Systems, apparatuses, methods, and computer storage media for measuring various phases of a multiphase fluid flow within a pipeline

Abstract

A multi-phase flow metering system has a hardware module for receiving a multi-phase flow comprising a plurality of phases, a measurement module having one or more sensors coupled to the hardware module for measuring data regarding characteristics of the multi-phase flow, a data acquisition (DAQ) module functionally for gathering the measured data from the measurement module, and a processing unit for using an artificial intelligence (AI) model for estimating at least flow rates of the plurality of phases of the multi-phase flow based on the measured data gathered by the DAQ module.

Description

CROSS-REFERENCE TO RELATED APPLICATIONSThis application claims the benefit of US Provisional Patent Application Serial No. 63 / 616,135, filed December 29, 2023, the content of which is incorporated herein by reference in its entirety . FIELD OF THE DISCLOSUREThe present disclosure relates generally to systems, apparatuses, methods, and computer storage media for measuring a multiphase fluid flow within a pipeline, and in particular to systems, apparatuses, methods, and computer storage media for measuring various phases of a multiphase fluid flow within a pipeline using artificial intelligence. BACKGROUNDMulti-phase flow metering, an essential aspect of fluid dynamics, involves measuring the flow rates of distinct phases within a complex mixture, such as gasses, liquids, and solids. This technology is instrumental for a multitude of industrial applications, providing critical insights into the distribution and dynamics of individual phases within the flow. Accurate quantification of phase fractions is pivotal for optimizing production processes, ensuring safety, and conserving resources in diverse industries, including oil and gas, chemical, environmental, and biomedical sectors.In the oil and gas industry, precise measurement of individual phase flow rates is imperative for maximizing resource recovery, optimizing reservoir performance, and enhancing operational efficiency. Moreover, multi-phase flow metering contributes to safety assurance by identifying potential hazards, thereby facilitating timely intervention to mitigate risks. It also aids in compliance with environmental regulations by assessing and monitoring the release of substances into the environment during various industrial activities.Additionally, multi-phase flow metering serves as a foundational tool for research and development, enabling the validation and refinement of models and simulations related to multi-phase flow behavior. These models play a crucial role in predicting system performance accurately. Furthermore, multi-phase flow metering finds applications in biomedical and pharmaceutical fields, providing essential data for diagnostics, treatments, and drug delivery systems.In summary, accurate multi-phase flow metering is a fundamental technology that significantly impacts diverse industries, allowing for precise measurement and analysis of multi-phase flows, thereby driving advancements, efficiency, and safety across various domains. SUMMARYAccording to one aspect of this disclosure, there is provided a multi-phase flow metering system or apparatus comprising: a hardware module for receiving a multi-phase flow comprising a plurality of phases and in some cases introducing intervention to the flow; a measurement module comprising one or more sensors coupled to the hardware module for measuring data regarding characteristics of the multi-phase flow; a data acquisition (DAQ) module functionally for gathering the measured data from the measurement module; a processing unit for using an artificial intelligence (AI) model for estimating at least flow rates of the plurality of phases of the multi-phase flow based on the measured data gathered by the DAQ module.In some embodiments, the one or more sensors are configured for measuring one or more of pressure data, temperature data, vibration data, conductivity, viscosity, density, electromagnetic, capacitance, optic, and acoustic data of the multi-phase flow.In some embodiments, the one or more sensors are further configured for measuring flow-opening size data related to the hardware module.In some embodiments, the processing unit is configured for: receiving measured data of a multi-phase flow, the multi-phase flow comprising a plurality of phases; pre-processing the measured data; filtering the pre-processed data; using the AI model for estimating at least the flow rates of the plurality of phases of the multi-phase flow based on the filtered data; and further processing and / or analyzing the estimated flow rates of the plurality of phases of the multi-phase flow.According to one aspect of this disclosure, there is provided a computerized multi-phase flow metering method comprising: receiving measured data of a multi-phase flow, the multi-phase flow comprising a plurality of phases; pre-processing the measured data; filtering the pre-processed data; using the AI model for estimating at least the flow rates of the plurality of phases of the multi-phase flow based on the filtered data; and further processing and / or analyzing the estimated flow rates of the plurality of phases of the multi-phase flow.According to one aspect of this disclosure, there is provided a multi-phase flow metering method comprising: receiving measurement data of a multi-phase flow, the multi-phase flow comprising a plurality of phases; estimating at least flow rates of the plurality of phases of the multi-phase flow using an artificial intelligence (AI) model based on the measurement data; and presenting the estimated flow rates as results of the multi-phase flow metering.In some embodiments, the multi-phase flow metering method further comprises: introducing passive or active intervention to the multi-phase flow.In some embodiments, the measurement data comprises pipe diameter, flow-opening size, pressure data, temperature data, vibration data, conductivity data, viscosity data, density data, electromagnetic data, capacitance data, optic data, acoustic data, or a combination thereof, of the multi-phase flow.In some embodiments, the multi-phase flow metering method further comprises: measuring pressure data of the multi-phase flow, measuring temperature data of the multi-phase flow, measuring vibration data of the multi-phase flow, measuring conductivity data of the multi-phase flow, measuring viscosity data of the multi-phase flow, measuring density data of the multi-phase flow, measuring electromagnetic data of the multi-phase flow, measuring capacitance data of the multi-phase flow, measuring optic data of the multi-phase flow, measuring acoustic data of the multi-phase flow, or a combination thereof.In some embodiments, the pressure data comprises inlet pressure, outlet pressure, pressure at one or more locations of the multi-phase flow intermediate inlet and outlet, pressure drop, or a combination thereof, of the the multi-phase flow; and wherein the temperature data comprises inlet temperature, outlet temperature, temperature at one or more locations of the multi-phase flow intermediate inlet and outlet, temperature difference between inlet and outlet, or a combination