Integrated production system modeling (IPSM) can help in a number of processes within exploration and production, including field development planning, forecasting and surveillance, but one of its most important uses is in production system optimization (PSO). One of the most useful tools for PSO that is used in Shell is the producing the limit (PtL) peer assist event.
IPSM, PSO and PtLs are all multidisciplinary activities, so it should be no surprise that they
Figure 1. IPSM for Draugen. Five platform producers and five subsea wells produce Rogn and Garn reservoirs. (Graphic courtesy of Shell)
bring together a diverse set of participants and expertise to look at improvements to the reservoir, wells and surface equipment. PtLs are most often focussed on opportunities to generate production gains that can be realized in the short term (1 to 2 years).
Conventionally, these gains are estimated during PtL events based on limited technical work upon which subsequent activities are prioritized. However, IPSMs are being developed and applied increasingly during PtLs, thereby improving the overall quality of opportunity evaluation. The IPSM links the expertise from the different disciplines (e.g., PT, reservoir, subsea, process), helping to create a consistent and integrated way of working. The IPSM allows the team to quickly test out the incremental effects of potential improvements on the total production system (reservoir through export).
The typical IPSM tool for PSO work is Petroleum Experts’ “IPM” suite which includes GAP, MBAL and PROSPER. GAP can form the basis of a simple integrated model with the reservoir modeled in MBAL (either with a decline model or as material balance), the wells modeled with PROSPER, and the gathering system up to the separators pulled together as an integrated model. GAP has optimization routines that can optimize electric submersible pump (ESP) speeds, choke settings, gas lift distribution etc, so that each individual change is evaluated against the performance of the total system.
Some PSO opportunities may pay out over an extended period of time. It is essential to model the whole production system in an integrated way to ensure that the optimization ideas are evaluated taking all critical constraints, interactions and processes into account, so that the results are as realistic and practical as possible. To quantify gains of PSO opportunities on a longer time scale, an appropriate reservoir model must be used.
Depending on the complexity of the reservoir and the availability of data and modeling resources, this may require a full reservoir simulator. When the surface equipment is relatively simple but critical to the optimization (as when it forms the primary constraints) these can be represented in GAP. In complex surface network systems, with mixed fluids or constraints on certain hydrocarbon components, a more comprehensive modeling of surface equipment may be required, which can be done with HYSYS. HYSYS is another third-party application (from AspenTech) with which process engineers can model surface processes in great detail. In order to link external applications to petroleum experts tools, the IPM suite includes RESOLVE. RESOLVE also has optimization routines and can run script that can be configured in detail to coordinate the calculation among the different applications as well as to map out complete development scenarios.
Where integrated models have been applied during PtLs, the PtL team found that this yields great benefits. On the other hand, when no IPSM model was available, benefits were sometimes missed or overestimated.
As more and more assets have up-to-date and realistic IPSMs available, these can be used during PtL events to identify and quantify gains. The remainder of this article describes one example of such usage during a “PtL plus” workshop for the Draugen field in Norway held in March 2006.
The Draugen model consists of the three main elements: a representation of the reservoir, well models and the surface network.
An IPSM was constructed including wells, major pipe work, flow lines and separators. In order to evaluate terms on a medium-term time scale, tank models were constructed. During the preparation for the PtL, the models were calibrated and matched to the available data.
Reservoir models
There are two main reservoirs in the Draugen field: Rogn and Garn.
To facilitate realistic short- and medium-term forecasting with the model, a material balance model was developed. The model was history-matched using output from a full-field reservoir simulator, which consisted of production projected until 2024. The relative permeabilities were taken from the existing full-field reservoir simulator.
The Rogn reservoir was split into two tanks: “Rogn main” and “Rogn south.” Each of these has different PVT data. Transmissibility was modeled between the two Rogn tanks and was matched with the data. Water injection was divided between the two tanks (35% Rogn South, 65% Rogn main).
Once this was done, the reservoir
simulator data could be matched well. Conversion of the reservoir simulator output to a material balance model is a simplification that has some drawbacks (material balance models cannot predict water breakthrough or transient effects very well) but does allow an easier integration between team members. It requires far less calculation time, and thus various optimization options could be evaluated more rapidly. To reduce run times even more, decline curves could be used. Depending on the need, each of three options (reservoir simulator, tank model or decline curves) can be attached to the network model.
