Real-time process optimization aids oil production operation

Glyn Westlake
MDC Technology Ltd.
Middlesbrough, U.K.

Wytch Farm, Dorset U.K., provides an example of how real-time process optimization (RTO) techniques benefit oil and gas production operations.

The application of RTO to petrochemicals and refinery facilities has a long and well-documented history for maximizing returns from assets.^{1 2}

Process optimization is a widely used tool throughout the process industry and essentially is applied in one of the following three forms (Fig. 1 [111,888 bytes]):

1. Off-line optimization for providing information for planning studies
2. Open-loop real-time optimization for advising operators
3. Closed-loop real-time optimization for supervisory process control.

Of these, the oil and gas production environment often uses off-line optimization as part of the process for scheduling production. Real-time optimization, both open and closed-loop, are less common, although both systems are in use. ^{2 3}

Solving problems

Process optimization systems solve the problem of choosing values for a group of facility set points that will maximize the economic benefit of an asset, while being within all operating constraints.

Off-line systems can make a valuable contribution in determining optimal operating conditions based upon one particular operating scenario or, as is often the case, upon a suitably averaged case. The solutions proposed by off-line systems are, however, limited in their applicability by the normal operational variability of the production process.

Production facilities are in a continual state of change because of changes in well availability and performance, for example.

These changes will, in turn, change the problem, which the optimizer needs to solve. Therefore, the optimal combination of facility set points will change. In essence, the purpose of RTO is to maintain the facility at the optimal set of conditions as circumstances change.

The Wytch Farm RTO system uses MDC Technology's RTO+ package. It is an example of an open-loop system. In this case, a closed-loop solution would not be possible because set points for the process variables considered by the optimizer must be set manually on local controllers in the field. If changes occur that require the generation of a new optimal point, a member of the operations team has to manually investigate.

Optimization problem

Wytch Farm is the largest onshore oil field in Western Europe, consisting of about 60 producing wells, spread over nine well sites.

The Wytch Farm oil field is operated by BP Exploration Operating Co. Ltd. on behalf of its partners ARCO British Ltd., Premier Oil plc, Onepm Ltd., Kerr McGee Oil (UK) plc, and Talisman North Sea Ltd.

Well fluids from all well sites go through either a 10-in. or 12-in. flow line to a gathering station where the fluids are separated into oil, gas, water, and LPGs (Fig. 2 [78,543 bytes]). A separate flow line is used for well testing.

After the gathering station, the crude and gas enter pipelines, and the LPGs are exported in railroad cars. Produced water is reinjected. Wytch Farm produces about 100,000 bo/d.

The primary separator pressure at the gathering station is controlled with the well site pressures floating, depending on the pressure drop, and therefore flow through the two flow lines. The wells are produced either with beam pumps or electric submersible pumps (ESPs). In most cases, the ESPs can be run at a variable frequency.

Production from the (fixed displacement) beam-pumped wells is essentially at a constant rate; however, the situation for the ESP wells is far more complicated.

Assuming an inflow performance relationship of Q = PI x DP, the well production (Q) is determined by reservoir pressure, well productivity index (PI), and bottom hole flowing pressure. The DP is the pressure drawdown (reservoir pressure minus bottom hole flowing pressure).

Of these, the reservoir pressure and productivity index are essentially fixed in the short term. The bottom hole flow rate, and with it the bottom hole flowing pressure, however, can change continuously because of changes in the head developed across the pump, the pressure drop across the choke, and the well site header pressure.

The relationships between the various parameters affecting well performance are well known for each well in isolation. What is less easy to determine is the overall effect of changes in operating parameters on the asset as a whole. This depends on:

• Change in production on the well concerned
• Site on which the well is situated
• Current production from all other wells
• Well fluid water cut and hydrocarbon composition
• Flow line pressure at the gathering station
• Well site tie-in to the flow lines.

For example, consider two wells at different well sites. For an identical increase in well fluid production, the well furthest from the gathering station, where the pressure is controlled, will cause a greater pressure drop in the flow line because the additional well fluid flows through a longer line (Fig. 3 [76,073 bytes]).

However, this increase in pressure drop will be mitigated as production from other ESP wells is backed out because of the increased pressure at the well sites.

Because of this complex combination, an optimizer is needed to determine the optimal operating conditions. The optimizer will recommend:

• Wells to be on production
• ESP frequency
• Flow line to which the well sites should be connected.