thereof.In some embodiments, the plurality of phases comprise gas, oil, water, solvent, solids, or In some embodiments, the AI model is a machine learning model.In some embodiments, the machine learning model comprises: a supervised learning model, an unsupervised learning model, a semi-supervised learning model, a reinforcement learning model, a few-shot learning model, an ensemble learning model, or a combination thereof.In some embodiments, the supervised learning model comprises: logistic regression, backpropagation neural networks, random forests, decision trees, XGBoost, or a combination thereof.In some embodiments, the unsupervised learning model comprises: a k-means clustering model, a hierarchical clustering model, or a combination thereof.In some embodiments, the semi-supervised learning model comprises: a Q-learning algorithm, a temporal difference algorithm, or a combination thereof.In some embodiments, the machine learning model comprises: one or more algorithms including: a regression algorithm, an instance-based method, a regularization method, a dimension reduction method, a decision tree method, a Bayesian method, a kernel method, a clustering method, an associated learning rule algorithm, an artificial neural network model, a deep learning algorithm, an ensemble method, a time series analysis method, or a combination thereof.In some embodiments, the AI model comprises a rectified linear unit (ReLU) activation function.In some embodiments, the multi-phase flow metering method further comprises: training the AI model using the measurement data.In some embodiments, said estimating at least the flow rates of the plurality of phases comprises: estimating gas flow rate, oil flow rate, water flow rate, liquid-to-gas ratio (LGR), water cut, solvent flow rate, solid content, or a combination thereof, of the multi-phase flow using the AI model based on the pre-processed measurement data.In some embodiments, said estimating at least the flow rates of the plurality of phases comprises: pre-processing the measurement data using normalization, scaling, standardization, smoothing, resampling, interpolation, feature extraction, missing data imputation, outlier detection and handling, data transformation, fast Fourier transform (FFT), principal component analysis (PCA), feature scaling, time-series decomposition, data reduction, calculating statistical moments, calculating correlations, calculating cross-correlations, or a combination thereof; and estimating at least the flow rates of the plurality of phases of the multi-phase flow using the AI model based on the pre-processed measurement data.In some embodiments, said estimating at least the flow rates of the plurality of phases comprises: filtering the measurement data using: low-pass filtering, high-pass filtering, band-pass filtering, notch filtering, Savitzky-Golay filtering, Kalman filtering, median filtering, Butterworth filter, wavelet filtering, Chebyshev filter, finite impulse response (FIR) filter, infinite impulse response (IIR) filter, Wiener filter, particle filter, extended Kalman filter (EKF), morphological filtering, adaptive filtering, anisotropic filtering, phase unwrapping, Kalman-Bucy filter, particle swarm filtering, extended information filter (EIF), or a combination thereof; and estimating at least the flow rates of the plurality of phases of the multi-phase flow using the AI model based on the filtered measurement data.In some embodiments, said estimating at least the flow rates of the plurality of phases comprises: processing the measurement data using data cleaning, data transformation, data reduction, data discretization, data normalization, data encoding, missing data handling, outlier detection and handling, feature engineering, feature scaling, data balancing, data augmentation, principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), data sampling, data binning, data reshaping, data aggregation, feature correlation analysis, anomaly detection, or a combination thereof; and estimating at least the flow rates of the plurality of phases of the multi-phase flow using the AI model based on the filtered measurement data.In some embodiments, said data encoding comprises: one-hot encoding.In some embodiments, said missing data handling comprises: data imputation.In some embodiments, said data sampling comprises: stratified sampling.In some embodiments, said data reshaping comprises: data reshaping using one or more pivot tables.In some embodiments, said processing the measurement data using comprises: processing the measurement data using anomaly detection, wherein values in the measurement data that deviate three times of standard deviation from a mean of the measurement data are identified as anormaly values.In some embodiments, the multi-phase flow metering method further comprises: processing the estimated flow rates using: data normalization, feature scaling, outlier detection, dimensionality reduction, data imputation, data transformation, ensemble techniques, model calibration, hyperparameter tuning, model evaluation metrics, interpretability analysis, visualization, reporting, documentation, deployment optimization, feedback loops, or a combination thereof.According to one aspect of this disclosure, there is provided a multi-phase flow metering system or apparatus for performing any of the above-described methods, the multi-phase flow metering system or apparatus comprising: a hardware module for receiving a multi-phase flow comprising a plurality of phases and in some cases introducing intervention; a measurement module comprising one or more sensors coupled to the hardware module for measuring data regarding characteristics of the multi-phase flow; a data acquisition (DAQ) module for gathering the measurement data from the measurement module; one or more processors; and one or more non-transitory computer-readable storage media comprising computer-executable instructions; wherein the instructions, when executed, cause the one or more processors to perform any of the above-described methods.According to one aspect of this disclosure, there is provided one or more non-transitory computer-readable storage media comprising computer-executable instructions, wherein the instructions, when executed, cause one or more processors to perform any of the above-described methods.Thus, the multi-phase flow metering system, apparatus, method, and one or more non-transitory computer-readable storage media disclosed herein provide a novel approach to multi-phase flow measurement by incorporating AI techniques, thereby revolutionizing the accuracy, efficiency, and adaptability of multi-phase flow metering across a spectrum of applications.By using the multi-phase flow metering system, apparatus, method, and one or more non-transitory computer-readable storage media disclosed herein, the flow rate of each flow component or phase may be accurately estimated without the need of separating the flow components before estimation. Thus, the multi-phase flow metering system, apparatus, and method disclosed herein provide flow-rate estimation solution with easy and cost-efficient access to data-driven flow metering, and significantly reduced training and operation burden for operators. BRIEF DESCRIPTION OF THE DRAWINGSFor a more complete understanding of