How to Optimize Gas-Oil Separation Plant Parameters to Maximize Oil Recovery

Abstract

To provide a method and system for improving output, efficiency, throughput, and stability of a GOSP process.SOLUTION: The method and system for optimizing a gas oil separation plant (GOSP) operating parameter comprises maximizing a stabilized crude oil recovery rate in an integrated framework for GOSP operation. Daily (day and night) and seasonal (summer and winter) ambient temperature changes are incorporated to show effect thereof on liquid recovery. An artificial intelligence model predicts the stabilized crude oil recovery rate over a range of ambient temperatures. Optimum GOSP separator pressure for maximum stabilized crude oil recovery is determined for each ambient temperature. GOSP separator pressure may be adjusted to achieve the maximum stabilized crude oil recovery under varying ambient temperatures.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] STATEMENT REGARDING PRIOR DISCLOSURE BY INVENTOR Aspects of this technology were presented at the APCChE 2019 conference in Sapporo, Japan on September 25, 2019 entitled "Novel Approach to Optimize GOSP using Artificial Intelligence", at the ISA Saudi Arabia Section Eastern Region Event 2019 on October 20, 2019 entitled "GOSP Optimization Using Artificial Intelligence", and at the International Petroleum Technology Conference on January 13, 2020 entitled "Decision Support System for Optimizing GOSP Operation", each of which is incorporated by reference in its entirety herein.

[0002] (Verification statement) The inventors acknowledge the support provided by Yokogawa Saudi Arabia for funding and supporting this work under project CIPR 2328, as well as the support of King Fahd University of Petroleum, Minerals College of Petroleum Engineering and Geoscience, and the Deanship of Scientific Research.

[0003] The present disclosure is directed to a method of operating a gas-oil separation plant to enhance liquid recovery. [Background technology]

[0004] The "Background" statement provided herein is for the purpose of generally presenting the context of the present disclosure. The work of the currently named inventors to the extent described in this Background section, as well as aspects of this statement that may not otherwise be considered prior art at the time of filing, are not admitted expressly or impliedly as prior art against the present invention.

[0005] Gas-Oil Separation Plants (GOSPs) are key oil processing facilities. GOSP facilities must be operated under optimal conditions to maximize production. Current practice is to operate GOSPs according to parameters and under conditions determined at the design stage based on simple laboratory tests. Traditional methods do not consider environmental variations and reservoir conditions that affect operating parameters, thus resulting in reduced liquid recovery and loss of profitability.

[0006] Crude oil extracted from an oil field goes through several operations before it can be exported to end users. Crude oil separation is one of the key processes for separating oil, gas and water. Thus, operation of a gas-oil separation plant (GOSP) under optimal conditions is considered highly desirable by oil operators. To improve plant operation, including the overall operation of the unit, the operating parameters of the individual units and external disturbances, a process that considers all variables and their influence on operation must be used. In reality, implementing optimal parameters and establishing optimal control conditions for all of the connected individual units for a GOSP is difficult in real non-simulated operation.

[0007] In general, due to the fundamental governing principles of different unit operation, optimization of each individual unit has been found to be the simplest way to improve the operation of a Gas-Oil Separation Plant (GOSP). Traditionally, GOSP improvements have been performed by optimizing units that operate with similar governing principles. Among them is the optimization of separator condition, size, and stage to maximize oil quantity and quality. Reservoir-related performance parameters such as oil API, formation volume factor (Bo) and gas-oil ratio (GOR) are included in the objective function used for optimization. (The following are each incorporated by reference herein: Mahmoud M. Tariq M. Kamal M. and Al-Naser M. “Intelligent Prediction of Optimum Separation Parameters in the Multistage Oil Production Facilities” Journal of Petroleum Exploration and Production Technology, 2019, https: / / doi.org / 10.1007 / s13202-019-0698-6; Edwin M., Abdulsalam S. and Muhammad I. “Process Simulation and Optimization of Crude Oil Stabilization Scheme Using Aspen-HYSYS Software”, International Journal of Recent Trends in Engineering & Research (IJRTER) Volume 03, Issue 05; May-2017 [ISSN: 2455-1457]; Elshafei M. and Doklah M., “Method for Optimizing and Controlling Pressure in Gas-Oil Separation Plants”, U.S. Patent #8,805,587, Aug. 2014.; Muhammad F., “Simulation of Crude Oil Stabilization”, B. Eng Project, Chemical Engineering Department. Universiti Teknologi PETRONAS, Malaysia, 2013; Ling, K., Wu, X., Guo, B., and He, J. 2013. “New Method to Estimate Surface-Separator Optimum Operating Pressures. Oil and Gas Facilities”, Journal 2(3): 65-76. Doi:10.2118 / 163111-PA; and K. Amminudin, “Yield and Energy Optimization Initiatives at Saudi Aramco Khurais Central Processing Facility”, SPE 138526, Abu Dhabi International Petroleum Exhibition & Conference held in Abu Dhabi, UAS, 1-4 November 2010 (see).

[0008] FIG. 1 shows a simplified process diagram of a typical GOSP. Crude oil (wet crude) extracted from an oil field or hydrocarbon reservoir is first processed in a gas-oil separation plant 100 where stabilized crude oil 110 is produced through a series of separation systems including a high pressure separator or high pressure production trap (HPPT) 112, a low pressure separator or low pressure production trap (LPPT) 116, and a stabilizer 102. Gas and condensate are collected in a gas train 104 and processed as acid gas 106 and natural gas liquids 108 for export. (Amminudin K. and (See Muaibed M. “Optimization Yields Gains for Saudi Crude Processing Giant” Oil and Gas Journal, December 2014.) Acid gas is natural gas or other gas that contains significant amounts of hydrogen sulfide (H2S). Natural gas is typically considered acidic if there is more than 5.7 milligrams of H2S per cubic meter of natural gas, which is equivalent to about 4 ppm by volume at standard temperature and pressure.

[0009] GOSP plants must efficiently process the inlet feed stream (wet crude oil) to meet the specifications for oil and gas production as driven by sales or production contracts, as well as the targets for produced water injection or disposal requirements. While the volume or outlet specifications are fixed, the inlet feed composition changes over time. Additionally, reservoir pressure and temperature changes over time affect the quality of the feed stream as well as the oil / water / gas ratio for the GOSP plant. These changes must be considered when operating a GOSP at high capacity and efficiency. Generally, GOSP optimization is performed based on the quantity and quality of the feed stream.

[0010] Gas-to-oil ratio (GOR), volume factor (Bo) and oil density (API) are traditionally considered as performance parameters for GOSP operations. The objective of GOSP optimization is to achieve minimum GOR and volume factor and maximum API. (Arnold, K.; Stewart, M. “Surface Production Operations (2 nd (See “Design of Oil-Handling Systems and Facilities”, International Conference on Petroleum Exploration and Research (ICER), 2003, Vol. 12, No. 1, 2003, Edition) Volume 1-Design of Oil-Handling Systems and Facilities, Elsevier 1999. However, the combined effect of these parameters is reflected by the liquids (oil) recovery. Thus, maximum liquids recovery from a GOSP plant results in low GOR, low volumetric coefficient (Bo) and low oil density (high API). Hence, optimization should aim at maximizing liquids oil recovery.

[0011] During the gas-oil separation process, the traditional goal is to liberate light components such as methane (C1) and ethane (C2) to form a separated gas stream while retaining the heavier components in the liquid stream, and thus the preferred gas-oil separation method is differential liberation, which retains the maximum amount of heavy and intermediate components in the liquid stream. The gas released earlier at higher pressure is not present at lower pressure, causing the intermediate and heavy components to co-evaporate or aerosolize. (See Clark, N., "Elements of Petroleum Reservoirs", Dallas: Society of Petroleum Engineers, 1960; Ahmed, TH, "Reservoir Engineering Handbook", Elsevier Inc., 2006, each of which is incorporated herein by reference). However, implementing differential liberation is costly and generally not feasible as a field operation.

[0012] The inlet feed stream (wet crude oil in FIG. 1) passes through a series of stage separation processes (HPPT 112, LPPT 116, stabilizer 102). The number of stages depends on the feed stream composition, pressure and temperature of the fluid stream, and the operating conditions of each individual operating unit. Crude oil processing also requires other processes such as heater treaters, desalination, and stabilization to meet demand requirements and to make it suitable for storage and transportation. Thus, liquid recovery from the GOSP depends on parameters such as the number of separation stages, inlet feed composition, and pressure and temperature of the individual operating units within the GOSP plant.

[0013] These processes and parameters are governed by different physical principles, but they are integrated to achieve target specifications. The performance of each individual unit is dependent on the other units, and therefore, an integrated process across multiple units is affected by changes in the operation of each individual unit. The different governing principles of units operating in an integrated network make GOSP optimization very complex and difficult to achieve and implement in field operations.

[0014] As mentioned above, improving the unit operation of individual units has traditionally been considered the simplest way to optimize an integrated GOSP plant. Previous studies have focused on determining the optimal operating conditions for the separation process. Al-Jawad and Hassan (2010) proposed an empirical approach to estimate the operating pressure for two-stage and three-stage separation processes. (See Al-Jawad M. and Hassan O., “Correlating optimum stage pressure for sequential separator systems”, SPE Projects, Facilities & Construction 5(1):13-16., 2010. https: / / doi.org / 10.2118 / 118225-PA, incorporated herein by reference in its entirety.) Also, Ling, et al. (2013) provided an approximation to determine the optimal separator pressure based on the wellhead pressure. (See Ling K., Wu X., Guo B., and He J., “New method to estimate surface-separator optimum operating pressures”, Oil Gas Facility 2(3):65-76, 2013, incorporated herein by reference in its entirety).Methods for determining optimal separator conditions based on summer and winter flash equilibria and simulation studies for optimizing separation conditions have been published, as well as artificial intelligence approaches for estimating the number of stages and operating pressure and temperature conditions in multistage separation processes (see Kylling O. “Optimizing separator pressure in a multistage crude oil production plant” MS Thesis, Norwegian University of Technology, 2009; Mahmoud M. Tariq M. Kamal M., and Al-Naser M. “Intelligent Prediction of Optimum Separation Parameters in the Multistage Crude Oil Production Facilities” Journal of Petroleum Exploration and Production Technology, https: / / doi.org / 10.1007 / s13202-019-0698-6, each of which is incorporated herein by reference). However, these publications focus on the separation unit and ignore other processes, such as stabilization, where efficiency improvements are essential to meet crude oil specifications for storage and transportation.

