Optimizing the Azeri Field asset

Using an advisory software package, subsystems were integrated into a model that optimized the facility to create a 3% production increase.

Nigar Jalilova and Abekir Tautiyev, BP; Mike Strathman and Sergi Sama, Aspen Technology

Increasing oil and gas production is a key business driver. As a result, companies are spending more time optimizing what they have already found. BP has achieved good results using sustainable best practices and process optimization solutions.

AZERI ASSET

The Azeri oil and gas production asset is within the top five BP investments worldwide and is possibly one of the world’s most complex oil systems. Present offshore facilities consist of three production platforms and a gas processing platform, located in the Caspian Sea. Each production platform has oil/water separation, gas compression and gas dehydration. The gas processing platform has four parallel compression trains driven by 25-MW gas turbines, which supply high-pressure gas for reservoir pressure maintenance and gas lift for oil production. The onshore terminal has four oil-stabilization trains, oil and gas receiving facilities and a gas plant for quality control.

SUCCESS FACTORS

The importance of addressing people, process and technology issues to ensure successful adoption of new technologies has been widely documented. Taking a simplistic approach and focusing on only one or two of the three areas has been identified as the single most important reason for failed technology adoption initiatives. In the case of oil and gas producing assets, addressing these areas is even more important, but there are hurdles associated with it:

• People: Often the adoption of new technology can be perceived by levels other than management as extra work or even as a job security threat. Additionally, new technology adoption may point out ways to operate an asset better, opening the door for blaming staff for not having found those better ways earlier. Therefore, communication is key to win support.
• Processes: How people work will change, since business and work processes are designed with a certain framework in mind. Technological changes may deem existing processes inadequate or obsolete. The plan to transition from “as-is” to the new workflows is a critical step.
• Technology: In field optimization, successful implementation requires the synergistic collaboration of all the technologies used.

These considerations make implementation methodology a key art and science when introducing new technology to the field.

ADVISORY SYSTEM

An offline advisory system, Azeri Field Optimizer (AFO), was conceived to determine the optimum values of key process settings in a number of business scenarios. Scenarios ranged from asset revenue maximization to maximization of injected gas, maximization of exported gas and others.

The system integrates disciplines, software and workflow. Key components include:

1. Well modeling software (Prosper): Production engineers can batch-run individual well simulation models to create simplified pressure-flow relationships for each individual well. Performance data can be later used for optimization.

2. Facilities simulation (Aspen Hysys): Process engineers use the well model results combined with the process models for the platforms and onshore facilities in a process simulation environment.

3. User interface (AFO Interface coded in Excel): This integration layer handles manual inputs for the optimization, retrieval of field data variables as inputs for model validation and calibration, data conditioning of field variables and aggregation of real-time data from the historian (Aspen IP.21) and the well test database (Microsoft SQL).

The field optimizer operates in several modes based on the business/technical needs for model initialization, mass balance, model calibration and optimization.

MULTIPLE OBJECTIVES APPROACH

The heart of the advisory system is a multi-purpose high-fidelity model with auto-calibration mechanisms. The advantage of this approach is that the same model that is tuned to plant conditions, can be extrapolated to new operating conditions, The tuned model can monitor the plant equipment performance, while performing offline studies for potential upgrading schemes, studies for different feed/economic conditions and optimization studies. This is a leap forward, since most optimization tools hard-code the configuration. In addition, optimization scenarios can be set to view the asset’s entire performance or can be restricted to certain subsystems, providing workflow aligned with operating decision making.

Strict rules have been applied during model construction in the hierarchy of systems and subsystems, so that activation/deactivation of model parts can be cascaded down to individual equipment.

WORKFLOW

The advisory system has a centralized graphical environment to collect all data required for initializing and calibrating the asset model from data historians. It interfaces with well models, depicts wellhead hydraulic behavior, handles the information traffic to and from the model and executes optimization runs. It is designed to help the end-user:

• Control the information transfer between different system components
• Perform plant data validation and pre-processing
• Calibrate well models against well test data retrieved from well test databases
• Execute the model/optimizer in its various modes.