the disclosure, reference is made to the following description and accompanying drawings, in which:FIG. 1 is a schematic diagram showing a multi-phase flow metering system, according to some embodiments of the present disclosure;FIG. 2A is a perspective view of a flow meter of the multi-phase flow metering system shown in FIG. 1;FIG. 2B is a side view of the flow meter shown in FIG. 2A;FIG. 2C is a cross-sectional view of the flow meter shown in FIG. 2A;FIG. 3A is a schematic front view of the flow meter shown in FIG. 2A;FIG. 3B is a schematic plan view of the flow meter shown in FIG. 2A;FIG. 3C is a schematic side view of the flow meter shown in FIG. 2A;FIG. 4 is a flowchart showing the steps of a procedure performed by a processing unit of the multi-phase flow metering system shown in FIG. 1 for estimating flow rates of various phases of a multi-phase flow, according to some embodiments of this disclosure;FIG. 5 is a plot showing an example of the standard deviation for anomaly detection performed by the processing unit of the multi-phase flow metering system shown in FIG. 1 for estimating flow rates of various phases of a multi-phase flow, according to some embodiments of this disclosure;FIG. 6 is a schematic diagram showing an artificial intelligence (AI) model in the form of a deep neural network (DNN) model;FIG. 7 is a plot showing the learning curve of the AI model shown in FIG. 6;FIG. 8 is a plot showing a comparison between the observed / actual data of the testing datasets and the predicted data obtained by the AI model shown in FIG. 6 using the testing datasets; andFIG. 9 is a plot showing the distribution of the residuals of the trained AI model shown in FIG. 6. DETAILED DESCRIPTIONThe present disclosure pertains to the field of multi-phase flow measurement, and more specifically, to a multi-phase flow metering system, apparatus, and method utilizing artificial intelligence (AI) (such as machine learning (ML)) techniques for accurately quantifying and analyzing multi-phase flows within diverse industrial, environmental, and scientific applications.Herein, a multi-phase flow is a fluid flow having a plurality of components or substances in various phases including gasses or vapours, liquids, or combinations thereof. Examples of multiphase fluid flow include combinations of oil, water, and gas; combinations of oil, water, gas, and solvent (such as solvent used in solvent-based oil recovery methods); combinations of oil, gas, and solvent, and / or the like.The multi-phase flow metering system, apparatus, and method disclosed herein integrate suitable AI methods to interpret complex data obtained from sensors and other measuring devices. By leveraging advanced computational models and pattern-recognition capabilities, the multi-phase flow metering system, apparatus, and method disclosed herein process signals representing different phases, enabling real-time determination of phase fractions, flow rates, and other relevant parameters. The integration of AI facilitates adaptive learning, enhancing the accuracy and reliability of multi-phase flow measurements across varying operating conditions.Herein, the term “flow rate” refers to the physical quantities of a substance (such as a flow component) passing a surface per unit of time. Depending on the substance and the use scenario, a flow rate may be a mass flow rate (that is, the mass of a substance passing a surface per unit of time) or a volumetric flow rate (that is, the volume of a substance passing a surface per unit of time).The multi-phase flow metering system, apparatus, and method disclosed herein is suitable for addressing challenges in traditional flow metering approaches, offering a dynamic and adaptable solution that caters to the intricacies of multi-phase flows. By harnessing the power of AI, the multi-phase flow metering system, apparatus, and method disclosed herein provide a versatile tool for optimizing industrial processes, resource allocation, safety monitoring, environmental impact assessment, and / or the like.Thus, the multi-phase flow metering system, apparatus, and method disclosed herein provide a novel approach to multi-phase flow measurement by incorporating AI techniques, thereby revolutionizing the accuracy, efficiency, and adaptability of multi-phase flow metering across a spectrum of applications.By using the multi-phase flow metering system, apparatus, and method disclosed herein, the flow rate of each flow component or phase may be accurately estimated without the need of separating the flow components before estimation. Thus, the multi-phase flow metering system, apparatus, and method disclosed herein provide flow-rate estimation solution with easy and cost-efficient access to data-driven flow metering, and significantly reduced training and operation burden for operators.Turning now to FIG. 1, a multi-phase flow metering system according to some embodiments of this disclosure is shown and is generally identified using reference numeral 100. The multi-phase flow metering system 100 comprises a hardware module 102 for receiving a multi-phase flow and in some cases introducing intervention. The hardware module 102 is coupled to or integrated with a measurement module 104 having one or more sensors, wherein the one or more sensors are instrumental in collecting various data regarding the flow characteristics. A data acquisition (DAQ) module 106 is employed to gather and organize data acquired from the measurement module 104. A processing unit 108, which in these embodiments incorporates a suitable AI method, is utilized to analyze the data gathered by the DAQ module 106 and estimate flow rates for various phases within the multi-phase flow. The obtained analysis results are accessible and comprehensible through a dedicated user interface 110, facilitating informed decision-making and operational optimization.Various modules 104 to 110 of the multi-phase flow metering system 100 may store raw and processed data in a versatile data storage unit 112 (which may be located on-site, off-site, or in multiple locations), and / or read data therefrom.Furthermore, for seamless operation, various modules 104 to 110 of the multi-phase flow metering system 100 may be powered by a power source module 114 having one or more electrical power supplies. In some embodiments, the hardware module 102 may also comprise one or more components powered by the power source module 114.In some embodiments, the multi-phase flow metering system 100 may be connected to a network 116 for communication with external sources and facilitating data transmission.In some embodiments, the hardware module 102 comprises a flow meter for integration onto a pipeline structure to receive a multi-phase flow. The dimensions of the flow meter's inlet and outlets are customized to conform to the specific dimensions of the associated pipeline, ensuring seamless alignment and efficient flow measurement within the system. The hardware is meticulously configured to exert deliberate influence on the multi-phase flow, instigating precise alterations aimed at enhancing the accuracy and efficacy of flow measurements.In diverse applications, these alterations may be categorized into either passive or active strategies. Passive alterations involve modifications to the flow section, such