Increasing Recovery Through New Reservoir Management Concepts

Olwijn Leeuwenburgh, Lies Peters, Frank Wilschut, Remus Hanea,
Oscar Abbink and Peter van Hooff, TNO

Increasing recovery by just one percent could add about 80 billion barrels of oil to global reserves. Oil companies have realized this potential and in response have expressed the goal to raise the recovery factor for their fields to as high as 70%. New developments in drilling and completion technology, as well as novel IOR techniques, will play an important part in realizing this goal. An equally important contribution will be delivered by the introduction of improved reservoir management concepts.

The typical current industry practice is that decisions on future field development are based on simulating future development scenarios on a single reservoir model, which has been matched to production data by time-consuming manual adjustment of a very limited number of reservoir parameters. In recent years, the realization has grown that future production scenarios should be tested on many geological realizations of the reservoir model in order to account for the inherent uncertainty in our knowledge of the reservoir properties on scales smaller than the inter-well spacing. Accordingly, many software providers have started to offer (parallel) simulation functionality that enables the reservoir engineer to simulate multiple geological scenarios. However, what has been lacking to a large extent is the possibility to adjust not just a select few, but all parameters and properties of the reservoir model, and to incorporate all available types of measurements in a more-or-less automated and consistent manner. For example, it is generally not possible to incorporate data from repeated seismic surveys by manual processess in a (geo-)statistically correct way. The development of new measurement techniques which generate very large numbers of data, including smart-well completions equipped with down-hole sensors, will pose serious challenges to existing company workflows for model history matching. Similarly, proper optimization of future production and development strategies requires approaches which are both robust to uncertainty and which can handle hundreds to thousands of controls (e.g. Leeuwenburgh et al., 2010). This will especially be important for so-called ‘smart’ or ‘intelligent’ fields, which enable much improved monitoring and provide many more control options than conventionally developed fields.

Closed-loop reservoir management
Research institution TNO, Delft University of Technology, and Shell International Exploration and Production in 2005 took the initiative, with the start of the ISAPP (Systems Approach to Petroleum Production) research program, to develop improved algorithms that will enable new reservoir management workflow concepts. Many methods for incorporating measured data with simulation models were explored and tested, and new concepts for integrating different elements of the complete workflow were pioneered, resulting in what has been called ‘closed-loop reservoir management’ (Jansen et al., 2009, see Fig.1). This concept introduces intelligence and integration to the standard company workflow, enabling the optimization process to be run at the ‘right time’, and with higher frequency.

Figure 1: The ISAPP closed-loop

Figure 2: The Brugge 2-phase benchmark model for closed-loop water-flooding optimization.

Figure 3: Results from the Brugge workshop on closed-loop reservoir management: improved history match quality (low error) tends to enable improved field development strategies and a considerable increase in realized asset value (high NPV). The red points are based on a history match over a longer time period than the blue points. The best results were all obtained with the Ensemble Kalman Filter (circles), while standard methods (squares) were found to deliver significantly poorer results.

These ideas were tested in an SPE workshop on Closed-loop Reservoir Management, held in 2008, for which the benchmark Brugge reservoir model was developed (Fig.2). The workshop results, documented in Peters et al. (2010), as well as subsequent studies, have validated the ideas behind the closed-loop concept, indicating that such an approach will result in more value, as measured for example by NPV or total recovery (Fig.3). Several clear conclusions could be drawn:

1.    The use of advanced computer-assisted history matching approaches provides improved consistency with both geological reservoir knowledge prior to history matching and with dynamic data.
2.    The incorporation of additional types of measurements, such as time-lapse seismic, in the history match improves the reservoir model.
3.    An improved set of reservoir models enables a more reliable (robust to uncertainty) forecast of reservoir value.
4.    The use of advances techniques for optimization of production strategy (scheduling) leads to higher value.
5.    An increased frequency of runs through the closed-loop (i.e. history matching and future production strategy or development optimization) provides higher value.

It has become clear than any serious effort to increase ultimate recovery will have to involve a change in reservoir management thinking along these lines.

An important element of the closed loop is generating a set of model realizations which are consistent with all available measured data, and which properly reflects the remaining uncertainty in the reservoir parameters. Companies tend to build ever larger and more complex models, accompanied by similarly increasing computing facilities. These models are becoming impossible to manage manually by a single reservoir engineer.

