You must log in to edit PetroWiki. Help with editing

Content of PetroWiki is intended for personal use only and to supplement, not replace, engineering judgment. SPE disclaims any and all liability for your use of such content. More information

Integrated asset modelling in production forecasting

Jump to navigation Jump to search

Production forecasting is the application of data and models to predict, plan and optimize the hydrocarbon recovery and production with as much accuracy and precision as possible. To effectively model the entire production system, the data and models must be representative of the upstream and downstream processes. Integrated asset modeling is a holistic modeling approach that allows upstream and downstream components to be modeled directly instead of making simplifying assumptions.


To further illustrate this point, consider any typical oil or gas field. The hydrocarbon pathway forms a continuous system such that each molecule of formation fluid flows through porous media to the wellbore, up the well, through the surface gathering network, and then to the processing facilities. Despite this continuum in the upstream and downstream processes, it is common practice to sub-divide the asset into discipline-related domains, with each domain being modeled separately. Domains are typically separated into subsurface, surface hydraulic systems and processing facilities.

Types of models

A reservoir engineer builds a reservoir simulation model to investigate, for example, the impact of an advancing water front. A production engineer dealing with hydraulics/flow assurance develops models to analyze, for example, the impact of increasing water production in the gathering surface network. Finally, the facilities engineer must also investigate the capacity of the plant to handle increases in water production. Reservoir, production and process engineering are all being handled in separate silos with separate data models and separate simulation tools, while patently, as illustrated by the example, they are dealing with common problem (i.e., impact of water breakthrough). This creates a challenge in dealing with boundary conditions, domain specific assumptions and production constraints in both upstream and downstream processing.

Integrated asset model (IAM)

An integrated asset model (IAM) serves to break down these barriers between domains and build a total simulation system for planning and optimization. An IAM is a holistic, “big-picture” model of the entire asset from pore to plant to provide decision support that takes into account the complexities of interactions between sub-surface and surface domains. This includes the following:

  • The impact of changing fluid composition on system backpressure (e.g., increasing GOR and water cut)
  • The interdependencies between wells sharing a common production system resulting from variation in backpressure
  • The dynamic nature of a field where multiple reservoirs are connected to a common surface network sharing common capacity constraints on production and injection
  • Variation of fluid composition resulting from the dynamic nature of injection / reinjection
  • Asset level optimization

The value of such a model is its ability to more quickly and accurately evaluate the relative impact of these interdependencies, thus enabling comprehensive assessment of a wider range of scenarios, uncertainties and development opportunities.

Defining objectives and making approximations

When constructing an IAM it is important to clearly define the objectives to allow it to be appropriately tailored.

This may be broadly split into two categories:

  • How will the model be used (e.g., life-of-field forecast or short-term “operational forecasts”)?
  • What are the dominant physical processes and system constraints and what degree of complexity is required to model them?

Ideally, to avoid model maintenance overhead, a single IAM should be capable of addressing both short-term and long-term production forecasts without the need to make changes when switching between the two. In some circumstances, however, this may not be feasible due to model performance. For example, given a field with 500 wells, it may be impractical to couple a hydraulic network model with a high-resolution full-field reservoir simulator; therefore, some approximations may be necessary.

This leads to the importance of considering dominant physical processes. It is critical to ensure that an appropriate level of fidelity is included in each connected domain model such that key physical process are not “smeared out” by simplifying assumptions, which defeats the purpose of using an IAM. For example, if a field is in primary depletion, it may be sufficient to use a lookup table to model reservoir performance. However, if the field is undergoing waterflood, the reservoir model should at least be capable of modeling the impact of variations of water injection; i.e., a material balance or reservoir simulator.

Coupling a reservoir model to a surface network model

Traditionally, reservoir simulation has relied on simplifications to model the backpressure impact of top side facilities on the reservoir deliverability. One such example of this type of simplification is the use of well performance curves with tubing-head pressure constraints to represent the minimum allowable wellhead pressure. To illustrate, consider a reservoir with a single well modeled using a simulation-based forecast. The well has an oil demand rate target and a minimum tubing-head pressure limit representing the facility backpressure below which it cannot operate. Initially, the well is able to meet its demand target and will be choked back; however, as the field blows down, the flowing conditions in the well will result in the reduction of the tubing-head pressure. At some point, the tubing-head pressure will hit the stated minimum constraint, which is approximating the impact of backpressure from the downstream facilities infrastructure. When this happens, the well will no longer be able to meet its demand target and will go into decline.