as contraction, expansion, corrugated walls, or vortex generators. These passive alterations inherently affect the flow dynamics, aiding in the accurate measurement of multi-phase flow. Conversely, active strategies encompass purposeful intervention, including acoustic pulsations, mechanical pulsations, and electrical or magnetic pulsations. These active strategies are strategically implemented to augment flow characteristics and optimize the precision of flow rate determinations within the multi-phase flow.FIGs. 2A to 3C show an example of a flow meter 140, according to some embodiments of this disclosure. In this example, the flow meter 140 comprises a hollow body 142 having an inlet 144A and an outlet 144B. A control compartment 146 is coupled to the hollow body 142 wherein the control compartment 146 comprises a flow intervention component 150, such as a rotatable disc controlled by a servo motor 148, in the hollow body 142.The measurement module 104 may comprise a wide array of sensors coupled to or otherwise integrated with the hardware module 102. The measurement module 104 may further comprise a communication component (not shown) received in the control compartment 146 for sending the data collected by the sensors to the DAQ module 106. The communication component may use any suitable sensor-data transmission techniques, which can be broadly categorized as wired transmission and wireless transmission. Wired transmission involves physical connections, ensuring reliability and simplicity. Serial communication, fieldbus, Modbus, Optical Fiber, Parallel communication, Industrial Communication, Ethernet and TCP / IP and Custom Communication Protocols are examples of wired communication. Wireless transmission encompasses radio frequency (RF), infrared (IR), NFC (Near Field Communication), cellular, satellite, power line communication (PLC), acoustic communication, optical transmission, ultrasonic communication, inductive coupling, Ethernet communication, mesh networking, and hybrid transmission methods.For example, as shown in FIGs. 3A to 3C, one or more sensors (not shown) may be coupled to the flow meter 140, including one or more pressure sensors and one or more temperature sensors (not shown) in the hollow body 142 of the flow meter 140 at suitable locations 152 such as in the upstream side (for example, about, near, or otherwise in proximity with the inlet 144A) thereof, in the downstream side (for example, about, near, or otherwise in proximity with the outlet 144B) thereof, in the middle section thereof, and / or the like.The one or more sensors collect the intervention status of the flow meter 140, the inlet temperature (that is, the temperature measured at the upstream side), the outlet temperature (that is, the temperature measured at the downstream side), the temperature at locations intermediate the inlet and outlet, the inlet pressure (that is, the pressure measured at the upstream side), the outlet pressure (that is, the pressure measured at the downstream side), the pressure at locations intermediate the inlet and outlet, the pressure drop (that is, the difference between the pressures measured at the upstream and downstream sides), and / or the like, and reports the collected data to the DAQ module 106 via the communication component.In various embodiments, the sensors may be pressure sensors, temperature sensors, vibration sensors, acoustic sensors, conductivity sensors, viscosity sensors, density sensors, electromagnetic sensors, capacitance sensors, optic sensors, and / or the like, which may generate a digital or analog output as required. In some embodiments, the measurement module 104 may also comprise a control circuit or a rotation sensor (not shown) coupled to the servo motor 148 to sense the flow-opening size in case of active intervention in terms of flow expansion and intervention.Any suitable types of pressure sensors may be used, such as absolute pressure sensors, gauge pressure sensors, differential pressure sensors, and / or the like. In some embodiments, piezoresistive pressure sensors, piezoelectric pressure sensors, capacitive pressure sensors, resonant pressure sensors, or microelectromechanical systems (MEMS) pressure sensors may be used.Any suitable types of temperature sensors may be used, such as thermocouples, resistance temperature detectors (RTDs), thermistors, infrared (IR) temperature sensors, semiconductor temperature sensors, bimetallic temperature sensors, gas thermometers, liquid-filled temperature sensors, fiber optic temperature sensors, solid-state temperature sensors, surface temperature sensors, MEMS temperature sensors, and / or the like.Any suitable types of vibration sensors may be used, such as piezoelectric accelerometers, MEMS accelerometers, velocity sensors, strain gauge-based vibration sensors, eddy current vibration sensors, capacitive accelerometers, and / or the like.Any suitable types of acoustic sensors may be used, such as piezoelectric transducers, capacitive transducers, electromagnetic acoustic transducers (EMATs), fiber optic acoustic sensors, microphone-based acoustic sensors, laser doppler vibrometers (LDVs), surface acoustic wave (SAW) sensors, MEMS acoustic sensors, and / or the like.Any suitable types of conductivity sensors may be used, such as contact conductivity sensors, inductive conductivity sensors, toroidal conductivity sensors, and / or the like. In some embodiments, two-electrode sensors, four-electrode sensors, or microelectromechanical systems (MEMS) conductivity sensors may be used.Any suitable types of viscosity sensors may be used, such as vibrational viscosity sensors, rotational viscosity sensors, capillary viscosity sensors, and / or the like. In some embodiments, resonant viscosity sensors, falling ball viscosity sensors, or microelectromechanical systems (MEMS) viscosity sensors may be used.Any suitable types of density sensors may be used, such as vibrating element density sensors, ultrasonic density sensors, Coriolis density sensors, and / or the like. In some embodiments, oscillating tube density sensors, radioactive density sensors, or microelectromechanical systems (MEMS) density sensors may be used.Any suitable types of electromagnetic sensors may be used, such as eddy current sensors, magnetic flux sensors, Hall effect sensors, and / or the like. In some embodiments, magnetoresistive sensors, Faraday effect sensors, or microelectromechanical systems (MEMS) electromagnetic sensors may be used.Any suitable types of capacitance sensors may be used, such as parallel plate capacitance sensors, differential capacitance sensors, interdigitated capacitance sensors, and / or the like. In some embodiments, cylindrical capacitance sensors, fluid level capacitance sensors, or microelectromechanical systems (MEMS) capacitance sensors may be used.Any suitable types of optic sensors may be used, such as fiber optic sensors, photodiode sensors, laser sensors, and / or the like. In some embodiments, infrared sensors, ultraviolet sensors, or microelectromechanical systems (MEMS) optic sensors may be used.Any suitable types of sensor connections may be used for connecting the sensors, such as circular connectors, DIN connectors, D-sub connectors, micro-D connectors, USB connectors, RJ45 connectors, miniature