We are pleased to announce the programme for the forthcoming Multiphase Conference in Banff, Canada. Download it here:
Keynote
As we approach the half-century milestone of utilizing mechanistic flow models in the design and operation of multiphase production systems, it becomes evident that these models have undergone significant evolution, paralleling the advancements in the simulators that employ them. In the present day, state-of-the-art models continue to undergo refinement and adaptation to address emerging challenges. This paper examines these models, scrutinizing their performance against both experimental and field data to identify areas of robust confidence and lingering uncertainties. Moreover, we highlight key domains of ongoing research aimed at further enhancing modelling capabilities to meet the demands of the next generation of challenges on the horizon.
Measurement
Two-phase and three-phase separators are widely used as a standard for well testing and production allocation. Test separators typically utilize highly accurate, single-phase flow meters to provide gas and liquid flow rates. As a result, the accuracy of separator-based measurements is largely driven by the degree to which the separator can effectively separate the incoming multiphase mixtures into their gas and liquid components.
For liquid measurements on three-phase separators, relatively small amounts of entrained gas for gas carry-under and/or gas breakout can result in significant errors in liquid rates. For two phase separators, errors in density-based watercut from Coriolis meters due to entrained gas can increase errors in reported net-oil significantly.
It is widely recognized that errors in the liquid rates from single-phase flow meters typically scale with the gas void fraction for bubbly liquids. However, without quantification of gas void fractions within the liquid flow meters, operators are often unable to quantify measurement error due to incomplete separation.
This paper presents analytical and experimental results quantifying gas void fraction levels within turbine and Coriolis meters. The approach utilizes SONAR-based beamforming techniques to measure the speed naturally-occurring, low-frequency, acoustic waves by interpreting the output of a pair acoustic pressure transducers spanning the inlet and outlet of the flow meter. The measured sound speed quantifies gas void fraction utilizing Wood’s equation.
The paper presents experimental data on the errors exhibited by several Coriolis meters or various size and tube geometries and by a turbine meter, and discusses methods such as machine learning to utilize real-time gas void fraction measurements to mitigate errors due to entrained gases. The approach has application across a wide range of flow measurement challenges, from conventional separators to measurement of fluids important for the energy transition, such as CO2 and LNG operating near phase transition boundaries.
Coriolis devices are constantly evolving to address various challenges for flow measurement, including wet gas flow metering. Despite being utilised in certain commercial wet gas meters, there is a lack of data concerning the fundamental aspects of this device when subject to different wet gas flow conditions. Indeed, such data would be helpful for the exploration of various features of Coriolis devices and adapting them to real conditions. Therefore, to establish a reliable model for measuring wet gas flow rates using Coriolis devices, it becomes crucial to conduct a thorough experiment under different conditions. In this paper, an experimental study on the influence of wet gas flow conditions on the metering performances of Coriolis device in a 50 mm inner diameter pipeline, using air, water, and oil as experimental fluids with Lockhart-Martinelli (XLM) values ranging from 0.01 to 0.30, was reported. The study included a comparison of Coriolis performance in different wet gas mediums. It also explored the impact of these different mediums on the over-reading (OR) of the device. Additionally, the effect of pressure on Coriolis response was investigated. It was found that the flow patterns observed in wet gas conditions were not confined to just annular flow, a factor largely overlooked in previous studies. Lastly, the study examined the effect of the Froude number (Fr) on Coriolis response under different conditions. The results indicated that OR is nearly independent of flow regimes across various wet gas conditions. However, OR does depend on the fluid properties passing through the tube.
Keywords: Coriolis flow meter, Flow pattern, Froude number, Over-Reading, Wet gas
Hydrates - Case Studies
This study presents a case of hydrate blockage during a PIG operation in an ultra-deepwater well in the Santos Basin, exploring methods for recovery the service line operations without rig intervention. The procedure utilizes the production line for depressurization and hydrate dissociation fluid reception, optimizing costs. Transient simulations guide liquid removal using natural gas due to bathymetry and lazy-wave riser configuration. Nitrogen completes cleaning and inertization, crucial for successful dissociation. Multidisciplinary analysis and alignment among technical and operational management aspects were required for the successful procedure, involving multiple cycles for hydrate dissociation and subsequent recovery of the foam PIGs.
Natural gas is the backbone of many electrical grids as it is used in both efficient combined cycle generation plants and for its renewables firming capability in peaking turbines. Further, it is a critical feedstock for the chemical processing industry, including fertiliser manufacturing, and has no alternatives for the foreseeable future. The development of new gas production capacity is therefore key to ensuring a cleaner, stable, and secure supply of energy, as recognised by its continued contribution in all plausible net-zero 2050 scenarios. New gas fields are increasingly located offshore in deep water, where their production lines are exposed to a range of flow assurance hazards. Of particular concern is the formation of hydrates – ice-like solids which may deposit on the pipe wall, eventually sloughing and collecting at low-points or bends.
The mainstay of hydrate mitigation historically has been avoidance through the injection of thermodynamic inhibitors (THIs) which disrupt the hydrogen bonds of their crystalline lattice, effectively shifting the equilibrium region. Significant volumes of THI are required for this process, where the distances and scale of modern subsea developments means that their use threatens the economic viability of projects. Various alternative chemical strategies, such as the use of kinetic inhibitors or anti-agglomerants (AAs), provide a nominally lower cost pathway for hydrate management: these have seen limited uptake globally. In addition, innovative new technologies have been developed around physical methods of managing hydrate severity, in particular: subsea separation and innovative materials. Subsea separation, from a hydrates management perspective, acts as a limiting reagent solution to the interaction of water and light gas molecules, while new thermoplastic composite materials have demonstrated a low affinity for hydrate deposition.
Concurrent with these new physical technologies, progress has been made in the modelling of hydrates for both gas and liquid dominated systems. The OLGA® Extensibility framework now enables the flow assurance effects of hydrate formation on the mass, momentum and energy conservation equations to be simulated. Ultimately this allows simulations to directly link a given degree of hydrate formation to an expected change in production rates, that is, a direct economic benefit can be computed from system designs which allow a limited degree of hydrate formation, while still avoiding blockages. In this work, we make use of this new capability to explore the synergistic nature of new chemical and physical methods in managing hydrate formation and deposition without the use of thermodynamic inhibitors. We demonstrate the potential for coupling a Pseudo-Dry-Gas in-line separator with a thermoplastic composite liquid return line dosed with anti-agglomerants. The gas and liquid return lines are analysed separately using computational tools appropriate for the differing flow mechanics in each, and the potential savings in both cost and emissions reduction are quantified.