INSERT Figure 1 Well constrained by tubing-head pressure limit. (Pending permission approval)

Backpressure impact

In reality, the backpressure impact from the surface facility is often a complex and dynamic which cannot easily be represented by a fixed tubing-head pressure constraint. As a consequence of making such an approximation, the timing and rate of decline has the potential to be under- or over-estimated depending on whether the pressure constraint is optimistic or pessimistic.

Variation of backpressure 

In order to more accurately model the variation of backpressure over the life of the field a surface flow line model can be dynamically coupled to the reservoir model. There are essentially two approaches to this: the implicit approach described by litvak et al (1997)[1] and the explicit approach similar to that described by Ghorayeb et al (2003)[2]. The implicit approach relies on constructing a new integrated model which requires that the reservoir, wells and network be solved as a single system of equations using a fully implicit approach. By contrast, the explicit approach involves using a controller to couple existing simulation models. Explicit coupling is typically achieved by exchanging information between the sub-surface and surface models at a coupling point (e.g. the bottom hole or tubing head). This information may include the well inflow performance relationships, compositions, and pressures. The ultimate aim of both implicit and explicit methods is to achieve balanced pressure, flow and compositions for each well. This balancing process is dynamic and is updated at multiple points through the life of simulation.

Improve forcest accuracy

To illustrate how coupling to a surface network model can help to improve forecast accuracy consider the scenario shown in the Fig 2 below. A well is initially producing at a constant rate with a relatively high tubing head pressure which is higher than the surface back-pressure described by the flow line pressure (FLP) curve. Note that the flow line pressure (FLP) experienced by the well is a function of a number of parameters including the plant inlet pressure, the geometry (length, diameter, roughness etc…) of the flow lines, heat transfer factors and the volume and composition of the fluid in the production system. The difference between the tubing head pressure and the well flow line pressure is the pressure loss across the well head choke. As the reservoir blows down, the tubing head pressure reduces and the choke dp reduces. At the point where the THP and FLP curves converge, the production system is bottlenecked and the well rate declines. As the well rate declines, so the volume of fluid in the surface network will decline. Consequently, there is less fluid in the flow line, so the back pressure will also be reduced and the rate of decline will be less compared to the previous case where the back pressure response was fixed.

INSERT Figure 2 - Well constrained by dynamic flow line pressure. (Pending permission approval)

Dynamic coupling

This is one example of how dynamic coupling to a surface network model can change the outcome of the forecast as a result of variation of backpressure response. Some other examples are shown below:

  • Model network equipment such as gas-lift valves, pumps, compressors and chokes
  • Account for flow-assurance issues, such as wax and hydrate formation
  • Facilitate reservoir coupling through a common surface network model
  • Incorporate network equipment (flow lines, chokes, compressors etc.) Into asset-management strategies (for example, network de-bottlenecking)

One must consider whether adding the additional complexity of surface flow line adds significantly to the forecast accuracy at all. In some cases, it definitely provides improved physical modeling and therefore more reliable results. However, in other cases, it may be considered as overkill. To illustrate, consider two illustrative examples.

Example one

INSERT Example 1: An onshore gas condensate field where gas from multiple reservoirs with varying fluid types is commingled and transported to the plant via a common trunk line. (Pending permission approval)

In this case, the commingled fluids in the multiphase flow line can be susceptible to significant variation in pressure loss over the life of the field. Factors such as seasonal temperature swing, compositional changes, volume of fluid in the pipe and solid deposition can impact the backpressure. All these factors can vary significantly over field life, resulting in dynamically varying backpressure response, which lends itself to modeling with a multiphase flow line simulation tool.