connectors, subminiature connectors, push-pull connectors, coaxial connectors, terminal blocks, screw terminal connectors, latch-locking connectors, bayonet connectors, circular push-pull connectors, magnetic connectors, waterproof connectors, wire leads, thermocouple connectors, M8 and M12 connectors, insulation displacement connector (IDC), Molex connectors, printed circuit board (PCB) connectors, ribbon cable connectors, soldered connections, and / or the like.The DAQ module 106 is a central module of the multi-phase flow metering system 100, performing various tasks for the acquisition and processing of data emanating from various sensors in use. The DAQ module 106 operates as an important interface bridging the physical domain of sensor outputs to the digital realm where comprehensive data analysis and interpretation take place. The DAQ module 106 comprises a versatile functionality for flexibly adapting to and handling diverse sensor types and their respective outputs. More specifically, the functions of the DAQ module 106 comprises signal conditioning, analog-to-digital conversion (ADC) if input signal is analog, digital input / output (I / O), data processing, data logging and storage, synchronization, customization and scalability, power management, fault tolerance, real-time monitoring, calibration, compensation, and / or the like. The DAQ module 106 is adept at supporting multiple communication protocols, including, but not limited to, USB, Ethernet, RS-232, and RS-485.The processing unit 108 receives data transmitted from the DAQ module 106 and executes instructions (such as computer-executable instructions stored in a memory as software and / or firmware) to determine the flow rate of individual phases of the multi-phase flow. The processing unit 108 may comprise one or more of any suitable processing units or processors such as one or more of: central processing units (CPUs), graphics processing units (GPUs), high-performance computing (HPC) clusters, tensor processing units (TPUs), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), quantum processors, digital signal processors (DSPs), microcontrollers coprocessors, parallel processing clusters, cloud-based data processing, distributed computing, data processing units (DPUs), application servers, neuromorphic processors, analog computers, edge computing devices, in-memory computing, server farms, parallel computing clusters, and / or the like.With the computer-executable instructions, the processing unit 108 in some embodiments provides four types of functions including: a processor engine, a set of AI software tools, a set of management tools, and a set of collaboration tools. The AI software tools may comprise various suitable components such as AI frameworks and libraries, development environments, version control, containerization, orchestration, and AI development tools, along with AI model deployment solutions.In some embodiments, the processing unit may utilize various hardware for implementation, including central processing units (CPUs), graphics processing units (GPUs), tensor processing units (TPUs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and / or other specialized hardware designed for AI computation and data processing.The management tools involve strategic choices such as utilizing cloud services or on-premises servers, deploying DevOps tools, integrating AI models and algorithms, and implementing robust monitoring and logging systems, as well as security and privacy measures.The collaboration tools facilitate seamless teamwork, while documentation and knowledge-sharing tools are invaluable. Furthermore, the collaboration tools may comprise compliance and regulation tools for ensuring adherence to established standards and guidelines, culminating in a comprehensive, and resilient infrastructure framework.The interface module 110 provides necessary information, such as the estimated flow rate of various phases in the multi-phase flow, to one or more users, and receives input and instructions from the one or more users. In various embodiments, an assortment of interfaces may be utilized, such as standalone desktop application, web-based interface, mobile app, cross-platform mobile frameworks, progressive web app (PWA), Internet of things (IoT) interface, augmented reality (AR) or virtual reality (VR) interface, voice user interface (VUI), chatbot or conversational interface, command-line interface (CLI), local touchscreen interface, Simple Display Screen interface, Physical Buttons and Knobs, LED Indicators, Keypad / Keyboard, Smart TV Interface, Industrial Human-Machine Interface (HMI), Programmable Logic Controller (PLC) Interface, API Integration, Text Message or SMS Interface, Email Interface, Gesture-Based Interface, Eye-Tracking Interface, Brain-Computer Interface (BCI), Command Line Interface (CLI), Printed Reports, Near-Field Communication (NFC), Bluetooth Interface, and / or the like.The power source module 114 provides power to various modules of the system 100. In various embodiments, a variety of energy sources may be used, such as alternate current (AC) power, DC power, solar power, wind power, battery systems, generators (using diesel, natural gas, propane, and / or the like), fuel cells, thermoelectric generators, kinetic energy harvesting, and / or the like.The data storage module 112 comprises one or more computer-readable storage media for various modules such as the measurement module 104, DAQ module 106, processing unit 108, and interface module 110 to store data thereto and to read data therefrom. The data storage module 112 may also store computer-executable instructions or code for the processing unit 108 to execute and perform flow-rate estimation.In various embodiments, various computer-readable storage media may be used, such as hard disk drives (HDD), solid-state drives (SSD), optical discs, USB flash drives, network attached storage (NAS), cloud storage services (for example, Dropbox, Google Drive, and / or the like), external hard drives, random access memory (RAM), redundant array of independent disks (RAID), storage area network (SAN), direct attached storage (DAS), magnetic disks (such as floppy disks), three-dimensional (3D) NAND flash memory, and / or the like.The network module 116 may be used for connecting to other computing devices and / or networks (such the Internet, a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), and / or the like). In various embodiments, various data-transfer methods and network technologies may be used, such as wired data transfer employing communication media such as USB cable, Ethernet cable, serial cable (RS-232 / RS-485), fiber optic cable, and / or the like. Wireless data transfer methods such as WI-FI® (WI-FI is a registered trademark of Wi-Fi Alliance, Austin, TX, USA), BLUETOOTH® (BLUETOOTH is a registered trademark of Bluetooth Sig Inc., Kirkland, WA, USA), near-field communication (NFC), infrared (IR), ZIGBEE® (ZIGBEE is a registered trademark of ZigBee Alliance Corp., San Ramon, CA, USA), 3G, 4G, 5G, 6G cellular communications, satellite communication , and / or the like. Data transfer via removable media may also be used which involves options such as USB flash drive, external hard drive, memory cards (such as SD, microSD, etc.), and / or the