[0015] Other research has been directed at the crude oil stabilization process: for example, simulations have been developed to optimize stabilization plants, crude oil stabilization and condensate stabilization to optimize operating conditions. (See Edwin M., Abdulsalam S. and Muhammad I., “Process Simulation and Optimization of Crude Oil Stabilization Scheme Using Aspen-HYSYS Software”, International Journal of Recent Trends in Engineering & Research (IJRTER) Volume 03, Issue 05; May-2017 [ISSN: 2455-1457]; Muhammad F., “Simulation of Crude Oil Stabilization”, B. Eng project, Chemical Engineering Department. Universiti Teknologi PETRONAS, Malaysia, 2013; Rahmanian, N., Bin I., and Nasrifar, K., “Process Simulation and Assessment of a Back-up Condensate Stabilization Unit”, Journal of Natural Gas Science and Engineering, Vol. 26, Issue Pp. 730-736, 2015, all of which are incorporated by reference herein).

[0016] However, no methodology or strategy has been developed to improve unit operations in an integrated GOSP across all components of a complete oil train. Optimization of individual components of GOSP units has been achieved, but is difficult to implement in the field because all components are interconnected and establishing optimal control conditions for each individual unit is difficult to implement. Summary of the Invention [Problem to be solved by the invention]

[0017] It is therefore an object of the present disclosure to provide methods and systems for improving the output, efficiency, throughput, and stability of a GOSP process, including stabilizer unit operations that apply an integrated approach and identify an operating pressure that results in maximum liquid recovery for each ambient temperature. A relationship between optimal air pressure and ambient temperature change between summer and winter is determined to achieve maximum product recovery. [Means for solving the problem]

[0018] In an exemplary embodiment, a method for maximizing stabilized crude oil recovery in a gas-oil separation plant (GOSP) is described, the method including: constructing, by a computing unit including a dynamic simulator, a process model of the GOSP, the process model including at least a plurality of separators, a stabilizer, and a storage tank of the GOSP; simulating stabilized crude oil recovery of the GOSP under variable operating pressures of at least one of the plurality of separators for different ambient temperatures surrounding the stabilizer to generate a simulated stabilized crude oil recovery dataset; and calculating, by the computing unit, a simulated stabilized crude oil recovery data set. The method includes normalizing the recovery dataset to generate a normalized stabilized crude oil recovery dataset; training, by a computing unit, an artificial intelligence unit on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained artificial intelligence model; inputting, by a computing unit, a second portion of the normalized stabilized crude oil recovery dataset to the trained artificial intelligence model; predicting stabilized crude oil recovery for each separator pressure at each ambient temperature; determining, by the computing unit, an optimal pressure that maximizes the stabilized crude oil recovery for each ambient temperature; and adjusting, by the computing unit, the operating pressures of a plurality of separators of the GOSP to compensate for ambient temperature conditions.

[0019] In another exemplary embodiment, a system for maximizing stabilized crude oil recovery is described, the system including a gas-oil separation plant (GOSP), the GOSP including the following components: a plurality of separators, a stabilizer, a storage tank configured to hold the stabilized crude oil, and a computing unit operatively connected to the components, the computing unit including a computer readable medium including program instructions, the program instructions causing the processing circuitry to build a process model of the GOSP using a dynamic simulator, the process model including at least the plurality of separators, the stabilizer, and the storage tank, and to simulate crude oil recovery of the GOSP under variable operating pressures of at least one of the plurality of separators for different ambient temperatures surrounding the stabilizer, and to generate a simulated stabilized crude oil recovery. the computing unit to generate a simulated stabilized crude oil recovery dataset, normalizing, by a normalization unit of the computing unit, the simulated stabilized crude oil recovery dataset to generate a normalized stabilized crude oil recovery dataset, training, by the computing unit, an artificial intelligence unit on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained artificial intelligence model, inputting, by the computing unit, a second portion of the normalized stabilized crude oil recovery dataset to the trained artificial intelligence model to predict oil recovery for each separator pressure at each ambient temperature, determining, by the normalization unit of the computing unit, a pressure that results in a maximum stabilized crude oil recovery for each ambient temperature, and adjusting, by the computing unit, operating pressures of the multiple separators of the GOSP to compensate for the ambient temperature condition.

[0020] In another exemplary embodiment, a non-transitory computer readable medium having instructions stored thereon, the instructions when executed by one or more processors, causes a method for maximizing stabilized oil recovery in a gas-oil separation plant (GOSP), the method including: constructing, by a computing unit including a dynamic simulator, a process model of the GOSP, the process model including at least a plurality of separators, a stabilizer, and a storage tank of the GOSP; simulating stabilized oil recovery of the GOSP under variable operating pressures of at least one of the plurality of separators for different ambient temperatures surrounding the stabilizer to generate a simulated stabilized oil recovery dataset; and The method includes: normalizing, by the unit, the simulated stabilized crude oil recovery dataset to generate a normalized stabilized crude oil recovery dataset; training, by the computing unit, an artificial intelligence unit on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained artificial intelligence model; inputting, by the computing unit, a second portion of the normalized stabilized crude oil recovery dataset to the trained artificial intelligence model; predicting, by the computing unit, a stabilized crude oil recovery for each separator pressure at each ambient temperature; determining, by the computing unit, a pressure that maximizes the stabilized crude oil recovery for each ambient temperature; and adjusting, by the computing unit, operating pressures of the multiple separators of the GOSP to compensate for ambient temperature conditions.

[0021] The foregoing general description of the exemplary embodiments and the following detailed description are merely exemplary aspects of the teachings of the present disclosure and are not restrictive.

[0022] A more complete understanding of this disclosure and many of the attendant advantages thereof will be better understood by reference to the following detailed description, when taken in conjunction with the accompanying drawings. [Brief description of the drawings]

[0023] [Figure 1] FIG. 1 illustrates a typical gas-oil separation plant process. [Diagram 2] FIG. 1 illustrates a basic multi-stage separation scheme. [Diagram 3] FIG. 1 illustrates a stabilizer tower having multi-stage separation. [Figure 4] FIG. 1 illustrates an optimization process, including model development, data generation, and an AI model for optimization. [Diagram 5] FIG. 1 illustrates a stabilizer tower with multi-stage separation in combination with HPPT, P1 and LPPT separators. [Figure 6] FIG. 1 illustrates a data normalization process. [Figure 7] 1 is a graph illustrating data fitting for an ANN model. [Figure 8] 1 is a graph illustrating data fitting for an SVM model. [Figure 9] 1 is a graph of pressure as a function of temperature showing confidence levels. [Figure 10] 1 is a graph illustrating oil recovery as a function of separator pressure for different ambient temperatures. [Figure 11] 1 is a graph illustrating oil recovery at optimal separator pressure (winter case). [Figure 12] 1 is a graph illustrating oil recovery at optimal separator pressure (summer case). [Figure 13] 1 is a graph illustrating liquid recovery at optimal separator pressure (156 psi) at 15° C. (winter case). [Figure 14] 1 is a graph illustrating liquid recovery at optimal pressure (174 psi) at 41° C. (summer). [Figure 15A] 1 illustrates an exemplary computing system in accordance with certain embodiments. [Figure 15B]15B is an illustration of a non-limiting example of details of computing hardware used in the computing system of FIG. 15A in accordance with certain embodiments. [Figure 16] FIG. 1 illustrates an exemplary schematic diagram of a data processing system for use within a computing system in accordance with certain embodiments. [Figure 17] FIG. 2 is an exemplary schematic diagram of a processor for use with a computing system in accordance with certain embodiments. [Figure 18] FIG. 2 is an illustration of a non-limiting example of distributed components that may share processing with a controller in accordance with certain embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0024] In the drawings, like reference numerals designate the same or corresponding parts throughout the several views. Furthermore, as used herein, the terms "a," "an," and "the" mean "one or more" unless otherwise specified.

[0025] Additionally, the terms "approximately," "approximate," "about," and similar terms generally refer to a range that includes the specified value within a 20%, 10%, or preferably, 5% margin, and any value therebetween.

[0026] Aspects of the present disclosure are directed to a method for maximizing stabilized crude recovery in a gas-oil separation plant (GOSP), a system for maximizing stabilized crude recovery, and a non-transitory computer-readable medium having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to perform the method for maximizing stabilized crude recovery in a gas-oil separation plant.

[0027] Stabilization of crude oil is necessary to improve the efficiency of the gas-oil separation process. The goal of stabilization is to maximize stock tank recovery while meeting vapor pressure and H2S specifications. Butane significantly increases the vapor pressure of crude oil because the true vapor pressures of i-C4 and n-C4 are 72.6 and 50.1 psi, respectively. Control of butane content controls Reid Vapor Pressure (RVP) and True Vapor Pressure (TVP). One goal of stabilization is to remove C1, C2, and C3 as completely as possible while leaving a maximum amount of C4, C5, and heavier components in the crude oil.

[0028] Reid Vapor Pressure (ASTM D323) is an ASTM test method for determining the vapor pressure of crude oils and refined petroleum products. The Reid Vapor Pressure test indirectly measures the evaporation rate of volatile petroleum solvents.

[0029] Most of the partial gas pressure in crude oil comes from low boiling compounds, especially low molecular weight hydrocarbons such as methane and ethane, which may only be present in small amounts. The goal of stabilization is to remove these low boiling compounds without losing the more valuable higher molecular weight components.

[0030] Adjusting the pentanes and lighter fractions retained in the stock tank liquid as part of the stabilization process can change the API gravity of the crude oil. API gravity is a measure of how a heavier or lighter petroleum liquid compares to water; if its API gravity is above 10, it is lighter and floats on water, and if it is below 10, it is heavy and sinks. Thus, API gravity is the inverse measure of the density of a petroleum liquid relative to the density of water (also called specific gravity). It is used to compare the densities of petroleum liquids. For example, if one petroleum liquid is less dense than another, it has a greater API gravity. Although API gravity is mathematically a dimensionless quantity, it is referred to as being in "degrees". API gravity is graduated in degrees on a hydrometer instrument. The API gravity values ​​of most petroleum liquids fall between 10 and 70 degrees, with 10 degrees being the same density as the density of water.