Hydrates - Modelling and Experiments
Gas hydrate is one of the major issues in flow assurance in gas and oil production. The hydrate formation in the pipelines impacts on the flow, increasing the pressure drop and the risk of pipeline blockage. The hydrate formation leads to macromechanisms such as:
agglomeration, deposition, bedding and slurry. Depending on the thermodynamic and flow conditions, different hydrate macromorphology is formed. In this work, measurements of gas hydrate wall deposition in gas-oil-water-hydrate systems under sheared conditions are investigated. Experiments are performed for synthetic natural gas,mineral oil and fresh water. Experiments are carried out in condition of water content ranging from 10 to 70%, liquid loading from 40 to 80 vol% and subcooling from 2.25 to 11.93. Images from the rock-flow cell are used in order to perform a phenomenological analysis of the gas hydrate deposition. Furthermore, a proposal of the agglomeration and deposition mechanisms is given. This work highlights two relevant fact for the hydrate
deposition process: (a) the importance of the hydrodynamic of the fluids before the hydrate formation, including the phase distribution, the size of the water droplet, and the shear rate impact on the macromechanism of hydrate deposition and, (b) the existence of
free water is the most important parameter to consolidate the hydrate wall deposit under flow.
Gas hydrates pose a critical challenge to flow assurance because of their rapid formation and potential for pipe blockage. One management approach is to use anti-agglomerants (AAs), thus allowing hydrate formation while preventing agglomeration and making the hydrate particles to flow within the liquid as a slurry. Despite the importance of this topic, there is scant literature available on the impact of these particles on key flow parameters such as pressure drop, liquid holdup and overall flow topology. The aim of this study is to address these gaps, with a special focus on the stratified and slug flow patterns.
Natural gas hydrates corresponds to one of the most serious flow assurance issues in Oil & Gas production and transportation. Although significant research efforts have been undertaken in the past decades, considerable challenges remain in order to attain a more robust understanding and modeling of the evolution of hydrate crystals in multiphase systems, as well as simulation tools for production pipelines. The present study’s goal is twofold. It aims at implementing and verifying an approach based on a population balance framework that relies on a topological assumption of hydrates particles as formed by a porous media whereas a porosity evolution equation is considered for a given subcooling degree. Another consideration is that colliding particle pairs interact with each other by forming capillary liquid bridges, turning into aggregates that may stick together as agglomerates by undergoing a consolidation process or remain as separate entities with the bridge’s disruption due to shear interactions with the flow. The models have been verified against flow-loop data, as well as a rock-flow cell apparatus, both operating oil-dominated systems. Further developments allowed testing the approach by coupling with a 1-D multiphase hydrodynamic and heating transfer solver. Representative field case studies, including comparisons with existing hydrate model frameworks, showed promising results for the prediction of thermohydraulic profiles and hydrate-related quantities.
Gas hydrate formation in multiphase oil and gas pipelines causes flow assurance problems, especially in deep water subsea tiebacks. Agglomeration and bedding of hydrate particles and its deposition on pipe walls may form blockages that stop production. This work will cover the systems approach that we, at the Center for Hydrate Research of the Colorado School of Mines, have followed to design mathematical models to predict gas hydrate formation in multiphase flow pipelines, making emphasis on the development of models for transient operations. Transient operations with substantial shut-in times allow the system to cool down within hydrate formation conditions and have increased probabilities of forming hydrate blockages. Our model development begins with the presentation of a series of conceptual pictures that describe the main phenomena related to hydrate formation in oil and gas transportation pipelines during transient operations. These phenomena have been observed in multiscale experiments performed at our laboratory facilities and in other collaborating facilities. Then, we provide an overview of the recently developed models for dispersed (oil-dominated with water dispersed as droplets) and segregated (partially dispersed with a continuous water layer) systems, along with field case examples, highlighting new physics that improve predictions of hydrate phenomena during transient shut-down and start-up operations. With the field case studies, we demonstrate the application of these models in addressing hydrate-related challenges in multiphase pipeline operations.
CO2 Flow in Pipes and Wells
CO2 Phase Behaviour
CO2 coming from capture processes is generally not pure and can contain different kinds and quantities of impurities. The presence of these impurities greatly impacts the single and multiphase flow behavior of the CO2 stream and could lead to challenging flow assurance issues. As a result, it is essential to account for the impact of the impurities on the design and operation of CO2 injection systems. This can be assessed by applying advanced thermodynamic and thermo-hydraulic models which can model the multiphase behavior of the system during normal and transient operations.
Fracture control is an integral part of the design of a pipeline and is required to minimize both the likelihood of failures occurring and to prevent or arrest long running brittle or ductile fractures. The saturation pressure of the fluid, and hence the toughness required to arrest a fracture, increases as the concentration of impurities increases. Impurities also have a major impact on the fluid decompression wave velocity, which alters the decompression path to the saturation pressure and ultimately the material toughness requirement. Poor understanding of fluid and Equation of State (EoS) selection leads to risk of an inadequate fracture control plan.
In this study, we address the impact of impurities on the physical behavior of CO2-rich systems and consequently on saturation pressure and fluid decompression velocity. The modeling results have been compared against extensive literature data on decompression behavior of pure CO2 and CO2 with impurities. Current published literature shows a comparison of experimental data against different EoS predictions and highlights that there are differences between predictions and measurements. However, to our knowledge efforts have not been made to tune EoS models to match decompression behavior. In this study, we have proposed tuning methods for binary mixtures of CO2 and impurities and adopted those corrections for multi-component mixtures.