like. Cloud-based data transfer may also be used which encompasses cloud storage services (for example, Dropbox, Google Drive, OneDrive), file transfer services (for example, WeTransfer), email attachments, and data transfer via application programming interfaces (APIs) and web APIs such as file transfer protocol (FTP), FTP server, and SSH file transfer protocol (SFTP), peer-to-peer (P2P) data transfer, P2P file sharing, and peer-to-peer networking may also be used for data transfer and networking.FIG. 4 is a flowchart showing the steps of a procedure 200 performed by the processing unit 108, according to some embodiments of this disclosure. In these embodiments, the procedure 200 may be implemented as the computer-executable instructions stored on one or more computer-readable storage media in the form of one or more software programs, or one or more firmware programs.At step 202, the processing unit 108 pre-processes the data received from the DAQ module 106. At this step, various pre-processing techniques may be used such as normalization or scaling, standardization, smoothing, resampling and interpolation, feature extraction, missing data imputation, outlier detection and handling, data transformation, fast Fourier transform (FFT), principal component analysis (PCA), feature scaling, time-series decomposition, data reduction, statistical moments, correlations and cross-correlations, and / or the like.At step 204, the processing unit 108 filters the pre-processed data using suitable filtering techniques such as low-pass filtering, high-pass filtering, band-pass filtering, notch filtering, Savitzky-Golay filtering, Kalman filtering, median filtering, Butterworth filter, wavelet filtering, Chebyshev filter, finite impulse response (FIR) filter, infinite impulse response (IIR) filter, Wiener filter, particle filter, extended Kalman filter (EKF), morphological filtering, adaptive filtering, anisotropic filtering, phase unwrapping, Kalman-Bucy filter, particle swarm filtering, extended information filter (EIF), and / or the like.At step 206, the processing unit 108 further processes the filtered data and provide the processed or prepared data for training an AI model and / or for use by an AI engine (also called an “AI core”) to determine the flow rates of various phases of the multi-phase flow. The use of the pre-processing 202, filtering 204, and data preparation 206 gives rise to data suitable for the AI engine to use and ensures that the AI model would not merely memorize patterns present in the unprocessed data, which may otherwise cause biases in the prediction of the AI model. Additionally, the data preparation 206 allows the AI engine to extract more meaningful features and learn more effectively.In various embodiments, the processing unit 108 may perform various important data-processing tasks for data preparation 206. Examples of such data-processing tasks include data cleaning, data transformation, data reduction, data discretization, data normalization, data encoding (for example, one-hot encoding), missing data handling (for example, imputation), outlier detection and handling, feature engineering, feature scaling, data balancing (for imbalanced datasets), data augmentation, principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), data sampling (for example, stratified sampling), binning, reshaping (for example, pivot tables), data aggregation, feature correlation analysis, and / or the like.In some embodiments, the data preparation unit 206 may perform an anomaly detection process, wherein the mean value and standard deviation are calculated. In these embodiments, the values which deviate three (3) times of standard deviation from the mean are considered anomaly. FIG. 5 shows an example of the calculated standard deviation.Following the preparation process, the prepared data is segregated into two primary datasets: a training dataset and a testing dataset.Referring back to FIG. 4, at step 208, the AI core uses an AI method (such as machine learning, statistical modeling, and / or the like) and employs an AI model for in-depth analysis of the data received from the data preparation step 206. This AI model utilizes a fusion of pipeline-specific features and diverse flow characteristics to precisely estimate the flow rate of specific components (such as gas, oil, water, solvent (such as when solvent-based oil recovery methods are used), and solids) within the multi-phase flow passing through the flow meter. Such a methodology entails the development of an AI-driven solution that comprehensively considers distinct features of the pipeline, discerns underlying patterns, and accurately forecasts the flow of different phases within the flow.In various embodiments, the machine-learning methods may comprise supervised learning (for example, logistic regression, backpropagation neural networks, random forests, decision trees, XGBoost, and / or the like), unsupervised learning (such as k-means clustering, hierarchical clustering, and / or the like), semi-supervised learning (such as Q-learning algorithm, temporal difference algorithm, and / or the like), reinforcement learning, few-shot learning, ensemble learning, and / or the like. Additionally, each machine learning method may implement diverse algorithms or methodologies such as regression algorithms, instance-based methods, regularization methods, dimension reduction methods, decision tree methods, Bayesian methods, kernel methods, clustering methods, associated learning rule algorithms, artificial neural network models, deep learning algorithms, ensemble methods, time series analysis, and / or the like.In some embodiments, the processing unit 108 utilizes a sophisticated machine learning algorithm in the form of a deep neural network (DNN) to enhance flow-rate estimation.FIG. 6 is a schematic representation of a DNN 250, which comprises an input layer 252, one or more hidden layers 254, and an output layer 256. The input layer 252 comprises one or more input nodes 262 for receiving the data prepared at step 206 such as measurement data, including inlet temperature, outlet temperature, temperature difference between inlet and outlet, inlet pressure, outlet pressure, pressure drop, pipe diameter, flow-opening size, and / or the like. The output layer 256 comprises one or more output nodes 266. Each hidden layer 254 comprises a plurality of nodes 264. The nodes of adjacent layers are interconnected. Therefore, the nodes 264, spanning between adjacent layers 254, establish connections to facilitate the flow of data from nodes within the preceding layer, be it the input layer 252 or another hidden layer 254, for processing. The processed results are subsequently transmitted to nodes in the subsequent layer, which may constitute another hidden layer 254 or the output layer 256.Furthermore, each node is equipped with activation functions such as Rectified Linear Unit (ReLU), LeakyReLU, Tanh, Sigmoid, and various others. These activation functions serve to introduce non-linearity to the data, thus enabling the extraction of concealed patterns.For example, in some embodiments, the AI model uses a rectified linear unit (ReLU) activation function. In some embodiments, the AI model is a five (5) layers, fully-densely connected convolutional network with the ReLU activation