[0031] The economic value of crude oil is affected by stabilization. First, the liquids can be stored and transported to market more profitably than gas. Second, it is safer when storing light crude oils to minimize gas losses from the light crude oil.

[0032] The ultimate goal of a Gas-Oil Separation Plant (GOSP) is to maximize liquids recovery, capture marketable gases, and remove by-products, such as oil smoke, hydrogen sulfide gas, and other contaminants from the output air stream.

[0033] Liquid recovery depends on: 1) Number of separation stages 2) Pressure during each separation stage 3) Daily and seasonal ambient temperature fluctuations

[0034] Improving the cumulative separation of a GOSP plant is difficult to achieve with traditional methods that involve optimizing each component and / or separation unit separately, without considering their interrelationships. It is essential to consider the complete GOSP plant, rather than considering the separate units in isolation. To do so, the various parameters that affect the process must be considered, the sensitivities of each process must be identified and quantified, and process optimization opportunities must be identified.

[0035] Aspects of the present disclosure include the development of a comprehensive integrated framework for maximizing oil (a mixture of heavier molecular weight hydrocarbons) yield by adjusting LPPT separator pressure in a Gas-Oil Separation Plant (GOSP).

[0036] The crude oil can be stabilized through a series of flash or gas-liquid separators operating at successively lower pressures. Tray towers with reboilers and stripping gas may be used in combination with the separators. Multistage separation is the most commonly used method for oil stabilization. Figure 2 shows a conventional multistage separation process for crude oil stabilization. The pressure of the final stage of separation for the separator vapor is set to atmospheric pressure. Theoretically, the maximum yield of stabilized crude oil 210 is obtained by differential separation corresponding to an infinite number of separation stages (e.g., 212, 214, 216, 218), but in practice more than three stages are rarely used.

[0037] In many plants, the final stage of the separator (218, FIG. 2) is replaced by either a stabilizer column with stripping gas or a reboiler stabilizer column. The reboiler produces stabilized crude oil as heated steam flows up the column, stripping methane, ethane, propane, and butane. Stripping is defined as a physical separation process in which one or more components, preferably having an average molecular weight lighter than the average molecular weight of the mixture being stripped, are removed from a liquid stream by a steam stream. In addition, the stripping gas reduces the partial water pressure of the liquid stream, allowing more water to evaporate from the liquid stream.

[0038] 3 shows a schematic of a stabilizer column tower 302. The stabilizer column comprises a series of trays intended for gas-liquid contact / interaction. The liquid feed 326 from the separator 316 is fed to the top of the stabilizer column 302. As the liquid falls through the tower, it falls from tray to tray (indicated by the horizontal lines in the stabilizer 302), becoming increasingly rich in the heavy components and increasingly lean in the light components.

[0039] The stabilized liquid is cooled in heat exchanger 322 by feed stream 326 and in heat exchanger 328 before exiting for recovery in storage vessels for export. Steam generated by reboiler 324 at the bottom of the column or by stripping gases such as methane, ethane and propane flow through the column while the butanes are retained in the stabilized crude 310. The liquid crude at the bottom of the column is heated by reboiler 324 and pumped to higher level trays to further separate the lighter components and may then condense. The crude exiting the bottom of stabilizer 302 is cooled (322, 328) before being output as stabilized crude 310. The separation provided by stabilizer tower 302 is much more effective than the single stage separation of separators 312, 314 and 316. More crude is produced using the stabilized column because the higher operating pressure of the column results in more crude. Several different configurations can be applied, such as multi-flash stabilization or the use of a stabilization column with feed reflux or refluxed by a dedicated cooling unit.

[0040] At the top of the tower, any intermediate components that leave with the gas are condensed, separated, pumped back into the tower, and sprayed down onto the top tray or one or more trays below. This liquid is called "reflux." In a feed-reflux stabilizer column, a percentage of the heated liquid at the bottom of the column is cooled and pumped back to the top of the column (not shown).

[0041] In one aspect of the disclosure, a computing unit 1500 is connected to sensors at the inputs and outputs of the separators (312, 314, 316), stabilizer 302, heat exchangers (322, 328), to each valve (represented by valve circuit symbols), and to a liquid level sensor (not shown) in the stock tank 330. The computing unit 1500 may be configured to receive data from the sensors, perform optimization according to the methods of the disclosure, adjust pressure in the separators and stabilizers, adjust valves, heat and pressure in the reboiler 324, etc. The sensors may be any of pressure, temperature, flow, O2, H2S, liquid level sensors, and the like to define the operating parameters of the GOSP plant as needed.

[0042] Liquid hydrocarbons (condensate) are brought into the system through an inlet separator and preheated in a stabilizer feed / bottom exchanger before entering the stabilizer feed drum. The liquid from the feed drum is fed to a stabilization tower at about 50-200 psi depending on whether it is sour (sour stabilization is performed at the lower end of the range, sweet stabilization at the upper end of the range). The condensate stabilizer reduces the vapor pressure of the condensate by removing the lighter components. Stabilization is typically performed in a reboiled absorber with tray-type internals (see U.S. Pat. No. 2,815,650). At the bottom of the tower, a portion of the liquid is circulated through a reboiler to add heat to the tower. The liquid leaving the bottom of the tower undergoes a series of stage flashes at ever increasing temperatures, driving off the lighter components. Therefore, these liquids must be cooled to a low enough temperature so that the vapor does not flash into the atmosphere in the condensate storage tank. (See Mokhatab, S.; Poe, W., “Handbook of Natural Gas Transmission and Processing,” 2012, incorporated herein by reference in its entirety.)

[0043] The feed crude oil pressure may be as high as 1,000 psi and must be significantly reduced before it reaches the stabilizer 302 and the stock tank 330. Temperature changes affect the production of crude oil. For example, there are 11 GOSPs in the Abqaiq region of Saudi Arabia. Abqaiq is one of the largest production plants with crude oil stabilizers. It can process up to 950,000 barrels of crude oil per day. However, in practice, the Abqaiq stabilizer can process 950,000 barrels of crude oil in summer but only about 860,000 barrels in winter. These variations in process capacity are due to the change in ambient temperature. Due to seasonal changes, the stabilization system / process design must be flexible enough to meet the demands under various operating conditions. During winter, higher concentrations of C3, C4, and C5 may be present in the oil. Therefore, the temperature set point for the stabilizer must be relatively low to increase the oil production during the winter months. In summer, the stabilizer may be operated at a higher temperature. For efficient stabilization, ambient conditions must be taken into account. (See Uwitonze, H., Hwang, K., Lee, I., “Modeling and Improving Natural Gas Condensate Process with Stripping and Heat Integration”, Chemical Engineering and Processing: Process Intensification, Volume 118, 2017; and Reza, V., Hossein, A. and Farshad, FT, “Simulation and Optimization of H2S Expulsion from Crude Oil with the Use of Equilibrium Mode”, International Conference on Chemistry and Chemical Process, Issue 10, 101-106, 2011, each of which is incorporated herein by reference in its entirety).

[0044] FIG. 4 depicts the process flow for the analysis of a GOSP plant. First, a process model simulating the GOSP plant is developed. A dynamic simulator was used to simulate the high pressure separator and the low pressure separator, respectively. A dynamic simulator was also used to simulate the stabilizer and the storage tank. In a non-limiting example, the dynamic simulator can be an OmegaLand simulator (Omegaland, Omega Simulation Company, Kosugi Building Nishiwaseda 2F, 2-20-9 Nishi-Waseda, Shinjuku-ku, Tokyo 169-0051, see https: / / www.omegasim.co.jp / contents_e / product / ol / overview / 169-0051).

[0045] Referring back to Figure 3, the oil train includes a series of separators (313, 314, 316), stabilizer 318, and stock tank 330. Figure 5 shows a non-limiting example of a GOSP plant as used in the process of Figure 4, which uses multi-stage separation including high pressure separator 512, P1 separator 514, low pressure separator 516 with stabilizer 518, and stock tank 550, each of which was simulated by a dynamic simulator.

[0046] The process model is then run under different pressure and ambient temperature conditions to generate multiple data sets.

[0047] In a non-limiting example of a process run, the following conditions were established: Model inputs were ambient temperature and LPPT pressure. Model output was liquid recovery in barrels / day. Ambient temperature range = 15~40℃. LPPT pressure range = 100-200psi. Stabilizer pressure = 55 psi.

[0048] The resulting dataset contained 714 data points. Of these, 70% were used for validation of the AI ​​and 15% for testing.

[0049] In the data processing step, the data of the dataset is normalized between [-1,1] or [0,1]. Figure 6 shows the data normalization between [-1,1] for the generated data.

[0050] Artificial intelligence (AI) techniques are being used to develop models for complex relationships between components in many fields such as construction, healthcare, transportation and management. AI methods enable the analysis of problems that were once considered too complex to analyze due to complex behavior or large amounts of data. Emerging trends in data analytics and integration of multidimensional and multimodal data enable value-added decision making in petroleum engineering applications. One such complex system is the Gas-Oil Separation Plant (GOSP), an integral part of the upstream petroleum industry. The complexity of GOSP operation is due in part to the integration of several different units and their governing principles. Establishing the relationships between the different units and each unit is difficult and depends on several operating parameters. In this disclosure, artificial intelligence techniques are applied to the measurement and control of the separator and stabilizer, individually and collectively, as representing the complete system of the separator and stabilizer.

[0051] AI techniques that may be used include Artificial Neural Networks (ANN), Functional Networks (FN), Support Vector Machines (SVM), Decision Trees (DT) and Fuzzy Logic (FL). These techniques are highly data-dependent and ensuring data quality has a significant impact on their predictive performance. (See Anifowose F, Adeniye S, Abdulraheem A., “Recent Advances in the Application of Computational Intelligence Techniques in Oil and Gas Reservoir Characterization: A comparative study”, Journal of Experimental & Theoretical Artificial Intelligence. 2014; 26(4):551-70, the entire text of which is incorporated herein by reference).