In this work we will look at the performance of equilibrium and transport property models for representing the general phase behavior and phase properties of CO2-rich streams relevant for CCS. We will start by analyzing the effect of impurities on the phase behavior of these mixtures, including the bubble point curves and the VLE critical points, which are required to design single phase transport. In addition, we will also consider the possibility of formation of other liquid phases (LLE), including water drop out which is particularly important for corrosion assessment and control. In a third section we will investigate the potential formation of solid phases such as hydrates and other pure solid phases such as ice or dry ice (solid CO2) which can lead to flow assurance problems. For this evaluation we will use standard equation of state models, such as the Cubic-plus-Association (CPA) equation of state (EoS), but also other more specific and accurate models such as GERG and its variants (EoS-CG). Cubic EoS models, such as CPA are recognized for their relative speed and robustness, albeit with compromised accuracy in density and thermal properties. These stand in contrast to the High Accuracy EoS models, revered for their precision in properties like density and phase behavior, but plagued by sluggishness and reduced robustness beyond their applicability range. The second part of this work will involve an assessment of models to describe the phase properties more relevant for designing CCS facilities (density, viscosity, heat capacity, Joule Thompson coefficient, compressibility, expansivity).
CO2-rich systems are becoming more prevalent around the world, including the re-injection of CO2 into the reservoir. CO2-rich systems pose challenges for flow assurance as CO2 can readily liquify and form at conditions of higher temperature/lower pressure relative to common natural gas systems. There exists a very limited amount of data CO2 phase changes and flow behavior with and without water, especially with nearly pure CO2 and with liquid CO2.
This contribution will present results for measurements and observations that quantify the phase behavior and flow behavior CO2-rich systems in support of the current needs for CCS (carbon capture and storage). A total of five systems are considered: pure CO2 and four CO2-rich mixtures with contaminants, which included N2, O2, H2, C1, C3. For the mixtures, the CO2 concentration was 95 mol% and the contaminants added to 5 mol%. The experiments are performed in the rock-flow cell setup, which allowed for controlled temperature and pressure under shear (rocking). Videos/images from the in situ borescope in the cell allowed for direct observations of the phase changes and flow behavior. The phase behavior/transition covered the entire phase space capturing vapor, liquid, vapor-liquid, and supercritical fluid.
Results will show the deviation of the bubble curves to predictions as well as the process associated with the phase changes as temperature and pressure are varied. It is shown that the phase changes can be impacted by the shear conditions and they do not happen precisely along thermodynamic boundaries. Water impacts the phase changes as well including the formation of gas hydrates, which are significantly different in CO2-rich systems.
The energy industry has decades of experience with modelling the multiphase flow of gas, oil, and water through wells and pipelines. This has led to state-of-the art design tools like OLGA and LedaFlow. Currently, the industry is working on upgrading the multiphase flow tools, to make them reliable for the design of CO2 transport. A particular challenge is modeling CO2 transport under two-phase flow conditions, which occurs during transient operations and even during steady-state operation for the case of injection into depleted gas reservoirs. This paper presents an accurate numerical method to solve the Homogeneous Equilibrium Model for single component CO2 transport under single-phase and two-phase flow conditions. The model performance is demonstrated through several test cases featuring fast transients. The same test cases are also simulated with OLGA, and a few of them also with LedaFlow and WANDA. The model and the test cases can be used as a basis for further developing, testing, and optimizing the numerical methods used to model CO2 transport. Our study shows that state-of-the art simulation models can provide numerically reliable solutions for fast transients with two-phase CO2 flow in pipes.
Pipelines carrying either CO2 or H2 will have increased risk of leakage as compared to conventional natural gas; CO2 because of Corrosion and H2 because of embrittlement. The most often discussed plan for H2 transportation is to mix it with Natural Gas and use existing pipeline networks. So, the H2 can be considered an impurity and will have limits on what composition is acceptable for embrittlement. For CO2 pipelines, there can be various impurities which can lead to increased corrosion and some CO2 Hubs will have up to 90 emitters and 10 different types of emitters. So, because of variations in temperature/pressure/flowrate/impurities the risk of leakage will vary along the length of the pipeline for both pipeline types. Both model-based and mass compensated LDS depend on accurate flowrate measurements at both ends of the network. However, predictions of density for both type of systems can have errors in the range of 2 to 10% which then directly affect flowrate measurements. The possible errors are large since accurate measurement of some impurities is difficult (including H2) and because of issues with the Equations of State (for H2 – large size difference between H2 and Methane and CO2 because near the critical point). This inaccuracy in density also effects the hydraulic modeling of the pipeline to predict pressure drop (which is a key signal for model-based leak detection). H2 Pipelines will be operated at higher velocities and have increased pressure drop. So, if you have H2 Composition that is varying in each pipeline between 0 and 20%, various segments will have different amounts of DP (pressure drop). Typical CO2 Pipelines will be operated in the dense phase region, however when a leak develops the flow will become multiphase. The conditions where/when this occurs will be dependent on the impurities and the temperature in the pipeline. For all of these reasons, in areas with higher probability of leakage and/or high impact areas, other types of LDS will be required. So, most networks may require 2 or 3 different types of LDS to get maximum coverage.
CO2 coming from capture processes is generally not pure and can contain different kinds and quantities of impurities. The presence of these impurities greatly impacts the single and multiphase flow behavior of the CO2 stream and could lead to challenging flow assurance issues. As a result, it is essential to account for the impact of the impurities on the design and operation of CO2 injection systems. This can be assessed by applying advanced thermodynamic and thermo-hydraulic models which can model the multiphase behavior of the system during normal and transient operations.
Fracture control is an integral part of the design of a pipeline and is required to minimize both the likelihood of failures occurring and to prevent or arrest long running brittle or ductile fractures. The saturation pressure of the fluid, and hence the toughness required to arrest a fracture, increases as the concentration of impurities increases. Impurities also have a major impact on the fluid decompression wave velocity, which alters the decompression path to the saturation pressure and ultimately the material toughness requirement. Poor understanding of fluid and Equation of State (EoS) selection leads to risk of an inadequate fracture control plan.
In this study, we address the impact of impurities on the physical behavior of CO2-rich systems and consequently on saturation pressure and fluid decompression velocity. The modeling results have been compared against extensive literature data on decompression behavior of pure CO2 and CO2 with impurities. Current published literature shows a comparison of experimental data against different EoS predictions and highlights that there are differences between predictions and measurements. However, to our knowledge efforts have not been made to tune EoS models to match decompression behavior. In this study, we have proposed tuning methods for binary mixtures of CO2 and impurities and adopted those corrections for multi-component mixtures.