This is a mixed-integer type of optimization problem because it combines both continuously variable quantities (ESP frequencies) and discrete quantities (well on/off status and flow line connections).

Optimizer components

The optimization system consists of several discrete parts that work together to generate the optimal operating point. Many of these are relevant to the real-time aspects of the system.

Communications interface

The communications interface handles all communication both with the facility and between the various optimizer processes.

In the Wytch Farm case, data are obtained from a facility MIS (management information system), although it is also possible for data to be obtained directly from the DCS (distributed control system).

Data conditioning, validation

The data conditioning and validation component is responsible for the identification of obviously bad measurements, such as out of range, and the scaling/filtering of data from the facility.

In the Wytch Farm system, the functionality is performed as part of the communications interface tasks.

Data reconciliation

Data reconciliation uses the system process models to test for consistency between input data points. By using process models, inconsistencies can be identified in sets of data points that are all within expected limits but are not consistent as a whole.

The reconciliation process applies biases for minimizing the inherent errors in the measured values. This methodology is applied across well sites, taking into account the well site pressures, ESP frequencies, and bottom hole flow and pressure measurements, where available.

Model performance updating

The model performance is updated by adjusting various model performance parameters to ensure that the process model used by the optimizer remains consistent with actual facility performance.

To avoid biasing the model based upon process noise, these calculations use sets of data collected over a period of time. In the Wytch Farm case, the key pieces of equipment are the wells and the flow lines. In both cases, the unit performance parameters are updated in this manner.

Process model

The process model is the core of the optimization system. This includes representation of all the wells, flow lines, and process equipment considered by the optimizer.

The model is used both by the optimizer and in isolation to establish a "base point" that is used as the start point for optimization.

A key aspect of generating the base point is to determine correct values for the constraints on the problem. This is required to account for differences between measured and modeled values and to estimate values for which no specific measurement exists.

In Wytch Farm, constraint updating/estimation is applied to the well bottom hole pressures (for reservoir management considerations) and the maximum motor current for the ESP pumps.

Optimization

The optimizer performs multiple runs of the process model evaluating the effect of changing the various set points. Based upon the model results, it first identifies a point that satisfies all the facility constraints and then determines the "best" solution within this feasible region.

Process executive

The process executive component is the "master of ceremonies" that coordinates the running and interaction of the other processes. The various processes operate in an asynchronous manner, exchanging data with each other via the communications interface, as shown in Fig. 4 [73,923 bytes].

RTO applicability

In looking at RTO applicability to a process optimization problem, one must review three basic criteria:

1. Problem complexity
2. Available degrees of freedom
3. Amount of change in the problem.

Problem complexity relates to the relative abilities of operators and the optimization system to determine the best set of operating conditions.

For an optimizer to be beneficial, the problem posed must not be so simple that the optimal operating conditions are obvious. Conversely, it is important that the optimizer represent the process with sufficient accuracy so that the answers generated represent optimal conditions on the facility.

In the case of the Wytch Farm optimizer, the problem solved by the optimizer was clearly one that lends itself to optimization.

Manually obtaining a solution to the optimization problem would theoretically be possible, but would take an unreasonable amount of time because of the interactions of the various facility set points.

Even to perform a single simulation of the facility would require an iterative approach to ensure consistency between all the system flows and pressures.

The strength of the optimizer lies in its ability to consider the wells and flow lines as a complete entity rather than as individual units. Because of this, the system is able to evaluate competing potential operating changes and determine which are the best for the asset as a whole.

The degrees of freedom available affect both the feasibility of RTO in general and the suitability of open and closed-loop applications. The larger the number of set points available for manipulation, the greater the scope for an optimizer to make process improvements. However, in an open-loop RTO application, as the number of recommended set point changes increase, an increased workload is placed on the operators.

This is important because the optimizer solutions are only valid in their entirety, and partial implementation can lead, at best, to suboptimal operation.

For Wytch Farm, there was never any danger of having insufficient degrees of freedom to make a worthwhile problem. In fact the difficulty of the problem is predominantly because of the interactions between the system inputs.

Of more concern, however, was the possibility of a large number of process changes being recommended for only a very small marginal gain. For this reason a 'pareto' strategy was taken.

The amount by which the problem changes, and the frequency with which these changes occur affects decisions as to whether real-time optimization is required or whether on-off/off-line calculations will suffice to determine the optimal facility set points.

Put simply, if the problem does not change, the optimal set point will not change and therefore, there is no need to do recalculations.