[0052] Neural network techniques have been used to predict the performance of GOSP multi-stage separation facilities, which may be used to plan and operate oil and gas separation facilities at the surface. Parameters such as gas / oil ratio (GOR) in different stage separations have been predicted. Furthermore, a genetic algorithm has been developed to optimize separator pressure in multi-stage separation. The genetic-based approach enhanced the oil separation process and increased the separated oil API. (See U.S. Pat. No. 8,805,587, the entire contents of which are incorporated herein by reference; see also Ghaedi M.; Ebrahimi, A.; Pishvaie, M. “Application of Genetic Algorithm for Optimization of Separator Production Units”, Chem Eng Commun. 2014;201(7):926-38).

[0053] In a non-limiting example of using artificial intelligence techniques, two AI models were developed to predict liquid recovery in GOSP by incorporating each separation / stabilizer unit and their relationships. Both an artificial neural network (ANN) model and a support vector machine (SVM) were constructed and their performances were compared in predicting liquid recovery. An optimization routine was then developed to determine the optimal pressure corresponding to the ambient temperature that results in maximum liquid recovery. The AI ​​technique is not limited to artificial neural network and support vector machine, but may be any type of AI model, including functional network (FN), artificial neural network (ANN), support vector machine (SVM), decision tree (Dt) and fuzzy logic (FL).

[0054] Regression models were constructed using artificial intelligence (AI) techniques, which are described as using either artificial neural networks (ANN) or support vector machines (SVM), but are not limited to ANN or SVM, and may be any of ANN, SVM, functional networks (FN), artificial neural networks (ANN), support vector machines (SVM), decision trees (Dt), and fuzzy logic (FL).

[0055] In the optimization step, the results of the AI ​​model were examined to determine the optimal operating pressure for the separator, as shown in Figures 7 to 14.

[0056] To generate the data necessary to build the AI ​​model, a GOSP process simulation model was built to mimic the driving process using a dynamic simulator. In a non-limiting example, the dynamic simulator may be an OmegaLand dynamic simulator.

[0057] The GOSP process, including instrumentation and process control, was configured and tested by running simulations for each individual unit and the complete plant. During the process, data from the initial PVT (pressure, volume, temperature) analysis was used as input for the dynamic simulator. After PVT data validation in the dynamic simulator, operational parameters were added based on accumulated historical data from gas flow, liquid flow and operating conditions. Simulation runs were performed by manipulating various process variables such as pressure and temperature with Reid Vapor Pressure (RVP) as a constraint to obtain the predicted oil recovery.

[0058] To perform the optimization, key operating variables and optimization decision parameters were identified. The operating pressure of the GOSP separator at various operating temperatures was used as the key optimization variable. Crude oil recovery was considered to be the optimization decision parameter. To analyze the effect of the process variables on oil recovery, the GOSP process simulator was run for different process variables. Data was generated by manipulating the process variables in the process simulator and the results from the simulation were used to train the AI ​​model. An exemplary data set of input and output variables is shown in Table 1.

[0059] The data set generated from the GOSP simulation model was used to train an AI model to predict liquid recovery. Before the operational data can be used, it must be normalized. Typically, the data is normalized between [-1,1] or [0,1], as shown in Figure 6, although other normalization methods such as log-normalization can be used. Once the data is normalized, an AI model such as an artificial neural network (ANN) or a support vector machine (SVM) can be trained. The output of the AI ​​model is then further analyzed.

[0060] After AI model development, process optimization was performed to determine the optimal separator pressure that would result in maximum liquid recovery for each ambient temperature. A correlation was then developed using regression techniques to determine the relationship between ambient temperature and optimal separator pressure.

[0061] Table 1: Example input and output data for the GOSP simulation model [Table 1]

[0062] Using a simulation model including the stabilizer unit, the ambient temperature (T a Data was generated for liquid recovery at different conditions of temperature, pressure, and separator pressure. Model inputs were ambient temperature and LPPT pressure. Model output was the amount of liquid recovered in barrels per day. Data was generated by setting the LPPT pressure in the range of 100-200 psi, the stabilizer pressure at one value (55 psi), and the ambient temperature over a range of values ​​(15-41 °C). The dynamic simulator with the GOSP model was run at the full range of separator pressure values ​​for a particular ambient temperature until the model reached steady state for each run. The next ambient temperature was input and the process was repeated to cover the entire ambient temperature range. The ambient temperature was increased in increments of 2 °C and the separator pressure was 2 psi at each step.

[0063] Artificial neural networks (ANNs) are based on a collection of connected units or nodes called artificial neurons that loosely model the neurons in a biological brain. Each connection can transmit a signal to other neurons, just like a synapse in a biological brain. An artificial neuron that receives a signal can then process it and send a signal to the neurons connected to it.

[0064] In an ANN implementation, the "signal" at the connections is real, and the output of each neuron is calculated by some nonlinear function of the sum of its inputs. The connections are called edges. Neurons and edges typically have weights that are adjusted as learning progresses. The weights increase or decrease the strength of the signal at the connection. Neurons have thresholds such that a signal is sent only if the aggregate signal exceeds that threshold. A neuronal net learns by processing examples, each of which contains a known "input" and "result," and forms probability-weighted associations between the two that are stored within the net's own data structure. (Herein, an "input" is more accurately called an input set, since it typically consists of multiple independent values, not a single value.) Thus, the "learning" of a neural net from a given example is the difference in the state of the net before and after processing the example. After being given a sufficient number of examples, the net becomes able to predict outcomes from the inputs using the associations built from the example set. If a feedback loop is provided to the neural net regarding the accuracy of its predictions, it will continue to refine its associations, resulting in ever-increasing levels of accuracy. In short, there is a direct relationship between the number and variety of examples processed by a neural net and the accuracy of its predictions. The ability of the artificial neural network to predict relationships and outcomes improves with successive input data sets. In this disclosure, the input to the artificial neural network is a data set simulated by a dynamic simulator for varying pressures and ambient temperatures as associated with liquid recovery.

[0065] A support vector machine (SVM) is a supervised machine learning model that uses a classification algorithm for two-group classification problems. After applying a set of labeled training data to the SVM, a support vector regression (SVR) model can be used to fit the data to a line or a hyperplane within confidence bounds defined by a distance of ±ε (epsilon). The confidence bounds are set by the SVM's operator and can be adjusted for a desired level of accuracy. The accuracy of the model is measured using the coefficient of determination (R 2 ) and mean squared error (MSE). A new data set can then be mapped into that same space and the data samples are fitted to the trained line within the confidence limits.

[0066] In statistics, the coefficient of determination, R 2 is the proportion of variance in the dependent variable (output value) that can be predicted from the (several) independent variables (input values). 2 The variance indicates how close the output value is to a set of desired output values. Variance is the expected value of the squared deviation of a random variable from its mean. Informally, it measures how far a set of numbers is from their mean.

[0067] Then, two different models for predicting liquid recovery were constructed using artificial neural network (ANN) and support vector machine (SVM). In that case, the input parameters were ambient temperature (T a ) and separator pressure, and the output was liquid recovery. The ranges of the two input parameters are shown in Table 2.

[0068] (Input parameter range) [Table 2]

[0069] From numerous runs of the simulator, 710 data points were used to develop the AI ​​model, where 70% of the dataset was used to train the model, while 15% of the dataset was used for validation and 15% of the dataset was used for testing. The selected data was normalized between [0,1]. For the ANN model, one hidden layer was used with 20 neurons and a tan-Sigmoid function was used as the activation function, while for the SVM, a Gaussian function kernel function was used.

[0070] FIG. 7 shows the data fitting for liquid recovery using the ANN model, and FIG. 8 shows the data fitting for the SVM model. Both the ANN and SVM models performed well in predicting liquid recovery. Table 3 shows the coefficient of determination (R 2 ) and mean squared error (MSE). From this comparison, it can be seen that the therapeutic method closely predicted fluid recovery, but the ANN was more accurate.

[0071] (Performance analysis) [Table 3]

[0072] Figure 9 shows a plot of pressure versus temperature, showing the predicted pressure for each temperature, a line fit to the data, and confidence limits. It can be seen that pressure versus temperature is linear within the confidence limits.

[0073] Figure 10 shows the ambient temperature T in the range of 15 to 41 °C. aFigure 1 shows oil recovery plotted as a function of GOSP separator pressure for T. Stabilizer pressure was held constant at 55 psi. The shape of the curve indicates that for any particular temperature, liquid (oil) recovery increases with increasing pressure up to a certain value and then decreases with further increases in pressure. The pressure corresponding to maximum oil recovery is the optimum pressure at that temperature. The results are shown in Figure 1 for a GOSP separator pressure (100-200 psi) and T a For the complete range (15-41°C), the following can be observed: for any fixed temperature, there is an optimum for the separator pressure at which maximum liquid recovery can be obtained for given feed and operating conditions. For example, at 15°C, the optimum separator pressure is about 150 psi, which results in a maximum liquid recovery of about 34,420 barrels per day, as shown by dotted line 1035. However, at 41°C, at a higher separator pressure of about 178 psi, as shown by dotted line 1037, the maximum liquid recovery drops to only 34,225 barrels per day. The down arrow shows the trend of decreasing liquid recovery over increasing temperature. In addition, the separator pressure required to achieve maximum liquid recovery increases with increasing temperature, which increases recovery costs.