Carbon capture, utilization, and storage (CCUS) is instrumental in reaching net-zero CO2 emissions by 2050. One of the CO2 utilizations in oil field operations is in the single-point injection gas lift, in which the gas unloading valves are eliminated, and the well is unloaded through the operation injection valve. However, single-point unloading requires high kick-off surface pressure during injection to overcome the hydrostatic head in production tubing. Alternatively, supercritical CO2 can be used to unload the well and lift the oil with low surface injection pressure due to its unique physical properties of liquid-like density and gas-like viscosity. The objective of this study is to investigate supercritical CO2 as an unloading and lifting fluid in a single-point gas lift operation through simulation using commercial and academic mechanistic models. The simulation models calculate CO2 pressure and temperature profiles along the tubing/casing annulus to predict the surface injection pressure. To illustrate the applicability and novelty of using CO2 in single-point well unloading, this study predicts and compares the required surface kick-off pressure using CO2 and Methane in a single-point injection unloading operation. The simulation prediction shows that for a 10,000-ft deep vertical onshore well, single-point injection gas lift using supercritical CO2 reduces the required surface kick-off pressure by approximately 3000 psi compared with that of Methane. Moreover, using an offshore case study with a 10,000-ft vertical well and 2000-ft water depth, the simulation predicts a 3300-psi reduction in kick-off surface pressure using supercritical CO2 compared to Methane. The advantage of using CO2 in unloading and lifting oil wells is not only to promote sustainability through CCUS, but also to eliminate conventional unloading valves, which minimizes well interventions and production downtime in the case of valve failure, especially in offshore operation.
Carbon, Capture & Storage (CCS) is an unavoidable technology to reduce shot and medium term Carbon Emissions. In the Dutch sector, there are several initiatives to store CO2 in depleted gas fields. Given that these fields are at low pressure, and considering that transport is more favourable at high pressure, a significant pressure drop is expected over the valves that control the pressure and mass flow during injection. As a result, a phase change occurs across the valve, leading to a substantial temperature drop along the phase boundary of CO2 that potentially affects the material of the walls and equipment downstream. Predicting the actual temperature profile and heat flux is a topic of current research as it involves simulations of turbulent multi-phase flow including phase change.
This paper explores three methods to model the expansion of supercritical CO2 in a nozzle: in order of complexity these are the isenthalpic model, the 1D Euler model and the Enhanced Mass Transfer (EMT) model which was set-up in ANSYS-Fluent. These models are compared against each other and to the experimental result by Nakagawa et al. [1][2]. The results showed that the most extensive model demonstrated the best agreement with the experimental data, particularly for the pressure data. The mass transfer mechanism incorporated in this extensive model shows the best technique for modelling cavitation. The onset of evaporation downstream of the throat of the nozzle has a minor impact on the total mass flow. The Euler and the ETM models are also able to capture the shock wave. The EMT model is the only model which accurately predicts the minimum of the pressure downstream of the nozzle. Although the results show a good agreement, there is still a room for improvement. Especially, including accurate PVT data near the critical point presents a challenge.
For Carbon Capture and Storage (CCS) systems, accurate multiphase flow predictions are important to achieve safe and cost effective design and operation. The multiphase flow simulator LedaFlow was originally developed for hydrocarbon production systems, but due to the rapid emergence of CCS technology, LedaFlow is currently being adapted to scenarios and conditions relevant for CO2 injection.
In CCS applications, the gas/liquid density difference tends to be smaller than in typical hydrocarbon production systems, and the gas/liquid surface tension can also be very small. Both these factors increase the degree of gas bubble entrainment into the liquid, and measurements show that the gas bubble fraction in the liquid film can become very high in stratified two-phase flows with CO2. In LedaFlow, bubble entrainment in stratified flows is presently not modelled because the overall effect of bubble entrainment on typical oil/gas scenarios is usually limited. It has however become clear that bubble entrainment needs to be included in LedaFlow to accurately model CO2-rich flows.
This paper outlines the derivation and implementation of a new gas entrainment model in LedaFlow for stratified flows in near-horizontal pipes. The model is based on a balance between turbulent diffusion and gravitational drift, mainly using validated closure laws from the public literature. The bubble size model was calibrated by fitting the model to experimental bubble concentration profiles in bubbly flows. Comparisons between the model predictions and the respective measurements shows very good agreement with the bubble entrainment data for data covering several different pressures and two different pipe diameters. Furthermore, we observe that the flow regime predictions and pressure drop predictions improve with the introduction of the new gas entrainment model in LedaFlow.
Multiphase Flow
An experimental campaign was conducted in Equinor’s high pressure flow loop in Porsgrunn with the objective of establishing a high-quality real gas condensate data set for use in modelling steady state flow as well as the specific surge wave phenomena in the Mikkel and Midgard (part of Åsgard) fields on the Norwegian continental shelf. The Mikkel and Midgard fields have been naturally flowing to Åsgard B since 2003 with introduction of the Åsgard Subsea Compression station in 2015 as a major upgrade. The surge wave phenomenon is per today observed in flowlines both upstream and downstream the compressor station, representing challenges with respect to liquid level control both in the scrubber upstream the compressor station as well as topside Åsgard B. Several flowlines are operated with so called ‘’minimum flow limitations’’ representing the minimum production rates at which the flowlines can be operated with acceptable liquid content and flow fluctuations due to surge waves. In order to plan future operations and restructuring initiatives, predictive tools able to capture these phenomena from a quantitative point of view are required. Data based on realistic fluid properties and conditions, have been vital for achieving the required flow simulator improvements in close collaboration with the software vendor modeling team. The first part of the paper gives a detailed description of the experimental matrix, the experimental execution, and the data analysis. Emphasis is also given to describe the state-of-the-art instrumentation used to achieve a unique set of high quality three-phase pipe flow data. The second part of the paper describes a modelling effort utilizing User Defined Functions in the multiphase flow simulator to tune/modify the closure laws to better match the experimental data. The improved model is compared with the Åsgard field data. The model will be used to aid future operations of the field. In an accompanying paper the software vendor has used the new data to improve the general multiphase flow simulator model.