Within the production environment, however, it is very unlikely that there will be no change. In fact, a major challenge involved with applying RTO to production systems is to design the system so that it accounts for the often-substantial changes in the operating scenario.

For Wytch Farm, the changes are, as would be expected, within a production environment and include over the short term:

• Shutdown and start-up individual wells and well sites.
• Changed flow of wells between the test and production flow lines.
• Changed well operation flexibility
• Failure of facility instrumentation.

Over the longer term, changes that can be expected are as follows:

• Decline and intervention in well productivity
• Changed reservoir pressure because of depletion and injection strategy
• Fowled flow line
• Start-up of new wells.

The short-term changes are all accounted for automatically by the optimization system. The status and connections for each well are read in and the model adjusted accordingly, as are the well flexibility (availability for switching on/off and allowable ESP frequency range).

Instrument failure is detected and accounted for by a combination of the data validation and data reconciliation processes. To maintain a high stream factor, it is essential that any RTO system be able to cope when some facility data are bad or unavailable.

This is especially the case in the oil and gas production environment where it is often both technically and logistically difficult to maintain/replace instrumentation.

Over the longer term, changes are handled by a combination of manual and automatic updating/modification. Wells and flow lines are automatically updated with manual entries for fine-tuning. New wells are configured within the process model prior to coming on-line and then initial performance data are entered once production has stabilized.

Many changes, little gain

A potential problem for an open-loop RTO system with many degrees of freedom is that a very large number of set point/status changes are recommended, but many of these are associated with only a marginal benefit.

A key feature of optimizer solutions is that they are valid only when implemented in full. It is important, therefore, that in an open-loop application, the optimizer does not recommend a large number of process changes that have only a very marginal benefit.

If this happens, operators may start to make their own decisions as to the important changes and ignore the rest, potentially causing suboptimal operation.

For Wytch Farm, the reduction of the number of recommended process changes that caused only a small improvement was especially important for two reasons:

1. Changes can only be implemented locally on the well sites and, therefore, require an operator to be dispatched to make the changes.
2. A disincentive for taking wells on and off-line because of increased wear and tear and the time needed for production to stabilize.

To reduce the possibility of too-frequent changes, penalties are applied to the changes made by the optimizer. A penalty is applied for each recommended process change, independent of the magnitude of the change.

The size of the penalty reflects the cost of making the change and, in the case of the well status, reflects the lost production while the well is brought on-line. This ensures that recommended changes must be associated with a significant improvement in the operating point.

Off-line optimization

A major hurdle to the increased use of off-line optimization techniques by production/process engineers is the amount of time required to obtain any useful answers.

To perform a meaningful off-line optimization run, one must first perform a number of steps, as follows:

• Obtain current facility operating data.
• Run a simulation based on current operating data, without optimization.
• Account for, and correct for, the differences between the model results and the facility. These differences are typically caused by changes in equipment performance or equipment degradation. Typically, this is an iterative procedure involving repeated runs of the process simulation.

As an RTO system must keep itself up-to-date with the current facility data and equipment performance, it can be used to provide a starting point for off-line investigations that represent the as-is state of the system. This facilitates the use of the optimizer to perform off-line studies.

This aspect of the optimizer is now being widely used at Wytch Farm. In particular, the optimizer has been used to examine the potential benefit of a number of well interventions to determine the overall benefit to the asset as a whole. In some cases, the optimizer has identified that the yield from the well concerned would increase after a proposed intervention (as expected), but that the extra production was offset by the effects of backing-off other wells.

The net effect was such that there was no net increase in oil export from the site. The benefit was not so much in what the optimizer indicated was right but in what it indicated was not worth doing at all.

Use of the optimizer to evaluate the effect of potential well modifications has now been included in the evaluation/justification procedure for future well modifications.

System benefits

The RTO+ optimizer at Wytch Farm is recognized as a very successful application that has produced benefits of about 3% as a result of the daily runs.

This has been instrumental in justifying well interventions resulting in about 4% increased production and a saving of £150,000 by avoiding expenditure on well interventions that would not have been beneficial.

References

1. Dewar, I., Lopez, S., and Brewer, M., Closed Loop Optimization of Olefins Facility.
2. Mortimer, A., and Purves, B., Tern Optimizer Project-The World's First Offshore Closed Loop Optimizer.
3. Witherwick, D., and Dormer, A., Gas Facility Optimization.

The Author