DATA VALIDATION

Field data tends to be scarce, and when available is sometimes of poor quality (e.g., multi-phase flow with frequent disruptions). It is therefore essential that the optimization system has robust, reliable and intuitive data validation mechanisms.

All field instruments have a fixed error (offset) and random error. This is especially true for flowmeters, which rely on an assumed fluid density that will almost certainly be different from the actual fluid density. The instrument offset can be estimated from present plant performance and is assumed to remain the same over a fairly large time range.

The gross-error detection engines check data against its corresponding high- or low-scale instrument-calibrated range and sounds an alarm when values are out of range, Fig. 1. Bad data must be detected and replaced with alternate data to prevent the model from stopping or becoming misleading. This alternate data may come from an alternate sensor (i.e., from redundant sensor installations) or from a calculation.

Fig. 1. Gross-error detection checks data against the instrument-calibrated range.

Ad hoc data-quality analysis engines are placed around critical equipment. These components monitor the consistency of a field measurement group and can pinpoint suspicious data points where the gross-error detection engines saw no problem or instrument failure. For instance, suction flowrates in parallel compression trains shall be consistent with the corresponding train status indicators, as well as with the power consumption readings, discharge pressure readings, etc.

WELL MODEL TUNING

Wells are simulated with the help of individual well models. These models are updated and calibrated regularly to reflect actual well performance. When well tests are available, these can be compared graphically with well model predictions and, if deviations are found to be acceptable, the wellhead performance relationship can be adjusted to match well test data (watercut, gas/oil loading, GOR and main flow rate volumes), Fig. 2.

Fig. 2. The tuned model can monitor equipment performance and perform offline and optimization studies.

CALIBRATION

A best practice in optimization is calibrating the model to actual performance. Any model of real equipment has an intrinsic error from the assumptions and simplifications embedded in the model, since certain real behaviors cannot be modeled in a form suitable for optimization (e.g., dynamic wax deposition in heavy oil pipelines). Calibration is a least-squares minimization problem that can be solved asset-wide, using the same mathematical algorithm employed for optimization. This avoids additional side models and model configurations, and facilitates model organization and maintenance.

OPTIMIZATION

Optimization can have many meanings in a process modeling environment. With present market prices, most operators are interpreting optimization as maximizing the current production rate. This is the core objective for this project. To that end, there are a number of recurring aspects:

• Process decision variables: A typical optimization problem will contain “obvious” true decision variables and “indirect” decision variables. An obvious decision variable is the pressure set point of a high-pressure oil separator. This is a controllable variable whose set point can be adjusted to an optimizer’s values. An indirect decision variable may be the speed of a centrifugal compressor. In most cases, this will be a controller-manipulated variable. Indirect decision variables are sometimes a convenient or efficient way of solving the mathematical problem. Indirect variables should be carefully analyzed, since their manipulation may require changing the configuration of corresponding process controllers.
• Constraints: These are as important as decision variables. Some constraints are easily identifiable from operating-personnel experience and/or from the set points of field alarms. Others are “fuzzy” constraints (e.g., the capacity of an oil/gas separator, the maximum allowed flowrate in a gas pipeline) whose values are established based on existing design and/or best operating practices. Sometimes fuzzy constraints are not real hard constraints, and can be challenged. Significant gains can sometimes be achieved in a step-wise manner.
• Objective function: Sometimes the objective function’s mathematical formulation is coincident with the business objective; for others, it is a combination of process economics and a number of weighted (penalized) key process constraints re-cast inside the objective function. A proper objective function formulation requires extensive system testing using a suite of relevant business scenarios covering the process situations that the algorithm will face.