[0074] FIG. 11 shows a typical temperature range for a period of time when the temperature is lower than the annual average temperature, such as during the winter months in some countries. In a non-limiting example, the lower temperature T l is the average temperature T avg 10℃ lower than T l and T avg The value of may be determined considering historical ambient temperature records for the GOSP location. In another non-limiting example, the winter season may be the winter season in Saudi Arabia, which has many GOSP plants. Curve 1139 shows that the optimum separator pressure for an ambient temperature Ta of approximately 15° C. is 156 psi. At 15° C. and a separator pressure of 100 psi, the liquid (oil) recovery is 34,345 barrels per day. However, if the separator is operated at 156 psi (T aIf the separator is set to operate at a pressure of 100,000 barrels per day (Ta = 15°C), the liquid recovery increases to 34,415 barrels per day (bbl / D). Thus, a separator pressure change of 56 psi at Ta = 15°C increases the liquid recovery by 70 barrels per day. For a GOSP with an inlet feed stream of 100,000 barrels per day, adjusting the separator pressure for each ambient temperature results in an improvement in GOSP production of 0.07% per day. The effect of temperature change at a fixed pressure (160 psi) also shows that oil recovery increases with decreasing temperature (see trend arrows). Daytime maximum temperatures reach 25°C, while nighttime average minimum temperatures are 15°C, with a variation (±10 during the day and night). This temperature range represents typical winter conditions in Saudi Arabia. For example, curve 1141 shows that a maximum separator pressure of 160 psi increases yield at 25° C. from 34,260 barrels / day to approximately 34,345 barrels / day, an increase of approximately 85 barrels per day or 0.2%.

[0075] Similarly, FIG. 12 shows typical temperature ranges for periods of time when temperatures are higher than the annual mean temperature, such as during the summer months in some countries. In a non-limiting example, the higher temperature T h is the average temperature T avg It can be 20°C higher than T h and T avgThe value of may be determined considering historical ambient temperature records for the GOSP location. In a non-limiting example, the summer season may be a summer season in Saudi Arabia, which may have many GOSP plants. The optimum liquid recovery was found at a temperature of 41° C. when the pressure is 174 psi. The liquid recovery corresponding to 41° C. and a separator pressure of 100 psi is 34,111 bbl / D. However, if the GOSP separator is set to operate at 174 psi (temperature=41° C.), the liquid recovery increases to 34,223 bbl / D as shown by curve 1244. The change in separator pressure can increase the liquid recovery by 112 bbl / D. The increase in oil recovery is 112 bbl / D, which is a 0.11% increase per day of GOSP production.

[0076] Also, the effect of temperature variation (wide arrow, see Figure 12) at fixed pressure (170 psi) shows that oil recovery increases with decreasing temperature (from about 34,223 at 41°C to about 34,310 at 31°C as shown by curve 1242). The temperature range represents a typical summer in Saudi Arabia: the average daytime maximum temperature reaches ~41°C while the average nighttime minimum temperature is ~31°C with variation (±10°C between daytime and nighttime).

[0077] The results show that for a given set of operating conditions, liquid recovery is significantly affected by temperature. At low temperatures (~15°C), liquid recovery was significantly higher than that at high temperatures (~41°C), affecting the liquid recovery and performance of the GOSP plant.

[0078] Although ambient temperature variations are not controllable, these results show that there is an opportunity to negate the effect of temperature on oil recovery by setting pressure conditions that allow maximum oil recovery. To maximize liquid recovery in GOSP plants, parameters can be adjusted by developing an operational decision support system (ODSS) that identifies the most productive pressure according to temperature variations.

[0079] An oil producing country may produce an average of 10 million barrels of crude oil per day. For example, in Saudi Arabia, a combination of several GOSPs can produce this amount of crude oil. Thus, small improvements in production have significant commercial impacts. For a 0.1% increase in crude oil recovery, the commercial benefits based on an average production capacity of 10 million barrels of crude oil and an average price of crude oil of $60 per barrel are very high, as shown in Table 4. A 0.1% improvement can generate revenues of $219 million per year.

[0080] (Commercial impact of a 0.1% improvement in production) [Table 4]

[0081] To illustrate the benefits of adjusting the GOSP separator pressure to accommodate ambient temperature variations, two cases are provided below.

[0082] (Case 1: Winter) Case 1 represents typical winter conditions with an average air temperature of 15° C. A graph of liquid recovery (bbl / day) versus separator pressure (psi) is shown in FIG. 13. In FIG. 13, the optimum separator pressure for the GOSP at an ambient temperature of 15° C. is 156 psi. The base GOSP separator pressure is 100 psi while all other parameters are kept constant. The additional liquid recovery gained in this case is 70 bbl / D compared to the base pressure.

[0083] The potential commercial benefits on a daily and seasonal basis using the GOSP analysis of this disclosure are shown in Table 5, which shows that adjusting the separator pressure to 156 psi could generate a premium revenue of $371,610 per 100,000 barrels of crude oil in the winter. In a non-limiting example of Saudi Arabian crude oil production for the winter season, a change to operating the separator at 156 psi during the winter season would result in a projected benefit of $37,161,000 for a daily 10 million barrel capacity of Saudi Arabian crude oil production.

[0084] (Commercial impact of adjusting separator pressure to 156 psi during winter at 15°C average ambient temperature) [Table 5]

[0085] (Case 2: Summer) Case 2 shows typical summer conditions with an average air temperature of 41° C. A graph of liquid recovery (bbl / day) versus separator pressure (psi) is shown in FIG. 14. In FIG. 14, the optimum separator pressure for GOSP at an ambient temperature of 41° C. is 174 psi. The base separator pressure is 100 psi while all parameters are constant. The additional liquid recovery obtained at 174 psi is 122 bbl / D compared to the liquid recovery at 100 psi.

[0086] FIG. 15A illustrates some aspects of a computing device 1500 as related to GOSP optimization and control of a GOSP plant. The computing device includes a network controller 1506 and an antenna 1560 for communicating with an operator and for communicating with wireless sensors, valves or other control devices of the GOSP. The CPU 1501 is a non-transitory computer readable medium that stores instructions therein, which when executed by the one or more processors, cause the one or more processors to perform the optimization and control steps of the disclosure. The CPU is connected to a dynamic simulator 1504 that models the separators, stabilizers, and storage tanks and generates a data set of pressure versus ambient temperature data, connected to a normalization unit 1506 that normalizes the data set, connected to an operator advisory station 1557 that includes an AI unit 1509 (which may be any of artificial neural networks, support vector machines, or function networks (FN), decision (Dt), and fuzzy logic (FL)) that correlates the data, and connected to an R 2and MSE values ​​to define the relationship between the training data and the test data. The optimization unit 1526 analyzes the statistical data to determine whether the pressure should be adjusted at any of the separators to increase the liquid recovery volume. The CPU is connected to the I / O device 1586 to send and receive commands / signals to the sensors and send actuation signals to the separator control valves of the GOSP. The computing unit includes a memory 1502 and a database 1580. The database may hold simulations, historical PVT data, and the like as required by the optimization unit to make a decision on whether to adjust the pressure at the separators for higher liquid recovery. The peripheral devices 1518 may include a keyboard, mouse, printer, display, or the like as required to operate the computing unit 1500. Further details of the computing unit are discussed below with respect to FIG. 15B and FIGS. 16-18.

[0087] The potential commercial benefits on a daily and seasonal basis using the GOSP analysis of this disclosure are shown in Table 6, which shows that adjusting the GOSP pressure to 174 psi could generate a premium revenue of $657,000 per 100,000 barrels of crude oil in the summer. In a non-limiting example of seasonal Saudi Arabian crude oil production, a change to running the separator at 174 psi during the summer would result in a projected benefit of $65,700,000 per 10 million barrels of Saudi Arabian crude oil production capacity.

[0088] (Commercial impact of adjusting separator pressure to 174 psi during summer with an average ambient temperature of 41°C) [Table 6]

[0089] The disclosed method maximizes oil recovery through pressure optimization for separators in GOSP plants. The method recommends optimal pressures for each ambient temperature that result in maximum liquid recovery for interconnected GOSP units including stabilizers. Based on the experimental results, the following conclusions can be drawn: 1) Separator pressure was found to have a significant effect on oil recovery from each individual separation stage. 2) The artificial neural network (ANN) was able to predict the optimal pressure with high accuracy. 3) AI techniques were used to establish relationships based on input fluid composition, gas and liquid velocities and process parameters. 4) LPPT pressure was found to have a significant effect on oil recovery from each separation stage. 5) The oil recovery of GOSP increases with increasing separator pressure until the separator pressure reaches the optimal separator pressure, and then the oil recovery decreases with further increase in separator pressure. 6) Oil recovery appears to be strongly dependent on ambient temperature. Liquid oil recovery decreases with increasing ambient temperature. 7) At high temperatures (summer), more barrels of liquid oil can be obtained per day by adjusting the separator pressure compared to low temperatures (winter). This result has significant economic value for countries with hot climates. 8) Small adjustments in separator pressure can lead to significant revenue generation. Although the improvement seems small, about 0.1%, the economic impact is large because of the large amount of oil recovered, and therefore significant additional revenue can be obtained. In a non-limiting example, a GOSP plant in Saudi Arabia can achieve over $219 million per year.

[0090] Thus, the method of the present disclosure provides an operational advisory solution that is a tool for oil operators that can add significant value to their oil processing facilities with little investment.

[0091] Embodiments of the present disclosure may be described in the following brackets:

[0092] (1) A method for maximizing stabilized oil recovery in a gas-oil separation plant (GOSP) 300, comprising: constructing, by a computing unit 1500 including a dynamic simulator 1504, a process model of the GOSP, the process model including at least a plurality of separators (312, 314, 316), a stabilizer 302, and a storage tank 330 of the GOSP; simulating stabilized oil recovery of the GOSP under variable operating pressures of at least one of the plurality of separators for different ambient temperatures surrounding the stabilizer to generate a simulated stabilized oil recovery data set; and calculating, by the computing unit, a simulated stabilized oil recovery data set. and normalizing the set of separator pressures (see FIG. 4, FIG. 6, 1506, see also FIG. 15A) to generate a normalized stabilized oil recovery dataset; training, by a computing unit, an artificial intelligence unit on a first portion of the normalized stabilized crude recovery dataset to generate a trained artificial intelligence model; inputting, by a computing unit, a second portion of the normalized stabilized crude recovery dataset to the trained artificial intelligence model; predicting stabilized oil recovery for each separator pressure at each ambient temperature; determining, by the computing unit, an optimal pressure that maximizes the stabilized oil recovery for each ambient temperature; and adjusting, by the computing unit, operating pressures of the multiple separators of the GOSP to compensate for the ambient temperature conditions.