Compared to a reservoir simulation model, material balance models are also a step “back to basics” in the study of a reservoir. In the analysis of these reservoirs, we had to model substantial aquifer support in Rogn, where until now we had the impression that the bulk of the pressure support came from water injection. This is one of the triggers to study the interactions that were matched in the tanks and the effects of water injection on the production system behavior.
Well models
Inflow performance of the wells is very sensitive to water cut and draw-down, mainly due to the very large permeability of the reservoir (> 1 Darcy). Small errors in the reservoir pressure estimation yielded large differences in inflow performance. This was also the case for water cut. A 1% difference in water cut yielded as much as a 5% change in the oil production, 1.06 to 3.53 Mcf/d (30 to 100 cm/d) depending on the well. In light of the accuracy of water cut measurements, this was an important source of uncertainty and required good data to provide a match. Originally, all models had 0% water cut. The productivity index (PI) was initially derived from analytical calculations with relative permeability curves, then adjusted to match the production data from well tests. Significant changes in the PI were necessary in the current well models to get a good match.
After matching the PIs for all wells, the most recent well tests could be reproduced with an accuracy of about 5 to 10%. Further calibration of the relative permeability models for each well would be a useful refinement. This can confirm the relative permeability from special core analysis and to distinguish the rel-perm effect from potential inflow impairment (skin) effects. If impairment were large enough, this would provide opportunities for stimulation of the wells.
Most subsea well models had lift curves created using generally available flow correlation, which gave a good match. Some platform wells were recalibrated using the Shell- recommended flow correlation, which yielded more consistent results, in particular for gas lifted wells.
As preparation time was limited, the lift curves were generated for a limited number of injection rates to save time. Generating curves based on more rates will improve the IPSM solutions speed and accuracy.
Flow lines
Some of the flow lines had already been matched prior to the preparation phase described in this paper. For other flow lines data had not yet been made available. Some flow lines were re-matched using the recommended Shell correlation for flow lines, which yielded better results.
Well test data was used to match some of the flow line models. The well test data included wellhead pressures and pressure at the test separator. Comparing measured data with the models indicated restrictions in the flow lines. Similar pressure drops were noted between the first stage separator and the wellhead. This indicated that restrictions in the line, or the use of various fittings, introduce a pressure drop that is higher than expected from a flow line of a length of 196.8 ft (60 m).
This illustrates an important point: Finding opportunities for improvement (PSO) often begins during the construction and calibration of the model, where measurements are found to conflict with the model predictions. This could mean, of course, that the model is somehow incorrect, but just as often it means that some of the basic data is wrong. In this case it could be that the actual internal diameter is effectively smaller than expected.
Topside model
A topside model was created with simple constraints to represent the known limits. If compressor constraints are included, the gas-lifted wells can be optimized given the compressor capacity and the associated gas produced from the oil wells.
The model was used to explore several opportunities and scenarios. Most of these were logged in the list of opportunities of the PtL and captured in a spreadsheet.
Bottlenecks and constraints
One of the advantages of IPM is that it has an open interface (called Open Server), which can be used to control and automate most functions in the model. Taking advantage of this interface, a Shell tool was built to automatically identify bottlenecks in the model and to then propose an optimal strategy for removing these bottlenecks. An early version of this tool was put to use during the PtL. The tool identifies bottlenecks by reviewing which constraints are active. Active constraints mean that the values in the production system are close to or equal to the constraint values. In the Draugen model, the only active constraints were the subsea water injection pipeline pressure constraints. This was judged not to be a real bottleneck, and no gains were recorded from removing this constraint. As Draugen has been in decline for some time, much of the equipment was more than large enough for the current production numbers.
There appeared to be more pressure drop between wellhead and separator than expected. A run without any restrictions was done, and this yielded a significant gain. Waxing was ruled out due to the high temperatures in the flow line (> 121°F or 50°C), but sand in the flow line could not be ruled out. This will be followed up using a more accurate flow line model that takes major fitting (including bends) into account.