The incorporation into such models of large numbers of data, such as produced by seismic surveys or by wells equipped with down-hole sensors, will only be possible using computer-assisted methods. This will require both trained personnel as well as user-friendly software tools, enabling the reservoir engineer to spend his time instead on making better decisions.

Research has suggested that only certain scales, regions or aspects of the model may be relevant in matching the model to data, or in controlling the output of simulations of future development scenarios. This offers the potential to use clever up-scaling or model-reduction methods, which could significantly reduce computing time and therefore cost, and enable increasing the frequency of the loop, resulting in better results.

Finally, the subsurface characterization may be improved by incorporating data which is traditionally not typically used in the same way as production data are used. Within ISAPP new developments in this area have been in the use of time-lapse seismic, subsidence, gravity and bore-hole radar measurements. Challenges also exist in the design of surface and subsurface sensor networks for such soft-sensing data types, and in extracting the relevant information from the resulting data sets.

New reservoir management tools
Implementation has started of some of the new concepts and methods coming out of the ISAPP program into tools that can be integrated in the workflows of oil companies. An example is the history matching tool which has been integrated with the JewelSuite® modeling package. The tool consists of an Ensemble Kalman Filter which has been integrated with JewelSuite® property modeling functionality, to simulate multiple geological realizations. It enables the reservoir engineer to adjust all grid block properties and parameters of the reservoir model in a semi-automated fashion in order to achieve a history matched set of model realizations consistent with all production data. This tool is the first of its kind and operates with a host of commercially available reservoir simulators. It represents the first step towards a full implementation of all elements of the closed reservoir management loop in a single modeling package. It demonstrates both the feasibility and potential of a major new development in reservoir management outside of academia and research groups. Results obtained for the model depicted in Fig. 4 are shown in Fig. 5. Figure 6 demonstrates the value of a computer–assisted history match approach relative to a manual approach on predictions for a new well.

Figure 5: Results obtained with the integrated history matching workflow for the 3-phase channelized benchmark model. Shown is the product of permeability (K) and thickness (h) a) True Kxh. b) Initial Kxh before history matching, c) Estimated Kxh after history matching with the Ensemble Kalman Filter

Figure 6: Comparison between predictions of oil production and the realized production (truth) for a new well.

The workflow described above will be extended with functionality for computer-assisted development planning. The experience gained in the ISAPP program has led to the development of a robust ensemble-based optimization tool that can be used for automated optimization of both production scheduling and well placement.  An example for well placement optimization is shown in Fig. 7.

Figure 7: Optimization of well placement in a simple rectangular reservoir containing an L-shaped oil-trapping fault. a) Oil sweep resulting from water injection and production from initially proposed well locations. b) Oil sweep resulting from production of the field using optimized well locations. Super positioned are the displacement vectors of the wells with respect to their original positions. The improved sweep efficiency represents 10% increased oil recovery.

Future developments
TNO and Delft University of Technology are continuing the ISAPP knowledge centre and are currently inviting oil companies to participate. The main aim of the ISAPP-2 program will be to transfer the new concepts developed in the first ISAPP program to the world of real operations. To this end the program is looking for partners who are interested in bringing in actual assets in order to demonstrate the added value of these concepts and to incorporate the required tools within their workflows.

Many challenges still remain. Therefore, the ISAPP-2 program will additionally continue the fundamental and exploratory research on closed-loop concepts and computer-assisted methods. Partners will be able to participate in this research by close cooperation with staff members of the involved institutions. The ultimate aim is to enable 10% or more increase in recovery by introducing improved methods and concepts into the reservoir management workflow of companies. We believe that this goal can be achieved best by close cooperation between those involved in R&D and the engineers who will use the results in daily operations.

References
ISAPP-2 website: http://www.isapp2.com/

Peters, E. et al. 2010: Results of the Brugge benchmark study for flooding optimization and history matching, SPE Reservoir Evaluation and Engineering, p.391–405, SPE 119094.

Jansen, J. D., S. D. Douma, D. R. Brouwer, P. M. J. van den Hof, O. H. Bosgra, and A. W. Heemink, 2009: Closed-loop reservoir management, SPE 119089.

Leeuwenburgh, O., P. J. P. Egberts, and O. A. Abbink, 2010: Ensemble methods for reservoir life-cycle optimization and well placement, SPE 136916.