[0093] (2) The method of (1), wherein the artificial intelligence unit includes an artificial neural network (ANN), and the method includes: training, by a computing unit, an ANN on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained ANN; inputting, by the trained ANN, a second portion of the normalized stabilized crude oil recovery dataset; predicting, by the trained ANN, stabilized crude oil recovery values ​​at each separator pressure for each ambient temperature; and determining, by the computing unit, an optimal pressure for each ambient temperature.

[0094] (3) The method of (2) or (3), further comprising: measuring a separator pressure for each ambient temperature of the GOSP; measuring an amount of stabilized crude oil recovered in the storage tank at each separator pressure and ambient temperature over a period of time; applying a current data set including the measured separator pressure, ambient temperature, and amount of stabilized crude oil recovered in the storage tank of the GOSP to the trained ANN to predict stabilized crude oil recovery at each separator pressure for each ambient temperature; determining a statistical distance of the current stabilized crude oil recovery from the predicted stabilized crude oil recovery; determining, by a computing unit, an optimal pressure that maximizes stabilized crude oil recovery for each ambient temperature; and adjusting the operating pressures of the multiple separators of the GOSP to compensate for the current ambient temperature conditions.

[0095] (4) Test the accuracy of the trained ANN by inputting a third portion of the normalized stabilized oil recovery dataset to generate a test output dataset and determining the coefficient of determination R between the predicted stabilized oil recovery and the test stabilized oil recovery for each pressure and each ambient temperature. 2 and performing a statistical analysis between the training data set and the test output data set by calculating:

[0096] (5) Any one of the methods of (2) to (4), further comprising determining a statistical distance by calculating a mean squared error (MSE) between the predicted stabilized oil recovery and the current stabilized oil recovery for each pressure and each ambient temperature.

[0097] (6) The method of (1), wherein the artificial intelligence unit includes a support vector machine (SVM), and the method further includes: training, by the computing unit, the SVM on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained SVM; predicting, by the trained SVM, the stabilized crude oil recovery at each separator pressure for each ambient temperature; and determining, by the computing unit, an optimal pressure that maximizes the stabilized crude oil recovery for each ambient temperature.

[0098] (7) The method of (6), further comprising: measuring a separator pressure for each ambient temperature of the GOSP; measuring an amount of stabilized crude oil recovered in a storage tank of the GOSP at each separator pressure and ambient temperature over a period of time; applying a current data set including the measured separator pressure, ambient temperature, and amount of stabilized crude oil recovered in the storage tank of the GOSP to the trained SVM to predict stabilized crude oil recovery at each separator pressure for each ambient temperature; determining a statistical distance of the current stabilized crude oil recovery from the predicted stabilized crude oil recovery; determining, by a computing unit, an optimal pressure that maximizes stabilized crude oil recovery for each ambient temperature; and adjusting operating pressures of the multiple separators of the GOSP to compensate for the current ambient temperature conditions.

[0099] (8) Test the accuracy of the trained SVM by inputting a third portion of the normalized stabilized oil recovery dataset to generate a test output dataset and determining the coefficient of determination R between the predicted stabilized oil recovery and the test stabilized oil recovery for each pressure and each ambient temperature. 2and performing a statistical analysis between the trained output data set and the test output data set by calculating:

[0100] (9) Any one of the methods of (6) to (8), further comprising performing a statistical analysis between the trained output data set and the test output data set by calculating a mean squared error (MSE) between the predicted stabilized oil recovery and the current stabilized oil recovery for each pressure and each ambient temperature.

[0101] (10) The method of (1), further comprising measuring an ambient temperature surrounding the separator, measuring an operating pressure of a low pressure production trap (LPPT) separator, and adjusting the operating pressure of the LPPT to an operating pressure that maximizes stabilized oil recovery for the measured ambient temperature.

[0102] (11) The method of (10), further comprising increasing the operating pressure of the LPPT by at least 50% from the base operating pressure during months of the year that are cooler than the average annual temperature, and increasing the operating pressure of the LPPT by at least 70% from the base operating pressure during months of the year that are warmer than the average annual temperature.

[0103] (12) The method of (1), wherein the artificial intelligence unit includes any one of an artificial neural network (ANN), a support vector machine (SVM), a function network (FN), a decision tree (DT) and a fuzzy logic (FL).

[0104] (13) A system for maximizing stabilized crude oil recovery, the system including a gas-oil separation plant (GOSP) 300, the GOSP including the following components: a plurality of separators (312, 314, 316), a stabilizer 302, a storage tank 330 configured to hold stabilized crude oil, and a computing unit 1500 operatively connected to the components, the computing unit including a computer readable medium including program instructions that cause the processing circuitry to build a process model of the GOSP using a dynamic simulator 1504, the process model including at least the plurality of separators, the stabilizer, and the storage tank, and to simulate crude oil recovery of the GOSP under variable operating pressures of at least one of the plurality of separators for different ambient temperatures surrounding the stabilizer, and to calculate the simulated crude oil recovery. and causing a normalization unit of the computing unit to generate a simulated stabilized oil recovery dataset, normalizing the simulated stabilized oil recovery dataset by a normalization unit of the computing unit to generate a normalized stabilized oil recovery dataset (see FIG. 4, FIG. 6, 1506, see also FIG. 15A), training an artificial intelligence unit on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained artificial intelligence model, inputting a second portion of the normalized stabilized crude oil recovery dataset into the trained artificial intelligence model to predict oil recovery for each separator pressure at each ambient temperature, determining by the normalization unit of the computing unit a pressure that results in a maximum stabilized oil recovery for each ambient temperature, and causing the computing unit to adjust operating pressures of the multiple separators of the GOSP to compensate for the ambient temperature conditions.

[0105] (14) The system of (13), wherein the artificial intelligence unit includes an artificial neural network (ANN), and the computing unit is further configured to train the ANN on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained ANN, input a second portion of the normalized stabilized crude oil recovery dataset to the trained ANN, predict, by the trained ANN, stabilized crude oil recovery at each separator pressure for each ambient temperature, and determine, by an optimization unit, an optimal pressure that maximizes the stabilized crude oil recovery for each ambient temperature.

[0106] (15) The system of (13) or (14), further configured to: (a) measure a separator pressure for each ambient temperature of the GOSP; (b) measure an amount of stabilized crude oil recovered in the storage tank at each separator pressure and each ambient temperature over a period of time; (c) apply a current data set including the measured separator pressure, the ambient temperature, and the amount of stabilized crude oil recovered in the storage tank of the GOSP to the trained ANN to predict stabilized crude oil recovery at each separator pressure for each ambient temperature; (d) determine, by a statistical analysis unit of the computing unit, a statistical distance of the current stabilized crude oil recovery from the predicted stabilized crude oil recovery; (e) determine, by an optimization unit of the computing unit, an optimal pressure that maximizes the stabilized crude oil recovery for each ambient temperature; and (f) adjust operating pressures of the multiple separators of the GOSP to compensate for the current ambient temperature conditions.

[0107] (16) The computing unit tests the accuracy of the trained ANN by inputting a third portion of the normalized stabilized crude recovery to generate a test output data set, and determines a coefficient of determination R between the predicted normalized stabilized crude recovery and the test stabilized crude recovery for each pressure and each ambient temperature. 2and performing a statistical analysis between the trained output data set and the test output data set by calculating, by the statistical analysis unit, a mean squared error (MSE) between the predicted stabilized oil recovery and the current stabilized oil recovery for each pressure and each ambient temperature.

[0108] (17) The system of (13), wherein the artificial intelligence unit includes a source vector machine (SVM), and the computing unit is configured to train the SVM on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained SVM, input a second portion of the normalized stabilized crude oil recovery dataset to the trained SVM, predict stabilized crude oil recovery at each separator pressure for each ambient temperature by the trained SVM, and determine, by an optimization unit, an optimal pressure that maximizes the stabilized crude oil recovery for each ambient temperature.

[0109] (18) The system of (17), wherein the computing unit is further configured to: measure a separator pressure for each ambient temperature of the GOSP; measure an amount of stabilized crude oil recovery recovered in the storage tank at each separator pressure and ambient temperature over a period of time; apply a current dataset including the measured separator pressure, the ambient temperature, and the amount of stabilized crude oil recovery recovered in the storage tank of the GOSP to the trained SVM to predict stabilized crude oil recovery at each separator pressure for each ambient temperature; determine, by a statistical analysis unit, a statistical distance of the current stabilized crude oil recovery from the predicted stabilized crude oil recovery; determine, by an optimization unit of the computing unit, an optimal pressure that maximizes the stabilized crude oil recovery for each ambient temperature; and adjust operating pressures of the multiple separators of the GOSP to compensate for the current ambient temperature conditions.

[0110] (19) The computing unit tests the accuracy of the trained SVM by inputting a third portion of the normalized stabilized crude recovery dataset to generate a test output dataset, and the statistical analysis unit calculates a coefficient of determination R between the predicted normalized stabilized crude recovery and the test stabilized crude recovery for each pressure and each ambient temperature. 2 and performing a statistical analysis between the trained output data set and the test output data set by calculating a mean squared error (MSE) between the predicted stabilized oil recovery and the current stabilized oil recovery for each pressure and each ambient temperature.

[0111] (20) A non-transitory computer readable medium having instructions stored therein, the instructions, when executed by one or more processors, causing a method for maximizing stabilized oil recovery in a gas-oil separation plant (GOSP) 300, the method comprising: constructing, by a computing unit 1500 including a dynamic simulator 1504, a process model of the GOSP, the process model including at least a plurality of separators (312, 314, 316) of the GOSP, a stabilizer 302, and a storage tank 330; simulating stabilized oil recovery of the GOSP under variable operating pressures of at least one of the plurality of separators for different ambient temperatures surrounding the stabilizer to generate a simulated stabilized oil recovery data set; and normalizing, by a computing unit, the simulated stabilized crude oil recovery dataset (see FIG. 4, FIG. 6; 1506, see also FIG. 15A) to generate a normalized stabilized crude oil recovery dataset; training, by a computing unit, an artificial intelligence unit on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained artificial intelligence model; inputting a second portion of the normalized stabilized crude oil recovery dataset into the trained artificial intelligence model; predicting stabilized crude oil recovery for each separator pressure at each ambient temperature; determining, by the computing unit, a pressure that maximizes the stabilized crude oil recovery for each ambient temperature; and adjusting, by the computing unit, operating pressures of the multiple separators of the GOSP to compensate for ambient temperature conditions.