Also, it was observed that the pressure drop over the flexible flow line of the subsea well to surface was 5 bar higher than expected. This could be due to sand.
Gas lift optimization
Gas lift distribution can often be optimized against a total gas lift volume constraint. Total gas lift available will also depend on gas contracts in place and the amount of associated gas that is produced. When the IPSM was used to optimize the gas lift distribution, an increase in oil production of about 7%, or 6,290 bbl, was predicted. This number was reduced based on previous experience to account for the uncertainty in dynamic behavior of gas lift in very undersaturated oil.
Optimizing gas lift amount
The amount of gas available for gas lift was limited to 28.2 MMcf/d (800 Mcm/d). This limitation is mainly due to the limited compressor capability. The IPSM was used to explore additional gains (in addition to the to the “optimized gas lift” scenario described above) of injecting 35.3 MM, 44.1 MM and 43 MMcf/d (1 MM, 1.25 MM and 1.5 MMcm/d) of gas. According to the model, this yields gains in the order of a few hundred cubic meters per day. This number needs to be verified after the model has been calibrated more extensively. In many cases, optimizing compressor settings is possible by integrating a more accurate process model into the IPSM.
Gas lifting standalone
Well A55 was originally an exploration well and not equipped with a gas lift valve. Retrofitting gas lift during a workover of A55 was considered. Several cases were reviewed: using the currently available amount of gas lift, and the other amounts mentioned above under “Optimizing gas lift amount.” Gains were recorded that were interesting enough to be included in the list of opportunities.
Optimizing subsea network
When the (new) E wells were put onstream in the model, the wellhead pressure of well D2 increased due to backpressure and it would hardly flow (see Figure 1 for location of the E and D wells). This confirmed the suspicion that if D2 is not converted to gas lift soon, it could die and the associated reserves could be lost.
Different separators
The platform wells are generally produced through the first-stage production separator. It has become common practice to produce one of the wells through the test separator when not performing well tests. At most one well can be produced through the test separator. It was investigated which well would be optimally produced through the test separator. Although there were differences, these were marginal, just in excess of 353 Mcf/d (10 Mcm/d), and so this was not included in the list of opportunities.
Subsequent decisions
The Draugen PtL defined unrisked gains of about 530 Mcf/d (150 Mcm/d) from one of the core PSO activities, gas lift optimization, based on optimization of the platform wells plus subsea well A55. This equates to an acceleration of just over 1% of total daily production.
The Draugen team reviewed the results of the IPSM on the PtL and had a framing session concerning the future use of the model. The conclusion was that a project to enhance and maintain the model was acceptable and was expected to give the following benefits:
• Enable consistency and better field understanding through integration of the disciplines in model construction, validation and maintenance;
• Streamline modeling effort through removal of duplication at discipline model boundaries; and
• Deliver accelerated production gains quicker and with reduced effort.
Lessons learned
Some lessons that can be applied elsewhere:
• IPSM allows full integration of decision-making between the disciplines and rapid estimation of opportunities;
• The use of an IPSM forces one to use a consistent PVT data set across the system: in the tanks, the well models and (where applicable) the process models;
• The tank models may need to be divided into more tanks to account for the differences in PVT in the reservoirs. They can be calibrated using individual well production data. This also allows a transparent way of dividing injection water among the different tanks in the system during the calibration;
• Gas lift distribution optimization will include both the benefits of gas lift to the wells and also the effect of natural gas production against a total lift gas compression limitation;
• Although reservoir simulator output can be helpful, history-match material balance models with actual production data first (since this helps in fundamental understanding of the reservoir and speeds up the simulation);
• A good understanding of relative permeability effects is important to basic predictions and to the estimation of inflow impairment;
• Comparing pressure drops in flow lines with the model predictions can help identify possible constrictions;
• Go through the model with the production operators and get input from them. This will provide a reality check for the model, identify areas for improvement and also provide them insight into the interaction of different elements under different conditions; and
• The topsides model should accurately predict how the system elements interact and where the system is truly constrained. Consider integrating a detailed process model of the topsides to improve these predictions.
Email
Print.jpg)