The flow dynamics in multiphase pipeline flow is generally influenced by parameters such as fluid properties, flow rates, pressure, pipe diameter and inclination. While the general topology of such flows is extremely different to model precisely, modelling efforts often require a rough categorization of topology by means of flow regimes. The type of flow regime plays an important role in pipe flow modelling, since it strongly influences computed results on holdup, pressure drop and flow structures. In pipeline simulators, the flow regime is typically determined from slip velocity and slug fraction calculations, or from quasi-empirical correlations. In some parameter regimes such schemes are imprecise and need to be constrained by experiments. In a lab setting, the flow regime has traditionally been identified by visual inspection of high-speed optical videos or holdup time series. This method, however, can be both expensive, imprecise, and subjective.
We have developed a machine learning scheme which automatically determines the flow regime from measurements of liquid concentration levels, obtained by gamma, X-ray or other measurements. The schemes are currently able to distinguish between several flow regimes; e.g., slug flow, stratified and large wave flow with good accuracy. We are working on a hybrid scheme accounting for physics-based constraints as well as correlations between X-ray data, gamma densitometer data, and other sensor types of data (e.g., pressure), with the aim of identifying flow regimes from data in a field scenario.
The studies of two-phase liquid-gas flow have mostly been limited to horizontal or vertical upward geometries over the years. However, downward multiphase flow in vertical pipes is gaining attraction in various industries, such as injection wells, refrigeration cycles with refrigerant phase changes, chemical distillation towers, cooling towers in nuclear reactors, heat exchangers, and CO2 sequestration systems. The purpose of this study is to review, analyze and connect findings from the available two-phase flow studies in downward vertical pipes, especially in the presence of CO2. The experimental works use various methods like wire-mesh sensors, gamma densitometers, and quick-closing valves to identify the flow patterns and measure the flow characteristics. Some studies utilize theoretical models to decipher the complexities of downward two-phase flow. At this study, machine learning techniques are employed to predict downward two-phase flow, based on the data from 9 previously conducted works. The available data identify churn-turbulent, bubbly, and annular flows as the main downward flow patterns. Flow pattern maps differ between large (>101.6 mm) and small (<51 mm) diameters, particularly for the case of slug flow. A relationship is observed between void fraction and flow
patterns. Notably, frictional pressure drops are higher in annular flows. Comparatively, some models predict experimental results within a 10% error, while others show large discrepancies, especially for slug flow and larger diameters. The drift-flux model’s accuracy decreases beyond a 0.83 void fraction. The homogenous model is optimal for predicting pressure drops in two-phase CO2 flows at lower void fractions. Machine learning techniques are introduced as an improved approach for flow pattern identification, based on the available data. Despite the recent advances, knowledge gaps persist, notably in prediction of interfacial shear, pipe diameter effects, and flow pattern observation for CO2. There is a need for comprehensive models for the entire spectrum of flow patterns.
In this study, we assess the accuracy of two leading commercial multiphase flow simulators in reproducing field data from Offshore Production Facilities (OPFs) located in the Parque das Baleias field, northern Campos Basin, which is part of the Brazilian pre-salt cluster. Simulator predictions are initially benchmarked against a dataset of over 250 quasi-steady pressure drop data points from six production pipelines and tubing. The results are further compared with those from our in-house computational model. All models, both in-house and commercial, were standardized with identical boundary conditions and operational parameters, such as thermal conductance for casings and insulating materials, and surface roughness for pipelines and tubing. They were also furnished with consistent PVT and fluid property data, derived from a thermophysical property calculation framework [1], and verified by laboratory measurements of actual reservoir fluids. The second part of the paper contrasts the commercial simulators’ efficacy in forecasting transient operations, such as well shut-in, cool-down time estimation, pipeline depressurization, and well restart.
[1] T.R. Gessner, J.R. Barbosa Jr., Integrated Modelling of Reservoir Fluid Properties and Multiphase Flow in Offshore Production Systems, Springer (2023) DOI: https://doi.org/10.1007/978-3-031-39850-6
Flow Evolution and Transient
Slack line occurs in oil pipelines when the local pressure drops below the vapor pressure of the oil, resulting in gas breakout and multiphase flow in that region. By contrast, ‘tight line’ exists where the local line pressure remains above the vapor pressure, and the liquid remains single phase. The presence of both single-phase and multiphase flow in different parts of the line, as well as the proximity of the vapor pressure to a zero-pressure condition, makes transient simulation of slack flow particularly problematic.
A simple method, using steady-state analysis, is presented for identifying the location of slack regions in an oil pipeline. In addition, a function for liquid holdup determination in the slack region is developed and implemented into the steady-state model. Lastly, a transient model, based on the Ellipsis multiphase solver, is presented which covers both the appearance and disappearance of slack line during pipeline transients.
Tyrihans is an offshore field in the Norwegian Sea that produces oil and gas to the Kristin platform via a 43 km long pipeline that is connected to four production manifolds. Due to the high production flow rate, which generates gas velocities above 10 m/s, the flow in the pipeline is operated under frictional-dominated conditions. Nonetheless, flow instabilities, resulting in liquid surges arriving at the platform topside at high velocities, can be observed.
Several field data series including unstable flow conditions have been collected. To better understand surge flow instabilities, the company and vendor have developed a surge wave flow simulator in a technology cooperation. The simulator can qualitatively replicate the observed instability, generating waves that increase and decrease in amplitude. In many of the flow cases, a regular high frequency instability is observed. In other cases, the flow remains relatively stable until an infrequent irregular instability occurs, resulting in liquid surges at the receiving facility.
Various numerical and model parameters influence the results of the simulation, and these are discussed in detail.
Slug flow presents a distinct modeling challenge due to its significant spatial changes. In large-diameter and long pipelines with many undulations, which are often found in the industry, the two-phase flow may never fully develop. In this study, we collected experimental data to characterize the evolution of the intermittent flow in a 2° upward inclined, 6-in ID, 85 m length pipe at an average pressure of 217 psi. Data comprise pressure and temperature, mass flow rates, differential pressure, and capacitance from nine probe stations. Translational velocity shows lower values at the entrance of the test section and reaches a higher plateau after crossing the middle part of the test section. Slug frequency shows higher values at higher superficial liquid velocities. Slug frequency decreases along the test section and reaches a plateau after passing the middle part of the test section. For the presented results, slug length does not show a defined trend at the entrance of the test section. On the other hand, both liquid film and unit cell lengths show a linear trend at the beginning of the test section. Results will contribute to the evaluation and improvement of the available slug and pseudo-slug flow models and closure relationships.