Example two

INSERT Example 2: an offshore oil field where production to wellhead platforms is initially separated and pumped via single phase lines to a central facility for processing. (Pending permission approval)

In the case of Example 2, it may be warranted to simplify the problem. The separation at the wellhead platforms creates a discontinuity between the upstream and downstream flow regimes by virtue of fixed separation inlet pressure. Consequently, there is little value in pressure-flow balancing of the upstream and downstream problems. 

As a final note, sub-surface/surface network coupling can be computationally expensive. For example, consider the following: a surface network model takes five minutes to solve (which is not uncommon, particular for large networks requiring constraints bound optimization). A conservative estimate of a loosely coupled system where the network is solved once per month and a forecast over a 40-year period would mean the total additional time taken by the network model would be 5 x 12 x 40 = 40 hours, which in all but the most complex reservoir models will far outstrip the run time of the sub-surface model. This obviously can present a significant performance bottleneck that represents a barrier to implementing such a solution and therefore pragmatic alternatives/assumptions must be made (i.e., simplification of the surface model, coupling at the manifold instead of the well level).

Interplay between multiple reservoirs

Interplay between multiple reservoirs connected through common capacity constraints where a field contains many reservoirs, which are not in pressure communication, it is common practice to construct and maintain separate models for each reservoir. In many cases, however, these reservoirs may be producing through shared surface facilities. Therefore, while the reservoirs have no sub-surface interdependency, they are connected through the surface. In effect, the reservoirs form a coupled system through common surface constraints.

There are two options for the generation of production forecasts when faced with this type of problem. The first is the use of forecast aggregation as described in Aggregation of forecasts. The second involves coupling the multiple reservoir models under an IAM.

Forecast aggregation

Forecast aggregation is a two-step process.

Step 1
Involves establishing reservoir deliverability.
May involve running multiple sensitivity scenarios using a reservoir simulation to generate deliverability curves.

Step 2
Generates the forecast by applying system constraints to limit production based on the deliverability generated in step 1.
May involve integration of these curves in a custom spreadsheet to apply the system constraints required for forecast generation.

While in many cases, forecast aggregation is a valuable tool for production forecasting, it suffers from a key draw back relating to the fact that steps 1 and 2 are decoupled. It illustrates the deliverability generated by step 1 and the constraints bound forecast generated by step 2. The oil deliverability curve has an associated water production curve and injection profile. When a field-level system demand constraint is applied to the oil production it is relatively straightforward to establish how the deferred production from constrained oil deliverability will translate to a plateau and an extended tail. However, it is less obvious how the reduced production will impact the production of water (and gas). In addition, reducing the production will create less requirements on injection, which in turn has the potential to impact both water and oil production.

Fig 3 illustrates how constraining the oil deliverability creates additional uncertainty in terms of the water production profile and the water injection profiles. For example, cutting back oil production will reduce water production, but not necessarily in a linear fashion (i.e., lower draw down may result in avoidance of water cusping). Similarly, assuming a voidage replacement strategy, reducing production would reduce requirement on water injection, which in turn has the potential to reduce the amount of produced water. In short, there is a degree of physical coupling between these phenomena that is not represented using a decoupled curve-based approach.

INSERT Figure 3 Forecast aggregation i. (Pending permission approval)

Coupling multiple reservoir models

Coupling multiple reservoir models under an IAM involves bringing all reservoir simulations into a common environment and applying system constraints over all reservoir models in a dynamic fashion. This is sometimes referred to as reservoir coupling and provides a way of generating production forecasts without the need to break the problem into two steps (i.e., first determine deliverability, second apply constraints).

Consider a scenario where a number of distributed fields are producing to a common surface facility.

INSERT Figure 4 - Forecast aggregation ii (Pending permission approval)

Assuming each reservoir has a deliverability potential of 10,000 b/d, the facility has an oil target of 3,000 b/d and the maximum liquid handling constraint from any one reservoir is 5,000 b/d. In addition, assume that the reservoirs are all injecting water for pressure maintenance under voidage replacement schemes. Initially, all reservoirs are producing with equal potentials and so may contribute equally to the production target (i.e., 1000 b/d of oil each). After some period of time, suppose the water cut in reservoir b and reservoir c starts to increase. As a result of this, the water injection to b and c must also increase to maintain voidage. Consequently, the water production will increase faster until the liquid production from these reservoirs hits the stated maximum limit of 5,000 b/d. At this point, the oil production from reservoir a will be increased to account for the curtailed oil production from reservoirs b and c.