Optimization provides brand new ideas about how to operate or revamp an existing process, instead of simply repeating a schema that has been used before. It provides an engineering baseline to benchmark a process engineer’s estimates, which often are strongly dependent on his or her particular experience and preferences. Optimization leads to rationalization of engineering efforts, time and engineering costs for competitive plants.

BEYOND OPTIMIZATION

Integration of process engineering with business processes is vital for E&P companies seeking to optimize recovery from existing assets. It provides greater transparency and accuracy in decision making. In addition to creating the data integrity and models for analytics, there are further benefits.

Equipment condition monitoring is a byproduct of model calibration. The economic penalty ($/day) of using new parameters can be traced and analyzed each time the model calibration system updates equation coefficients (tuning factors) to reflect changes in equipment performance from fouling or degradation. Plant engineers and management may use this information to schedule maintenance cycles and/or equipment replacement.

What-if studies can be used to evaluate the impact of operating asset equipment at significantly different operating conditions. These differences can range from changes in input data to different model parameters to different equipment. Examples include studies to evaluate major changes on specific set points, the impact of starting idled equipment or stopping some currently operating equipment and constraint sensitivity analysis (the value of eliminating a constraint). The initial data for a what-if study could come from a plant data tuned system, an optimized system or a previously saved study.

RESULTS

Producing and calibrating a detailed model yields a major benefit. It shows which instruments and equipment are working reliably, and identifies areas that require attention or adjustment.

The Azeri project identified some meters that showed incorrect readings. Some of these readings were active optimization variables, so their accuracy directly affected plant economics. For example, ambient temperature needs to be correctly specified, since gas turbine power output is affected by it. High ambient air temperature results in reduced power output when temperature exceeds 22ºC for about three months each year. During this period, capacity is reduced. For the remainder of the year, gas compression capacity can be maintained and may even exceed design capacity.

The Azeri system exhibits a rather complex network of pressure flow interactions. The gas and oil compression, pumping and transport systems are tightly coupled. In the Sangachal area, high climate condition variability during the year plays an important role in the performance (and capacity) of the gas conditioning and oil stabilization facilities. Capacity constraints can be observed anywhere from the wellheads to the terminal delivery points. The business requirements of a particular day and/or equipment availability strongly affect the operating philosophy followed to best use the facilities.

In most cases, the facilities’ ability to use associated gas by injecting it into the reservoir or transporting it onshore and dew-pointing it to pipeline specifications for inland distribution, constrains maximum oil production. Four factors influence that ability and therefore affect oil handling capacity:

• Gas injection compressor performance: This includes the number of trains in service and performance of the gas turbine drivers that are affected by ambient conditions.
• Gas dehydration column pressure setting on the compression platform: The column receives all produced gas streams. The lower pressure, the higher the drawdown from the wells. But lower compression platform capacity to the Sangachal gas export pipeline and lower discharge pressure of the gas injection compressor result in fewer volumes being delivered. This trade-off is among the most important factors determining maximum production quotas.
• Propane refrigeration circuits’ performance: The propane condensers installed in the Sangachal Gas Dew Pointing Unit are seriously affected by ambient conditions.
• Low-pressure separator’s pressure setting on production platforms: This affects the flash gas generated and hence the capacity utilization of the highly loaded gas pipeline to the compression platform. Controlling oil stabilization can optimize available transport capacity.

In a typical case, where the optimizer was run to predict the maximum attainable oil production for a particular day of interest, it was found that the pressure set points could be optimized to increase production an average of 3%.

Use of the advisory system has helped Azeri Field asset managers better understand the overall facilities’ performance and propose refined operating strategies. The optimizer has been used for a number of different situations such as platform shutdown, maximizing exported or injected gas to meet specific business requirements and varying ambient conditions. In all cases, it has provided sound answers and revealed interesting behavioral aspects that can continue to be explored.    

ACKNOWLEDGEMENT

This article was derived from a paper presented at the SPE Gulf Coast Section 2008 Digital Energy Conference and Exhibition held in Houston, Texas, May 20-21, 2008.