still capturing its influence. This feature, coupled with automation, makes it possible to generate and run small, local studies within a short time frame for example, during drilling operations. The input model to the NWM tool is the full-field simulation model, which captures the global behavior of the reservoir. The tool also generates boundary conditions for a fullfield model simulation forecast so that a near-wellbore model can be simulated independently. Alternatively the nearwellbore model can be reintegrated into the full-field simulation model in order to improve its accuracy. The generation of a near-wellbore model is a three-step process: defining the volume of interest defining the well gridding. A volume of interest (VOI) is designed interactively. The full-field simulation model cells within the VOI are extracted and taken as the basic NWM geometry, including rock and fluid property data. All wells within the VOI are extracted from the full-field model to give a coarse well description in the NWM. If any of the wells are horizontal or highly deviated, the NWM tool is capable of using a more detailed well path design for new infill well locations. The tool allows explicit well path definition or import of a well path from drilling engineering software such as Drilling Office* integrated drilling software. Perforation intervals can be edited by the user, imported from a file, or inherited from the full-field simulation model. Multilateral wells are supported. A multisegmented well model can be generated automatically to improve modeling of internal wellbore flow control devices such as intelligent chokes, wellbore multiphase fluid flow, frictional pressure losses and multilateral well configurations. Near wellbore rock and properties can be modified within annular zones to improve near wellbore flow. The NWM tool provides several automatic local grid refinement options. It also provides facilities for adding, moving and modifying well tracks and allows automatic implementation of a multisegmented wellbore model. Simulation in Saudi Arabia A well-model study conducted recently in Saudi Arabia helped field operators to make an informed choice between drilling a single horizontal well and developing the field with stacked multilateral branches. For the single-well option, the study also compared the performance predictions of a realistic model that included wellbore friction effects with one that did not. The multilateral design used to conduct the simulations has a vertical wellbore and three branches equally spaced around the main well in a tripod configuration (Figure 5.9a). The simulation recorded early pressure disturbance when the well was first opened (Figure 5.9b). The most productive branch will produce most of the reservoir fluids and, therefore, cause the largest pressure drop. This means that the pressure disturbance around the main (vertical) wellbore is not symmetrical. However, once the well has been on production for some time, the pressure distribution changes to produce a more cylindrical drainage pattern. In later stages of the depletion process there is direct pressure communication between the branches and the pressure distribution resembled the kind of pattern that might be expected from a single giant vertical well (Figure 5.9c). A comparison of production performance (Figure 5.10) contrasts the situation in a single, horizontal well assuming no wellbore friction, a single, horizontal well with wellbore friction, and a stacked multilateral well. The prediction from the single, horizontal well without wellbore friction is clearly far more optimistic than the expectation when wellbore friction is included. The performance of the stacked multilateral well is much better than the horizontal well the production potential is greater and the anticipated sweep efficiency is higher as a result of improved access to reservoir zones that would be poorly produced by a single vertical or horizontal well. Individual branch flow rates vary, with each branch flowing according to its flow capacity. Bringing it all together The multisegment well model is a major improvement on conventional, horizontal or highly deviated well modeling. The key benefits include improved handling of multilateral topology. Individual branches are identified by their branch number and may be defined using planned or actual well deviation surveys. Branches may, if required, have subbranches. improved modeling of multiphase flow. Use of the drift-flux model or precalculated pressure drop tables can produce more accurate pressure gradients than the homogeneous flow treatment used in the other well models. The pressure gradient can vary from segment to segment throughout the well; it is calculated fully implicitly in each segment using the local flowing conditions. greater flexibility in modeling the wellbore path. This flexibility is achieved by using a variable number of gridblock completions per segment. This approach allows reservoir engineers to use many segments for greater accuracy or fewer segments for faster computation. 53

8 Frictional Effects on Pressure Distribution Figure 5.8: Pressure profile plot along the wellbore comparing models made with and without a wellbore friction variable Pressure, psia J cell along wellbore (dimensionless) The multisegment well model contains several options for modeling advanced wells: Specific segments can be configured to model flow-control devices, such as chokes, by providing a pressure-drop table that describes their pressure-loss characteristics as a function of flow. Various built-in, flow-control-device models are also provided. Variable pressure-loss multipliers can be used to represent smart devices that react to isolate high gas oil-ratio or water oil-ratio regions. Flow-limiting valves can also be used to represent smart devices that react to limit the flow of oil, water or gas through a segment. Other advanced well components, such as labyrinth devices and downhole separators, can be modeled. Crossflow can be modeled more realistically, as the fluid mixture can vary throughout the well. Wellbore storage effects can be modeled more accurately by dividing the wellbore into several segments. The drift-flux model allows phases to flow in opposite directions at low flow rates, so that phase redistribution within the wellbore can occur during shut-in well tests. Figure 5.9a: This Middle East simulation for Saudi Aramco used a tripod configuration for the multilateral wells Figure 5.9b: The pressure profile around a stacked multilateral well immediately after production startup Figure 5.9c: This pressure profile around a stacked multilateral well during the later stages of production shows the pressure communication between branches x axis x axis x axis z axis z axis HW-1 HW-1 HW-1 y axis y axis y axis z axis 54

9 A step in the right direction The exploration and production industry has used reservoir modeling and simulation techniques for several decades. The detailed models that can now be developed for individual wells are vital tools for efficient reservoir development. The well-simulation techniques can be applied in established and mature fields where asset team wants to avoid wateredout zones and high-permeability streaks that could promote further water breakthrough. However, the predictive modeling technique can also be applied to fields in the early stages of development using a model optimise drill revise cycle that makes full use of available well data. Well placement is one of the most challenging aspects of field development and errors can prove extremely expensive in terms of rig costs for redrilling or damage to the reservoir. By adopting a modeling system that delivers realistic predictions of well performance and helps geoscientists to locate the optimum trajectory for horizontal and multilateral wells, the industry is taking a major step towards improved production and reduced costs. STB/D 30,000 27,500 25,000 22,500 20,000 17,500 Single well (with friction) 15, Figure 5.10: A comparison of production predictions for a single well with no friction, a single well with friction and a stacked multilateral well Stacked multilateral well 1200 Number of days Single well (no friction)