This is a simple example, but it illustrates how the interdependencies between the three separate reservoirs relating to both production and injection can be captured dynamically in a single model. Scaling the problem up to include many more reservoirs and multiple levels of system constraints on both production and injection highlights the importance of considering this as an integrated production/injection system.

Fluid modeling and integration

Effective and accurate fluid modeling is important for all types of simulation models (reservoir, well, surface, facility). The needs of each discipline may vary (i.e., a detailed composition required for facility model is likely to be too detailed for reservoir simulation), therefore IAM must be able to translate fluid streams between simulation models.

INSERT Figure 5 - Fluid transition in an IAM. (Pending permission approval)

Fig 5 shows an example of how PVT models may vary from upstream to downstream domains. For effective integration, the prediction of fluid behavior across all models needs to be consistent. For example, a well-stream transferring compositional information from the reservoir to the surface model must satisfy the following compositions:

  • Mass and composition should be conserved between surface and sub-surface models
  • Volumetric rates (i.e., oil, water and gas rates) should match as closely as possible
  • PVT properties (e.g., phase densities, viscosities etc…) should be consistent

The complexity of the problem in ensuring consistency at the boundaries between the domains in the integrated asset model is governed by the type of the upstream and downstream fluid models. Table 1 summarizes some of the key combinations.

Upstream Downstream Comment
1 Black Oil Black Oil Oil, water and gas stock-tank flow rates as wells as phase stock-tank densities (and other pvt properties) may be obtained directly from the reservoir model. This information may then be set directly in the well/network model without any requirement to modify.
2 EOS model (N components) EOS model (N components) If there is a one-to-one mapping between surface and reservoir component slates, mass and/or molar rates may be transferred to ensure mass/composition conservation. Assuming the same equation of state is used on each side, common volumetric and pvt properties should ensue.
3 EOS model (N components) EOS model (M components) If the component slate used in the surface model is different from the reservoir it is necessary to de-lump the well streams from n to m components as part of the integration process. Typically, n < m because the level of compositional detail required by surface/process domains is greater than of reservoir.

Assuming the reservoir and surface fluid models have been derived from the same experimental data, it is relative straightforward to determine how reservoir eos heavy fraction pseudo-components (e.g. c7+) will translate to surface eos pure components (e.g. c7, c8, c9…).

4 Black oil EOS model (N components) The black oil model used in the reservoir model will typically have been derived from PVT lab experiments. Assuming the surface eos is matched to the same experimental data, it is possible to translate the black oil well stream to a compositional stream in a surface model. This is achieved using a technique known as black oil de-lumping.

It is good practice to ensure that fluid models from all models are consistent before embarking on model integration. For example, overlaying phase envelopes generated using the reservoir and surface EOS models to ensure similar volumetric prediction.

Table 1– Key combinations


An often-overlooked benefit of IAM is the level of collaboration that can be achieved between upstream and downstream teams. IAM initiatives are often “owned” by a single domain (i.e., reservoir or production engineering), and in many cases, this does not lead to the expected unfortunate “breakdown” of barriers between domains. Ideally, construction and maintenance of an iam should be a cross-domain initiative, which helps to ensure that the objectives of all parties are met and that key physical processes are conserved.

Torrens et al.[3] presents such an example. During the construction of an iam for a gas condensate field, a detailed review of the process facilities was carried out. The IAM was to be used for long-term planning (i.e., production forecasting) and so was “owned” by the re function. The construction process included a detailed review of the process facility requirements and revealed that the separator model previously used in the reservoir simulation was missing a key component; namely a high-pressure gas separation stage, which resulted in approximately 20-30% under-prediction in condensate production. The separator train models were adjusted to match the findings of the facilities, and excellent volumetric consistency was observed between the two domain models. Fig 6 Shows the observed condensate production plotted alongside the predicted value before and after the modification to the separation model in the IAM.