function. Table 1 shows an example of the hyper-parameter settings of this AI model.Table 1: An Example of Hyper-parameter settingsHyper-parameterSettingNumber of layers5Number of neurons or nodes in each layer128, 64, 32, 16, 1Activation functionReLUOptimizationAdaptive moment estimation (Adam) optimizerLearning rate0.001Number of epochs200Batch size32Batch normalizationNo The AI model may be trained using a supervised training procedure with various input data such as measurement data, including but not limited to inlet temperature, outlet temperature, temperature difference between inlet and outlet, inlet pressure, outlet pressure, pressure drop, pipe diameter, flow-opening size, and / or the like, where the correct answers (that is, labels) including, flow rate of individual phases (such as gas flow rate, oil flow rate, and water flow rate), liquid-to-gas ratio (LGR), water cut, solvent flow rate, solid content, and / or the like, are provided during training. Once the AI model is trained, unlabeled data such as the data obtained from the sensors to estimate or otherwise predict the flow rates of various phases, the LGR, water cut, solvent flow rate, solid content, and / or the like.Those skilled in the art will appreciate that various machine learning methods may be used, such as regression algorithms, instance-based methods (e.g., k-nearest neighbor), regularization methods, dimension reduction methods (e.g., principal component analysis), decision tree methods, Bayesian methods, kernel methods, clustering methods, associated learning rule algorithms, and / or the like.In some embodiments, a thorough hyperparameter tuning process is executed to optimize the AI model’s hyperparameters, adapting to the dynamic context within each specific target scenario. Examples of such hyperparameters include variables such as learning rate, epochs, batch size, activation function, number of hidden layers, number of neurons per layer, dropout rate, weight initialization, optimizer, loss function, kernel size, pooling size, strides, padding, learning rate decay, momentum, regularization strength, activation threshold, early stopping, mini-batch strategy, and / or the like.In various embodiments, the multi-phase flow metering system 100 collects sensor data, which can manifest in diverse formats including numerical data, categorical data, time series data, binary data, hierarchical data, network data (such as graphs), sensor data, log data, environmental data, customer interaction data, and more sourced from one or more sensor data repositories (not shown) within the system 100.In some embodiments, the AI engine uses a deep learning model (exemplified by a dense neural network) such as a high-level feature extraction model for extracting high-level features from the sensor data. The process of feature extraction may be accomplished through the utilization of a high-level deep learning model, such as a dense neural network, or in some embodiments, employing long short term memory (LSTM) structures.Additionally, the ensemble for multi-feature extraction in machine learning comprises multiple sub-machine learning models. Each sub-model is designated to perform specific tasks related to feature detection and / or classification of features. This ensemble approach significantly augments the system’s capability to effectively manage diverse feature types.The estimated flow rates of various phases of the multi-phase flow are output from the AI engine at step 208. Then, the processing unit 108 performs post-processing and / or further analysis on the estimated flow rates to evaluate the importance and contributions of various features. In various embodiments, various post-processing techniques may be used at this step, such as data normalization, feature scaling, outlier detection, dimensionality reduction, data imputation, data transformation, ensemble techniques, model calibration, hyperparameter tuning, model evaluation metrics, interpretability analysis, visualization, reporting, documentation, deployment optimization, feedback loops, and / or the like.As described above, after post-processing, the estimated flow rates of various phases of the multi-phase flow the analysis results thereof are sent to the interface 110 for presenting to one or more users.In some embodiments, the AI model may be trained using diverse data origins, which may be obtained from historical data derived from oil and gas field operations. Alternatively or additionally, the AI model may leverage data generated through simulation software. Alternatively or additionally, data acquired during lab testing may also be used for training the AI model.A training dataset is utilized to train the machine learning model, while the evaluation of model performance depends on the validation set. This synchronization allows for precise performance evaluation under practical conditions. Various machine learning architectures may be used, with the validation dataset playing a critical role in refining their hyperparameters. A comprehensive range of hyperparameter configurations is methodically evaluated, covering aspects such as layer count, neuron quantity per layer, activation functions, epochs, batch sizes, optimizers, normalizations, and other pertinent parameters. This exhaustive investigation ensures the selection of an optimized machine learning architecture, finely calibrated to yield resilient and accurate results.In an example as shown in Table 2, the total number of trainable parameters in these embodiments is 11,905.Table 2: Example of Model parametersThe processing time of each training epoch is one (1) millisecond (ms) in the example. The loss value for the example training set is 0.0254, and for the example validation set is 0.0208. FIG. 7 shows the learning curve.FIG. 8 is a plot showing a comparison between the observed / actual data of the example testing datasets and the predicted data obtained by the AI model using the testing datasets.FIG. 9 is a plot showing the distribution of the residuals of the example trained AI model (that is, the differences between observed / actual values and predicted values).Thus, the systems, apparatuses, methods, and computer storage media disclosed herein provide accurate and real-time flow-rate estimation of various flow components of a multiphase flow such as an oil and / or gas production flow without prior separation of the flow components.While in above examples, estimation of the flow rates of water, gas, and oil flow components is described, in some embodiments, the above described systems, apparatuses, methods, and computer storage media may be used for estimating lightness / heaviness of oil.In some embodiments, various modules of the multi-phase flow metering system 100 may be integrated and the multi-phase flow metering system 100 may be implemented as a multi-phase flow metering apparatus.Those skilled in the art will appreciate that such various embodiments and / or features thereof may be customized and / or combined as needed or desired. Moreover, although embodiments have been described above with reference to the accompanying drawings, those of skill in the art will appreciate that variations and modifications may be made without departing from the scope thereof as defined by the appended claims.