In oil production, the prevalence of multiphase flows involving oil, water, and gas phases is very common, in various sections of pipelines. Measuring the fractions of each phase is a challenge in both industrial and academic applications, as it requires the use of highly specialized equipment. In this context, our research endeavors to introduce a novel tool designed to compute local volumetric fractions in a transient three-phase flow from images of this flow. This tool consists of in-house Python codes based on convolutional neural networks within the scope of machine learning. These codes process the flow images through a U-NET to identify/segment the phases, and subsequently estimate the fractions by incorporating a third dimension to the analysis. To assess and validate the tool, we conducted visualization experiments using a high-speed camera in a transparent 3-inch tube mounted on an oscillatory apparatus. Here, a transient three-phase air-water-kerosene flow was investigated under varying operational conditions, encompassing oscillation frequencies, volumetric fractions, and dye concentrations. As a result, the neural network successfully quantified local volumetric fractions in the tube with a high degree of accuracy. As part of our ongoing efforts, the team is actively enhancing the multilayer neural network’s ability to discern bubbles and droplets embedded within the phases as well. This improvement aims to provide comprehensive metrics, including the quantity of bubbles and droplets, diameter distribution, and other relevant parameters. In summary, the tool proposed in this study introduces a new possibility for the quantitative analysis of intricate flows in agreement with the growing trend of artificial intelligence usage within the oil and gas sector.
Within the crude oil tankage facilities, above-ground piping is preferred due to ease of inspection. However, when the pipeline feeding tanks is mounted on a rack at lower tank levels/lower flow rates, there is a significant potential for column separation at high points. Liquid column separation can occur at high points of liquid pipelines when the pressure drops below the vapor pressure. The liquid column separation introduces a huge risk for vapor collapse due to transient events in the pipeline. In this study, we first assessed the risk of liquid column separation at high point of a light crude oil pipeline. Then, we simulated the risk of vapor collapse due to transient events and quantified the transient forces and the likelihood of failure. The approach to study this used a multiphase hydraulic model, Computational Fluid Dynamics (CFD) model and Finite Element Analysis (FEA).
Slugs
This paper investigates the flow management challenges within a deepwater oil field operated by TotalEnergies, situated in water depths between 1.2 and 1.5 kilometers off the African coastline, with a specific focus on the impact of pipeline corrosion inhibitor injection on field production. To prevent corrosion in the flowline network, corrosion inhibitor is injected continuously, leading to unstable flow in some of the risers. To better understand the impact of corrosion inhibitor on flow stability, different field trials of stopping and restarting the corrosion inhibitor injection have been conducted.
The unstable multiphase flow, characterized by slugging, leads to operational issues such as increased flaring and compromised separation processes. To accurately model these conditions, the study leverages a detailed fluid characterization, LedaFlow’s advanced Slug Capturing Method, and a fitted emulsion viscosity model based on fluid rheology data.
Comparative analysis of simulation results with field data indicates that LedaFlow’s dynamic simulations, coupled with an adequate fluid characterization, effectively capture the slugging phenomena, offering valuable insights into the fluid dynamics within the field’s complex subsea environment. The alignment of model results with actual field data reinforces confidence in the software’s ability to predict complex fluid dynamics, which is instrumental for design testing of new transport systems and informed decision-making in slugging mitigation strategies.
Due to the high variation of momentum and density in pipeline spans, slug flow has been identified as a concerning two-phase flow pattern with regard to fatigue life reduction. This flow pattern can easily arise in pipelines and affect many industries. However, due to the complexity involved in accurately modeling the impact of the flow on fatigue life, few modeling studies have been conducted, and fewer are verified with experimental data. Despite this, numerical modeling and simulation remain the most realistic approaches for understanding this field since the scale, environment, and geometric configurations of interest are difficult to reproduce experimentally.
This study presents theoretical models for predicting fluid flow, structural response, and fatigue life, with benchmarked results against experimental data. The results reveal the importance of accurately modeling the fluid parameters of slug liquid holdup, film liquid holdup, and slug frequency. The contribution of gravitational and centrifugal forces is sufficient to account for the fluid forces acting on the structural system. The numerical model benchmarking indicates that the one-way coupling and the assumptions adopted in the models are reasonable simplifications and can accurately reproduce the fluid-structure interactions for the flow conditions analyzed in this study (which did not produce forcing frequencies in resonance with the structural frequencies). The sensitivity of flow pattern parameters is assessed by comparing two different fluid models. It is concluded that a theoretical model can accurately simulate a horizontal pipeline fatigue life surface for a range of flow velocities and provide guidance on the critical flow conditions that may induce fatigue for the given configuration.
This work describes a comprehensive experimental study on the flow characteristics of high-density gas-liquid systems with a focus on slug flow. Oil and sulfur hexafluoride (SF6) were employed as model fluids. The experimental loop consisted of a 60-m long, 1-inch diameter pipe. The experimental procedure involved the controlled injection of oil and SF6 into the pipe to mimic the conditions of high-density gas-liquid flows, with five measuring stations along the loop. Each measuring station has a pair of conductive
sensors, a pressure and a temperature sensor. The frequency of slug formation, bubble and slug lengths and bubble velocities were measured with the conductive sensors and the images obtained with a high-speed camera. By systematically varying parameters such as the flow rate and the pressure, the study explored the intricate details of each one of these slug flow characteristics. These parameters serve as crucial indicators of the flow dynamics and offer valuable insights into the underlying mechanisms governing the flow.