[0112] Next, referring to FIG. 15B, further details of a hardware description of the computing environment of FIG. 15A according to an exemplary embodiment will be described. In FIG. 15B, the computing device 1500 described is a representation of the computing unit 1500 of FIG. 15A. The computing device includes a CPU 1501 that executes the above / below described processes. Process data and instructions may be stored in a memory 1502. These processes and instructions may be stored in a storage medium disk 1504, such as a hard drive (HDD) or a portable storage medium, or may be stored remotely.

[0113] Furthermore, the claims are not limited by the format of the computer readable medium storing the instructions for the inventive process, for example, the instructions may be stored on a CD, DVD, FLASH memory, RAM, ROM, PROM, EPROM, EEPROM, hard disk, or any other information processing device with which a computing device, such as a server or computer, communicates.

[0114] Further, the claims may be provided as a utility application, background daemon, or operating system component, or combination thereof, executing in conjunction with the CPU 1501, 1503 and an operating system, such as Microsoft Windows 7, Microsoft Windows 10, UNIX, Solaris, LINUX, Apple MAC-OS, and other systems known to those skilled in the art.

[0115] The hardware elements for achieving the computing device may be realized by various circuit elements known to those skilled in the art. For example, the CPU 1501 or CPU 1503 may be a Xenon or Core processor from Intel of America, or an Opteron processor from AMD of America, or other processor types recognized by those skilled in the art. Alternatively, the CPU 1501, 1503 may be implemented on an FPGA, an ASIC, a PLD, or using discrete logic circuits, as recognized by those skilled in the art. Furthermore, the CPU 1501, 1503 may be implemented as multiple processors operating cooperatively in parallel to execute instructions of the inventive process described above.

[0116] The computing device of FIG. 15B also includes a network controller 1506, such as an Intel Ethernet PRO network interface card from Intel Corporation of America, for interfacing with a network 1560. As can be appreciated, the network 1560 can be a public network, such as the Internet, a private network, such as a LAN or WAN network, or any combination thereof, and can include a PSTN or ISDN sub-network. The network 1560 can be wired, such as an Ethernet network, or wireless, such as a cellular network, including EDGE, 3G and 4G wireless cellular systems. The wireless network can also be Wi-Fi, Bluetooth, or any other wireless form of communication known in the art.

[0117] The computing device further includes a display controller 1508, such as an NVIDIA GeForce GTX or Quadro graphics adapter from NVIDIA Corporation of America, for interfacing with a display 1510, such as a Hewlett Packard HPL2445w LCD monitor. A general purpose I / O interface 1512 interfaces with a keyboard and / or mouse 1514, and a touch screen panel 1516 on or separate from the display 1510. The general purpose I / O interface also connects to a variety of peripherals 1518, including printers and scanners, such as an OfficeJet or DeskJet from Hewlett Packard.

[0118] A sound controller 1520, such as the Sound Blaster X-Fi Titanium from Creative, may be provided in the computing device to interface with a speaker / microphone 1522, thereby providing sound and / or music.

[0119] A general purpose storage controller 1524 connects the storage media disk 1504 to a communication bus 1526, which may be ISA, EISA, VESA, PCI, or the like, for interconnecting all of the components of the computing device. A description of the general structure and functionality of the display 1510, keyboard and / or mouse 1514, as well as the display controller 1508, storage controller 1524, network controller 1506, sound controller 1520, and general purpose I / O interface 1512 is omitted here for brevity, since these structures are known.

[0120] Exemplary circuit elements described in the context of this disclosure may be substituted with other elements and may be configured differently from the examples provided herein. Moreover, circuits configured to perform the configurations described herein may be implemented in multiple circuit units (e.g., chips), or the configurations may be incorporated into circuitry on a single chipset, as shown in FIG.

[0121] 16 illustrates a schematic diagram of a data processing system for performing the functions of the exemplary embodiments, in accordance with certain embodiments. The data processing system is an example of a computer in which code or instructions for performing the processes of the exemplary embodiments may be located.

[0122] In FIG. 16, data processing system 1600 utilizes a hub architecture including a northbridge and memory controller hub (NB / MCH) 1625 and a southbridge and input / output (I / O) controller hub (SB / ICH) 1620. A central processing unit (CPU) 1630 is connected to NB / MCH 1625. NB / MCH 1625 also connects to memory 1645 via a memory bus and to a graphics processor 1650 via an accelerated graphics port (AGP). NB / MCH 1625 also connects to SB / ICH 1620 via an internal bus (e.g., an integrated media interface or a direct media interface). CPU processing unit 1630 may include one or more processors and may even be implemented using one or more heterogeneous processor systems.

[0123] For example, FIG. 17 illustrates one implementation of the CPU 1630. In one implementation, an instruction register 1738 retrieves instructions from a high-speed memory 1740. At least some of these instructions are fetched from the instruction register 1738 by the control logic 1736 and interpreted according to the instruction set architecture of the CPU 1630. Some of the instructions may also be directed to the registers 1732. In one implementation, the instructions are decoded according to a hardwired method, and in another embodiment, the instructions are decoded according to a microprogram that converts the instructions into a set of CPU configuration signals that are applied sequentially over multiple clock pulses. After fetching and decoding the instructions, the instructions are executed using an arithmetic logic unit (ALU) 1734, which loads values ​​from the registers 1732 and performs logical and mathematical operations on the loaded values ​​according to the instructions. Results from these operations can be fed back to registers and / or stored in the high-speed memory 1740. Depending on the particular implementation, the instruction set architecture of the CPU 1630 may use a reduced instruction set architecture, a complex instruction set architecture, a vector processor architecture, a very large instruction word architecture. Furthermore, the CPU 1630 may be based on the Von Neuman model or the Harvard model. The CPU 1630 may be a digital signal processor, an FPGA, an ASIC, a PLA, a PLD, or a CPLD. Furthermore, the CPU 1630 may be an x86 processor by Intel or AMD, an ARM processor, e.g., a Power architecture processor by IBM, a SPARC architecture processor by Sun Microsystems or Oracle, or other known CPU architectures.

[0124] 16, data processing system 1600 may include SB / ICH 1620 coupled through a system bus to an I / O bus, read only memory (ROM) 1656, universal serial bus (USB) ports 1664, flash binary input / output system (BIOS) 1668, and graphics controller 1658. PCI / PCIe devices may also be coupled to SB / ICH 1688 through PCI bus 1662.

[0125] PCI devices may include, for example, Ethernet adapters, add-in cards, and PC cards for notebook computers. Hard disk drive 1660 and CD-ROM 1666 may use, for example, Integrated Drive Electronics (IDE) or Serial Advanced Technology Attachment Interface (SATA). In one implementation, the I / O bus may include Super I / O (SIO) devices.

[0126] Additionally, a hard disk drive (HDD) 1660 and an optical drive 1666 may also be coupled to the SB / ICH 1620 through the system bus. In one implementation, a keyboard 1670, a mouse 1672, a parallel port 1678, and a serial port 1676 may be connected to the system bus through an I / O bus. Other peripherals and devices may be connected to the SB / ICH 1620 using mass storage controllers such as SATA or PATA, Ethernet ports, ISA bus, LPC bridges, SMBus, DMA controllers, and audio codecs.

[0127] Moreover, the disclosure is not limited to the particular circuit elements described herein, nor is the disclosure limited to the particular sizing and classification of these elements. For example, one of ordinary skill in the art will understand that the circuit configurations described herein may be adapted based on variations in the sizing and chemistry of the battery or based on the requirements of the intended backup load to be powered.

[0128] The functions and configurations described herein may be performed by various distributed components of a system. For example, one or more processors may perform these system functions, and the processors are distributed among multiple components that communicate within a network. The distributed components may include one or more client and server machines that may share processing, as illustrated by FIG. 18, as well as various human interface and communication devices (e.g., display monitors, smartphones, tablets, personal digital assistants (PDAs)). The network may be a private network, such as a LAN or WAN, or a public network, such as the Internet. Input to the system may be received via direct user input, received remotely in real time or as a batch process. In addition, some implementations may be performed on modules or hardware that are not identical to those described. Thus, other implementations are within the scope of what may be claimed.

[0129] The above hardware descriptions are non-limiting examples of corresponding structures for performing the functions described herein.

[0130] Obviously, numerous modifications and variations of the present disclosure are possible in light of the above teachings. It is therefore understood that, within the scope of the appended claims, the invention may be practiced otherwise than as specifically described herein.

Claims

1. 1. A method for maximizing stabilized oil recovery in a gas-oil separation plant (GOSP), comprising: constructing, by a computing unit including a dynamic simulator, a process model of the GOSP, the process model including at least a plurality of separators, a stabilizer, and a storage tank of the GOSP; simulating the stabilized oil recovery of the GOSP under variable operating pressures of at least one separator of the plurality of separators for different ambient temperatures surrounding the stabilizer to generate a simulated stabilized oil recovery data set; normalizing, by the computing unit, the simulated stabilized oil recovery dataset to generate a normalized stabilized oil recovery dataset; training, by the computing unit, an artificial intelligence unit on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained artificial intelligence model; inputting a second portion of the normalized stabilized oil recovery dataset into the trained artificial intelligence model; predicting the stabilized oil recovery for each separator pressure at each ambient temperature; determining, by the computing unit, an optimal pressure that maximizes stabilized oil recovery for each ambient temperature; and adjusting, by the computing unit, the operating pressures of the plurality of separators of the GOSP to compensate for ambient temperature conditions. method.

2. The artificial intelligence unit includes an artificial neural network (ANN), and the method includes: training, by the computing unit, the ANN on the first portion of the normalized stabilized crude oil recovery dataset to generate a trained ANN; inputting the second portion of the normalized stabilized oil recovery dataset into the trained ANN; predicting stabilized oil recovery values ​​at each separator pressure for each ambient temperature with the trained ANN; determining, by the computing unit, the optimum pressure for each ambient temperature. The method of claim 1.