Claims

1. A multi-phase flow metering method comprising:introducing intervention to a multi-phase flow in a pipe, the multi-phase flow comprising a plurality of phases;measuring the intervened multi-phase flow to obtain measurement data thereof;estimating at least flow rates of the plurality of phases of the multi-phase flow using an artificial intelligence (AI) model based on the measurement data; andpresenting the estimated flow rates as results of the multi-phase flow metering;wherein said introducing the intervention comprises:introducing passive or active intervention to the multi-phase flow for optimizing said estimating the at least flow rates of the plurality of phases of the multi-phase flow.

2. The multi-phase flow metering method of claim 1, wherein said introducing the passive intervention to the multi-phase flow comprises:introducing the passive intervention to the multi-phase flow to affect flow dynamics of the multi-phase flow for optimizing said estimating the at least flow rates of the plurality of phases of the multi-phase flow; andwherein said introducing the active intervention to the multi-phase flow comprises:introducing the active intervention to the multi-phase flow to augment flow characteristics and optimize precision of flow rate determinations for optimizing said estimating the at least flow rates of the plurality of phases of the multi-phase flow.

3. The multi-phase flow metering method of claim 2, wherein the passive intervention comprises:modifying a section of the multi-phase flow for affecting the flow dynamics of the multi-phase flow for optimizing said estimating the at least flow rates of the plurality of phases of the multi-phase flow; andwherein the active intervention comprises:introducing acoustic pulsations, mechanical pulsations, electrical or magnetic pulsations, or a combination thereof, to augment the flow characteristics and optimize the precision of the flow rate determinations for optimizing said estimating the at least flow rates of the plurality of phases of the multi-phase flow.

4. The multi-phase flow metering method of claim 1 or 1, wherein the measurement data comprises pressure data of the multi-phase flow, temperature data of the multi-phase flow, vibration data of the multi-phase flow, conductivity data of the multi-phase flow, viscosity data of the multi-phase flow, density data of the multi-phase flow, electromagnetic data of the multi-phase flow, capacitance data of the multi-phase flow, optic data of the multi-phase flow, acoustic data of the multi-phase flow, or a combination thereof.

5. The multi-phase flow metering method of claim 4, wherein the pressure data comprises inlet pressure, outlet pressure, pressure at one or more locations of the multi-phase flow intermediate inlet and outlet, pressure drop, or a combination thereof, of the multi-phase flow; and wherein the temperature data comprises inlet temperature, outlet temperature, temperature at one or more locations of the multi-phase flow intermediate inlet and outlet, temperature difference between inlet and outlet, or a combination thereof.

6. The multi-phase flow metering method of any one of claims 1 to 5, wherein the AI model is a machine learning model.

7. The multi-phase flow metering method of claim 6, wherein the machine learning model comprises: a supervised learning model, an unsupervised learning model, a semi-supervised learning model, a reinforcement learning model, a few-shot learning model, an ensemble learning model, or a combination thereof.

8. The multi-phase flow metering method of claim 7, wherein the supervised learning model comprises: logistic regression, neural networks, random forests, decision trees, XGBoost, or a combination thereof.

9. The multi-phase flow metering method of claim 7 or 8, wherein the unsupervised learning model comprises: a k-means clustering model, a hierarchical clustering model, or a combination thereof.

10. The multi-phase flow metering method of claim 7 or 8, wherein the semi-supervised learning model comprises: a Q-learning algorithm, a temporal difference algorithm, or a combination thereof.

11. The multi-phase flow metering method of any one of claims 6 to 10, wherein the machine learning model comprises: one or more algorithms including: a regression algorithm, an instance-based method, a regularization method, a dimension reduction method, a decision tree method, a Bayesian method, a kernel method, a clustering method, an associated learning rule algorithm, an artificial neural network model, a deep learning algorithm, an ensemble method, a time series analysis method, or a combination thereof.

12. The multi-phase flow metering method of any one of claims 1 to 11, wherein said estimating at least the flow rates of the plurality of phases comprises:estimating gas flow rate, oil flow rate, water flow rate, liquid-to-gas ratio (LGR), water cut, solvent flow rate, solid content, or a combination thereof, of the multi-phase flow using the AI model based on the pre-processed measurement data.

13. The multi-phase flow metering method of any one of claims 1 to 12, wherein said estimating at least the flow rates of the plurality of phases comprises:pre-processing the measurement data using normalization, scaling, standardization, smoothing, resampling, interpolation, feature extraction, missing data imputation, outlier detection and handling, data transformation, fast Fourier transform (FFT), principal component analysis (PCA), feature scaling, time-series decomposition, data reduction, calculating statistical moments, calculating correlations, calculating cross-correlations, or a combination thereof; andestimating at least the flow rates of the plurality of phases of the multi-phase flow using the AI model based on the pre-processed measurement data.

14. The multi-phase flow metering method of any one of claims 1 to 13, wherein said estimating at least the flow rates of the plurality of phases comprises:filtering the measurement data using: low-pass filtering, high-pass filtering, band-pass filtering, notch filtering, Savitzky-Golay filtering, Kalman filtering, median filtering, Butterworth filter, wavelet filtering, Chebyshev filter, finite impulse response (FIR) filter, infinite impulse response (IIR) filter, Wiener filter, particle filter, extended Kalman filter (EKF), morphological filtering, adaptive filtering, anisotropic filtering, phase unwrapping, Kalman-Bucy filter, particle swarm filtering, extended information filter (EIF), or a combination thereof; andestimating at least the flow rates of the plurality of phases of the multi-phase flow using the AI model based on the filtered measurement data.

15. The multi-phase flow metering method of any one of claims 1 to 14, wherein said estimating at least the flow rates of the plurality of phases comprises:processing the measurement data using data cleaning, data transformation, data reduction, data discretization, data normalization, data encoding, missing data handling, outlier detection and handling, feature engineering, feature scaling, data balancing, data augmentation, principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), data sampling, data binning, data reshaping, data aggregation, feature correlation analysis, anomaly detection, or a combination thereof; andestimating at least the flow rates of the plurality of phases of the multi-phase flow using the AI model based on the filtered measurement data.

16. The multi-phase flow metering method of claim 15, wherein said data encoding comprises: one-hot encoding.

17. The multi-phase flow metering method of claim 15 or 16, wherein said data sampling comprises: stratified sampling.

18. The multi-phase flow metering method of any one of claims 15 to 17, wherein said processing the measurement data using comprises:processing the measurement data using anomaly detection, wherein values in the measurement data that deviate three times of standard deviation from a mean of the measurement data are identified as anomaly values.

19. A multi-phase flow metering system or apparatus comprising:a hardware module for receiving a multi-phase flow comprising a plurality of phases;a measurement module comprising one or more sensors coupled to the hardware module for measuring data regarding characteristics of the multi-phase flow;a data acquisition (DAQ) module for gathering the measurement data from the measurement module;one or more processors; andone or more non-transitory computer-readable storage media comprising computer-executable instructions;wherein the instructions, when executed, cause the one or more processors to perform the method of any one of claims 1 to 18.

20. One or more non-transitory computer-readable storage media comprising computer-executable instructions, wherein the instructions, when executed, cause one or more processors to perform the method of any one of claims 1 to 18.  

Images