Furthermore, the investigation revealed significant findings regarding the influence of operating conditions on the observed flow pattern, helping to understand the pattern transition between intermittent and stratified flow. The study also provided a detailed analysis of the associated phenomena. The flow pattern transition and the slug flow parameters found experimentally were also compared with previous works, correlations and models. The deviations found were analyzed and discussed. Moreover, the
experimental results presented in this work further the knowledge on multiphase flow behavior, with direct implications for industries such as oil and gas production and transportation. The insights from this study may contribute to the development of more accurate models for predicting and optimizing high-density gas-liquid transport systems, ultimately enhancing the efficiency and reliability of industrial processes. In addition, models for the pattern transition can be optimized for these conditions.
Churn flow is regarded as one of the most challenging liquid-gas flow patterns to comprehend. The literature reveals a gap in churn flow modeling, especially at low liquid rates found in natural gas wells. Commonly available models, like OLGA and TUFFP unified, consider a direct transition between annular and slug flow patterns. These models provide unsatisfactory predictions for pressure drop and liquid holdup in the churn region. Slug flow is assumed in these models to predict the behavior of churn flow. Experimental observations reveal that churn flow behaves differently than slug flow, especially at low liquid rates. This demonstrates the importance of developing a model to predict the churn flow behavior. This study develops a 1-D mechanistic model based on the mass and momentum balances to predict the pressure drop and liquid holdup of liquid-gas churn flow in vertical tubulars. The model considers the effect of droplet entrainment in the gas core. Various available interfacial friction factor (fi) and entrainment closure models are tested. Olieman’s entrainment correlation provides the best performance for oil-air churn flow. Jayanti and Hewitt’s fi correlation results in the best performance for churn flow. An improved correlation is developed based on this correlation to predict fi using the experimental data of water-air and oil-air flow. The correlation includes the effect of liquid properties on interfacial shear stress. The proposed mechanistic model with the developed fi correlation outperforms the other tested models, particularly at low liquid rates and for oil-air flow. This study provides a more accurate tool to predict liquid-gas churn flow in vertical pipes. The results can play a critical role in understanding multiphase flows in natural gas wellbores, nuclear facilities, and renewable energy production systems.
Solids
Due to the lack of mechanical moving parts, airlift pumps are utilized in many industrial applications for transporting liquid solid mixtures, corrosive or toxic fluids, sludges, or other complex fluid mixtures. The pump’s performance is dependent on the pump design parameters, such as the pump riser length and diameter, as well as the physical properties of the carrying liquid or mixture, and the characteristics of the solid particles when operating under three phase flow conditions. Understanding these pumps’ behaviour is key for optimizing their energy efficiency. The present study experimentally investigates the performance of an airlift pump operating under different liquid phase densities in two-phase flow conditions and evaluate its performance when handling three-phase mixtures of gas, liquid, and solids. Critical performance parameters such as pumps efficiency for operating under two phase flow conditions as well as effectiveness for three-phase cases are measured and discussed. Under two phase conditions, the pump is used to lift water with a density increased from 1000 to 1100 kg/m3 using water-salt solutions. Also, the pump is tested under three phase conditions, using 5 mm ceramic spheres entrained in the lifted water with a mixture density up to 1094 kg/m3. The pump utilizes an annulus air injector design and is tested at a constant submergence ratio of 0.7. Flow visualization using high speed imaging is used to understand the gas, liquid and solid distributions in the pump riser and evaluate the effect of these distributions on the pump performance. Lower liquid phase densities found to improve the pump efficiencies as high as 19% at inlet air flow rate of 14 LPM. As the liquid density increases, the flow patterns in the pump riser changes from slug to churn flow reducing the buoyancy forces causing the efficiency to drop. On the other hand, under three-phase flow conditions, the maximum effectiveness of the pump while lifting solid particles found to be approximately 11% at a mixture density of 1094 kg/m3 and inlet air flow rate of 41 LPM. Also, under the three-phase flow conditions, solid particles deliverability improves on the expense of the amount of lifted liquid. These performance criteria are presented in term of Stokes and Weber numbers to evaluate the characteristics of solid particle suspension in the liquid and its relation to the inertia or surface tension forces encountered in the three-phase flow. The results of this research can be used for optimizing the pump performance for lifting slurries and provide a broader understanding of the hydrodynamic behaviour in multiphase systems.
Asphaltene deposition in oil wells is a complex and challenging phenomenon, with poor understanding and prediction capabilities. This study aims to develop a mechanistic and integrated well model to predict asphaltene precipitation, aggregation, transportation, and deposition in time and space along oil wells. In addition, this work experimentally analyzes and models asphaltene particles/aggregate growth, which is incorporated into the integrated model to investigate its effect on the asphaltene deposition behavior. Additional objectives of this work are investigating the effect of various well operational parameters, fluid physical properties, flowing temperature, and flow hydrodynamic characteristics such as flow pattern on asphaltene deposition behavior. A database of crude oil asphaltene onset pressure (AOP) measured by dynamic light scattering (DLS) and solid detection measured by near-infrared red (NIR) light transmittance techniques is analyzed and modeled for particle and aggregate size using advanced computer vision techniques. Preliminary results show a significant aggregate growth rate of approximately up to 470% as pressure decreases below the AOP. Incorporating these results into the integrated asphaltene deposition model resulted in an improved model, which accounts for asphaltene aggregate growth. A field validation study revealed a close match between caliper log deposition thickness measurement and model prediction along an oil well produced from a Jurassic reservoir located in Kuwait. Furthermore, sensitivity analysis was carried out and showed that asphaltene deposition is significantly sensitive to asphaltene particle/aggregate size, indicating its importance in modeling asphaltene deposition behavior. In addition, the deposition layer’s physical and thermal characteristics are found to be affected by the deposited aggregate size, which in turn influences the thermal and hydrodynamic behavior of the flow, thus the asphaltene deposition behavior. Moreover, the fluid flow temperature effect on asphaltene precipitation and deposition is investigated and preliminary results show an appreciable effect under specific conditions.