3. Measuring the separator pressure for each ambient temperature of the GOSP; measuring the amount of stabilized crude oil recovered in the storage tank at each separator pressure and ambient temperature over a period of time; applying a current data set, including measured separator pressures, ambient temperatures, and the amount of stabilized crude oil recovered in the storage tanks of the GOSP, to the trained ANN to predict the stabilized crude oil recovery at each separator pressure for each ambient temperature; determining a statistical distance of the current stabilized oil recovery from the predicted stabilized oil recovery; determining, by the computing unit, an optimal pressure that maximizes stabilized oil recovery for each ambient temperature; and adjusting the operating pressure of the plurality of separators of the GOSP to compensate for current ambient temperature conditions. The method of claim 2.

4. testing accuracy of the trained ANN by inputting a third portion of the normalized stabilized crude oil recovery dataset to generate a test output dataset; The coefficient of determination R between the predicted stabilized oil recovery and the test stabilized oil recovery for each pressure and each ambient temperature. ２ and performing a statistical analysis between the training data set and the test output data set by calculating The method according to claim 3.

5. 5. The method of claim 4, further comprising determining the statistical distance by calculating a mean square error (MSE) between the predicted stabilized oil recovery and the current stabilized oil recovery for each pressure and each ambient temperature.

6. The artificial intelligence unit includes a support vector machine (SVM), and the method includes: training, by the computing unit, the SVM on the first portion of the normalized stabilized oil recovery dataset to generate a trained SVM; inputting the second portion of the normalized stabilized oil recovery dataset into the trained SVM; predicting stabilized oil recovery at each separator pressure for each ambient temperature with the trained SVM; determining, by the computing unit, an optimal pressure that maximizes stabilized oil recovery for each ambient temperature. The method of claim 1.

7. Measuring the separator pressure for each ambient temperature of the GOSP; Measuring the amount of stabilized crude oil recovered in the storage tank of the GOSP at each separator pressure and ambient temperature over a period of time; applying a current data set, including measured separator pressures, ambient temperatures, and the amount of stabilized crude oil recovered in the storage tank of the GOSP, to the trained SVM to predict the stabilized crude oil recovery at each separator pressure for each ambient temperature; determining a statistical distance of the current stabilized oil recovery from the predicted stabilized oil recovery; determining, by the computing unit, an optimal pressure that maximizes stabilized oil recovery for each ambient temperature; and adjusting the operating pressure of the plurality of separators of the GOSP to compensate for current ambient temperature conditions. The method according to claim 6.

8. testing the accuracy of the trained SVM by inputting a third portion of the normalized stabilized oil recovery dataset to generate a test output dataset; The coefficient of determination R between the predicted stabilized oil recovery and the test stabilized oil recovery for each pressure and each ambient temperature. ２ and performing a statistical analysis between the trained output data set and the test output data set by calculating The method according to claim 7.

9. 10. The method of claim 8, further comprising performing a statistical analysis between the training output data set and the test output data set by calculating a mean squared error (MSE) between the predicted stabilized oil recovery and the current stabilized oil recovery for each pressure and each ambient temperature.

10. Measuring the ambient temperature surrounding the separator; and Measuring the operating pressure of a low pressure production trap (LPPT) separator; and adjusting the operating pressure of the LPPT to the operating pressure that maximizes stabilized oil recovery for the measured ambient temperature. The method of claim 1.

11. increasing the operating pressure of the LPPT by at least 50% from a base operating pressure during months of the year that are cooler than the average annual temperature; and increasing the operating pressure of the LPPT by at least 70% from a base operating pressure during months of the year that are warmer than the average annual temperature. The method of claim 10.

12. 2. The method of claim 1, wherein the artificial intelligence unit comprises any one of an artificial neural network (ANN), a support vector machine (SVM), a function network (FN), a decision tree (DT) and a fuzzy logic (FL).

13. 1. A system for maximizing stabilized oil recovery, comprising: The system includes a gas-oil separation plant (GOSP), The GOSP comprises the following components: A plurality of separators; A stabilizer; a storage tank configured to hold the stabilized crude oil; a computing unit operatively connected to the components, the computing unit including a computer readable medium including program instructions that cause processing circuitry to: constructing a process model of the GOSP using a dynamic simulator, the process model including at least the plurality of separators, the stabilizer, and the storage tank; simulating oil recovery of the GOSP under variable operating pressures of at least one separator of the plurality of separators for different ambient temperatures surrounding the stabilizer to generate a simulated stabilized oil recovery data set; normalizing, by a normalization unit of the computing unit, the simulated stabilized oil recovery dataset to generate a normalized stabilized oil recovery dataset; training an artificial intelligence unit on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained artificial intelligence model; inputting a second portion of the normalized stabilized oil recovery dataset into the trained artificial intelligence model; predicting said oil recovery for each separator pressure at each ambient temperature; determining, by an optimization unit of said computing unit, a pressure that results in maximum stabilized oil recovery for each ambient temperature; adjusting, by the computing unit, the operating pressures of the plurality of separators of the GOSP to compensate for ambient temperature conditions; and system.

14. The artificial intelligence unit includes an artificial neural network (ANN); The computing unit is training the ANN on the first portion of the normalized stabilized oil recovery dataset to generate a trained ANN; inputting the second portion of the normalized stabilized oil recovery dataset into the trained ANN; predicting stabilized oil recovery at each separator pressure for each ambient temperature with the trained ANN; and The optimization unit determines the optimal pressure that maximizes stabilized oil recovery for each ambient temperature. Further configured as follows: The system of claim 13.

15. The computing unit is measuring the separator pressure for each ambient temperature of the GOSP; measuring the amount of stabilized crude oil recovered in the storage tank at each separator pressure and each ambient temperature over a period of time; applying a current data set, including measured separator pressures, ambient temperatures, and the amount of stabilized crude oil recovered in the storage tanks of the GOSP, to the trained ANN to predict the stabilized crude oil recovery at each separator pressure for each ambient temperature; determining, by a statistical analysis unit of the computing unit, a statistical distance of a current stabilized crude oil recovery from the predicted stabilized crude oil recovery; determining, by the optimization unit of the computing unit, an optimal pressure that maximizes stabilized oil recovery for each ambient temperature; and Adjusting the operating pressure of the separators of the GOSP to compensate for current ambient temperature conditions. Further configured as follows: The system of claim 14.

16. The computing unit is testing the accuracy of the trained ANN by inputting a third portion of the normalized stabilized crude oil recovery to generate a test output data set; The coefficient of determination R between the predicted normalized stabilized oil recovery and the test stabilized oil recovery for each pressure and each ambient temperature. ２ performing a statistical analysis between the trained output data set and the test output data set by calculating performing a statistical analysis between the training output data set and the test output data set by calculating a mean square error (MSE) between the predicted stabilized oil recovery and the current stabilized oil recovery for each pressure and each ambient temperature by the statistical analysis unit; Further configured as follows: The system of claim 15.

17. The artificial intelligence unit includes a source vector machine (SVM); The computing unit is training the SVM on the first portion of the normalized stabilized oil recovery dataset to generate a trained SVM; inputting the second portion of the normalized stabilized oil recovery dataset into the trained SVM; predicting stabilized oil recovery at each separator pressure for each ambient temperature with the trained SVM; and The optimization unit determines the optimal pressure that maximizes stabilized oil recovery for each ambient temperature. It is configured as follows: The system of claim 13.

18. The computing unit is measuring the separator pressure for each ambient temperature of the GOSP; measuring the amount of stabilized crude oil recovery recovered in the storage tank at each separator pressure and ambient temperature over a period of time; applying a current data set, including measured separator pressures, ambient temperatures, and the amount of stabilized oil recovery recovered in the storage tank of the GOSP, to the trained SVM to predict the stabilized oil recovery at each separator pressure for each ambient temperature; determining, by a statistical analysis unit, a statistical distance of the current stabilized crude oil recovery from the predicted stabilized crude oil recovery; determining, by the optimization unit of the computing unit, an optimal pressure that maximizes stabilized oil recovery for each ambient temperature; and Adjusting the operating pressure of the separators of the GOSP to compensate for current ambient temperature conditions. Further configured as follows:

20. The system of claim 17.

19. The computing unit is testing the accuracy of the trained SVM by inputting a third portion of the normalized stabilized crude oil recovery dataset to generate a test output dataset; The statistical analysis unit determines a coefficient of determination R between the predicted normalized stabilized crude recovery and the test stabilized crude recovery for each pressure and each ambient temperature. ２ performing a statistical analysis between the trained output data set and the test output data set by calculating Perform a statistical analysis between the training output data set and the test output data set by calculating a mean square error (MSE) between the predicted stabilized oil recovery and the current stabilized oil recovery for each pressure and each ambient temperature. Further configured as follows:

20. The system of claim 18.

20. A non-transitory computer readable medium having instructions stored thereon, the instructions, when executed by one or more processors, cause a method for maximizing stabilized oil recovery in a gas-oil separation plant (GOSP), the method comprising: constructing, by a computing unit including a dynamic simulator, a process model of the GOSP, the process model including at least a plurality of separators, a stabilizer, and a storage tank of the GOSP; simulating the stabilized oil recovery of the GOSP under variable operating pressures of at least one separator of the plurality of separators for different ambient temperatures surrounding the stabilizer to generate a simulated stabilized oil recovery data set; normalizing, by the computing unit, the simulated stabilized oil recovery dataset to generate a normalized stabilized oil recovery dataset; training, by the computing unit, an artificial intelligence unit on a first portion of the normalized stabilized crude oil recovery dataset to generate a trained artificial intelligence model; inputting a second portion of the normalized stabilized oil recovery dataset into the trained artificial intelligence model; predicting the stabilized oil recovery for each separator pressure at each ambient temperature; determining, by the computing unit, a pressure that maximizes stabilized oil recovery for each ambient temperature; and adjusting, by the computing unit, the operating pressures of the plurality of separators of the GOSP to compensate for ambient temperature conditions. Non-transitory computer-readable medium.

Images