INSERT Figure 6 - Improved simulation accuracy through IAM collaboration. (Pending permission approval)


IAM is a simulation methodology that focuses on creating a holistic, representative model of an entire asset rather than breaking it into separate domain-focused chunks.

IAM provides a systematic way of reducing or eliminating simplifying assumptions associated with domain-specific simulators, which can lead to a more physically realistic solution and improve accuracy.

An IAM can act as a focal point for collaboration between engineering disciples focusing on a variety of different goals (i.e., reservoir management, production engineering and flow assurance as well as process facility optimization). If properly constructed and maintained, an iam may be dual purpose because it has potential application to long-term planning and forecasting as well as short-term “tactical” look-ahead forecasting and optimization (for example, answering operation “what-if scenarios”).

An IAM typically has many “moving parts” and therefore can be complex and hard to construct and maintain. The following key aspects should be considered when putting together an IAM.

  • Over/under complicate the problem – don’t add complexity if it is not required; don’t simply if it will smear out critical detail.
  • Consider the key objectives and focus efforts on obtaining them.
  • Fluid consistency – fluid models must be consistent moving from the sub-surface to the topside domain. In this case, “consistent” does not mean “the same”; rather it means that predicted fluid properties and volumetric behavior at different temperatures and pressures should be equivalent.


  1. Litvak, M. L., & Wang, C. H. 2000. Simplified Phase-Equilibrium Calculations in Integrated Reservoir and Surface-Pipeline-Network Models. Society of Petroleum Engineers.
  2. Ghorayeb, K., Holmes, J., Torrens, R., & Grewal, B. 2003. A General Purpose Controller for Coupling Multiple Reservoir Simulations and Surface Facility Networks. Society of Petroleum Engineers.
  3. Torrens, R., Mohamed, M. E., Al Bairaq, A., & Kumar, A. (2014, November 10). Integrated Asset Modeling of a Gas Condensate Field Operating Under Gas Recycling Mode. Society of Petroleum Engineers.

Noteworthy papers in OnePetro

Ghorayeb, K., & Holmes, J. A. 2007. Black Oil Delumping Techniques Based on Compositional Information from Depletion Processes. Society of Petroleum Engineers.

Hepguler, G., Barua, S., & Bard, W. 1997. Integration of a Field Surface and Production Network With a Reservoir Simulator. Society of Petroleum Engineers.

Roadifer, R. D., Sauve, R. E., Torrens, R., Mead, H. W., Pysz, N. P., Uldrich, D. O., & Eiben, T. 2012. Integrated Asset Modeling for Reservoir Management of a Miscible WAG Development on Alaska. Society of Petroleum Engineers.

Noteworthy books

Society of Petroleum Engineers (U.S.). 2011. Production forecasting. Richardson, Tex: Society of Petroleum Engineers.  SPE Bookstore or WorldCat

External links

See also

Production forecasting glossary

Aggregation of forecasts

Challenging the current barriers to forecast improvement

Commercial and economic assumptions in production forecasting

Controllable verses non controllable forecast factors

Discounting and risking in production forecasting

Documentation and reporting in production forecasting

Empirical methods in production forecasting

Establishing input for production forecasting

Integrated asset modelling in production forecasting

Long term verses short term production forecast

Look backs and forecast verification

Material balance models in production forecasting

Probabilistic verses deterministic in production forecasting

Production forecasting activity scheduling

Production forecasting analog methods

Production forecasting building blocks

Production forecasting decline curve analysis

Production forecasting expectations

Production forecasting flowchart

Production forecasting frequently asked questions and examples

Production forecasting in the financial markets

Production forecasting principles and definition

Production forecasting purpose

Production forecasting system constraints

Quality assurance in forecast

Reservoir simulation models in production forecasting

Types of decline analysis in production forecasting

Uncertainty analysis in creating production forecast

Uncertainty range in production forecasting

Using multiple methodologies in production forecasting