Computational Fluid Dynamics (CFD) has become a common modelling tool for erosion prediction for complex and uncommon geometries, especially when prediction of erosion pattern is important. However, there are multiple challenges involved in CFD-based erosion modelling. Erosion under multiphase flow conditions can be observed frequently, which will bring additional challenges in particle tracking for multiphase flow to capture effect of liquid damping and solid shielding. Depending on multiphase flow regime, fluid-particle interaction can significantly affect particle tracking when particle travels from one phase to another. The presence of liquid layer on the pipe surface damps out particle momentum and particle impact velocity, which will significantly reduce erosion damage. This paper presents an algorithm which improves particle drag model that properly accounts for multiphase effect in particle trajectory and effectively captures liquid cushioning. This algorithm allows the volume fraction weighted mixture velocity and properties of the continuous phase to be used for particle drag calculation. It will significantly improve the erosion predictions, not just the value of peak erosion rate but also the erosion pattern, under multiphase flow conditions. Multiphase CFD modelling was validated for erosion rate and erosion scar location on elbows against published flow-loop data for slug/churn as well as low liquid loading annular multiphase flow regimes. This paper will present results of these validation studies.
Liquid-Liquid Flow
The flow of non-miscible liquids is frequent in many industrial applications and multiple scales, from microfluidic systems to oil transport in subsea production wells. The complexity of two-phase flow often limits rigorous numerical modelling of liquid-liquid flow in actual practice. One of the challenges is the prediction of velocity profiles, which are related to the average velocities of the phases, pressure drop, and heat transfer. Another issue is the values of the shape factors, which may be related to flow pattern transition. We propose experimental correlations as functions of the phases’ Reynolds numbers based on a Couette-Poiseuille velocity profile and the separate cylinder model to predict the velocity profiles at the diametrical vertical plane. A one-dimensional liquid-liquid flow model is used to calculate the in-situ average quantities of the flow. The proposed velocity profiles are validated against new experimental data obtained in a horizontal stratified oil-water pipe flow via particle image velocimetry (PIV) in two pipe diameters and data reported in the literature, considering a fair range of Reynolds numbers. The results indicate that increasing both Reynolds numbers produces a more uniform velocity profile with water and oil shape factors tending towards unity. However, for lower Reynolds numbers the shape factors differ considerably from unity, reaching values as high as 1.49 for water and 1.25 for oil. The oil shape factor is always lower than the water due to the greater oil viscosity. When the Reynolds numbers are constant, a decrease in the interface height resulted in significantly lower water velocities than oil. Although simple, the proposed model can be of great value for engineers and the development of numerical simulation codes.
Dispersed flows of two immiscible liquids are significant for the energy industry. The transportation and separation of multiphase mixtures is essential in the exploitation of offshore and unconventional oil sources, where water produced with oil. Dispersions provide large interfacial areas which enhance efficiencies in processes such as biofuel production and spent nuclear fuel reprocessing. Despite their importance, it is still difficult to accurately predict their evolution due to the lack of data and its inherent complexity. This work aims to use a combination of experiments and physics-based modelling to predict the evolution of oil-water dispersions in pipes, in the framework of Process Intensification. At low velocities, gravity-induced settling, and coalescence lead to the separation of dispersions. Gradually four layers are formed: the continuous phase, followed by a dilute settling/floatation layer (FL), where the drops settle towards their homophase, a densepacked layer (DPL) where drops coalesce with each other and with the interface, and an evolving continuous layer of the dispersed phase. When the drop-interface coalescence rate exceeds the floatation rate, DPL depletes, and the FL is in contact with the new continuous layer. This case has not been studied before.
Dispersed flows of silicon oil (828 kg m-3 density and 5.5 10-3 Pa s viscosity) and water were studied in a 26mm ID pipe. A novel multi-hole injector was designed to investigate the flow evolution after DPL depletion. Flow evolution is investigated with high-speed imaging and ultrasound techniques, to determine oil fraction, velocity profile, and droplet size distribution. The influence of flow hydrodynamics on drops floatation is carefully studied. A mechanistic model has been developed to predict the dispersed flow evolution. A data-driven modelling is also developing to include the influence of flow conditions on drops floatation. This is a first attempt to evaluate different settling modelling approaches.
Poster Session
Methods, Procedures, Process: The traditional method for determining pressure and temperature profiles in a pipeline involves applying mass, momentum, and energy balance equations to a control volume of the pipe and using a marching algorithm. However, the proposed method uses a matrix formulation with a vectorized procedure to determine pressure and temperature simultaneously, taking into account the connection between enthalpy and the balance equations. The proposed model is applied with both Black oil table correlations and a fully compositional model and can be used for pipelines with any inclination angle, from horizontal to vertical.
Results, Observations, Conclusions: The proposed method results in a faster and more accurate calculation compared to the traditional approach. It organizes the problem in a structured manner and provides fast and precise calculation of pressure-temperature profiles. Additionally, the vector and matrix approach improves the performance of the code and opens the possibility of coupling with reservoir simulators in an integrated approach. The results from the model are compared with commercial software.
Novel/Additive Information: The proper calculation of pressure-temperature profiles is crucial for solving flow assurance problems that are related to pressure and temperature dependence, such as paraffin and hydrate formation. This method offers a fast and precise calculation of pressure-temperature profiles and can be applied in pipelines with any inclination angle.
Gas-liquid concurrent flow in pipes is a common occurrence in the oil and gas industry. Accurately predicting the two-phase pressure gradient and holdup is critical for well and pipeline design and operations. There are two main approaches for computing the flow parameters: mechanistic and data-driven (machine learning) models. Mechanistic models require simplifications and have low flexibility to adapt to new data. In contrast, machine learning models depend on the training dataset and may have poor performance in predicting beyond the range of the training data. One limitation of both approaches is that they often produce deterministic results without indicating the uncertainty of the prediction. The lack of prediction uncertainty can lead to poor decision-making and negatively affect production optimization and reservoir management. In this work, we propose a novel framework for machine learning uncertainty quantification applied to multiphase flow; the model is based on neural networks and is used to predict not only the deterministic value of pressure gradient and holdup but the probability distribution of those parameters according to the flow conditions for steady-state pipe flow. A dataset
created using a multiphase flow commercial simulator was used for the model’s supervised learning training and validation. The trained model was then tested against experimental data collected in the lab, and the results were evaluated qualitatively (graphically) and quantitatively (coefficient of determination and negative log-likelihood).
Results show that when the model has good accuracy, the uncertainty bar is also low. However, the uncertainty bar is high when the model’s accuracy is low. This means the machine learning framework can inform the user whether its results should be trusted or not.
