Multi-sensor data assimilation and predictive analysis for optimizing well operations

Abstract

The examples described herein provide a computer-implemented method that includes analyzing a first dataset by applying it to a first model to generate a first result. The method further includes analyzing a second dataset by applying it to a second model to generate a second result. The method also includes performing validation on the first and second models by comparing the first and second results. The method further includes modifying the operational actions of a surface component based on at least one of the first or second results in response to determining that the first and second results match. The method also includes updating at least one of the first or second models in response to determining that the first and second results do not match.

Description

[0001] Cross-references to related applications

[0002] This application claims the benefit of U.S. Application No. 17 / 012,646, filed on September 4, 2020, the entire contents of which are incorporated herein by reference. Background Technology

[0003] The implementation scheme described in this article covers downhole exploration and mining operations in the resource extraction industry, and more specifically, it relates to technologies for multi-sensor data intervention and predictive analysis for well intervention.

[0004] Downhole exploration and production operations involve the deployment of a variety of sensors and tools. Sensors, for example, provide information about the downhole environment by collecting data on temperature, density, saturation, and resistivity, as well as many other parameters. This information can be used to control various aspects of drilling and tools or systems located in bottom-hole components, along the drill string, or on the surface. Summary of the Invention

[0005] The embodiments of the present invention relate to techniques for multi-sensor data intervention and predictive analysis for well intervention.

[0006] A non-limiting exemplary computer-implemented method includes analyzing a first dataset by applying it to a first model to generate a first result. The method also includes analyzing a second dataset by applying it to a second model to generate a second result. The method further includes performing validation on the first and second models by comparing the first and second results. The method also includes modifying the operational actions of a surface component based on at least one of the first or second results in response to determining that the first and second results match. The method also includes updating at least one of the first or second models in response to determining that the first and second results do not match.

[0007] A non-limiting exemplary system includes a surface component and a processing system. The processing system includes a memory that includes computer-readable instructions. The processing system also includes a processing device for executing the computer-readable instructions, which control the processing device to perform operations. The operations include analyzing the first dataset by applying it to a first model to generate a first result. The operations also include analyzing the second dataset by applying it to a second model to generate a second result. The operations further include performing validation on the first and second models by comparing the first result with the second result. The operations also include modifying the operational actions of the surface component based on at least one of the first or second result in response to determining that the first and second results match. The operations also include updating at least one of the first or second models in response to determining that the first and second results do not match.

[0008] Non-limiting exemplary methods for processing downhole multi-sensor data representing well operations include using a physical notification model representing the well operation, encoded into one or more computer systems. The method further includes using a fitting constant associated with at least one of the following: reservoir rock, fluid, heat, completion, and wellbore intervention characteristics. The method also includes using the physical notification model to generate a prior probability distribution function. The method further includes repeatedly fusing input data from at least a first and a second downhole sensor with the prior probability distribution function to generate a posterior probability distribution function. The method also includes processing the posterior probability distribution function using one or more computing systems to generate an output in real time.

[0009] Other embodiments of the present invention implement the features of the above-described method in computer systems and computer program products.

[0010] Additional technical features and benefits are achieved through the technology of this invention. Embodiments and aspects of the invention are described in detail herein and are considered part of the claimed subject matter. For a better understanding, refer to the detailed description and accompanying drawings. Attached Figure Description

[0011] Referring now to the accompanying drawings, in which similar elements in several drawings have similar numbers:

[0012] Figure 1 A cross-sectional view depicting wellbore operation according to one or more embodiments described herein;

[0013] Figure 2 The present invention describes one or more embodiments according to the present invention. Figure 1 A block diagram of the processing system that can be used to implement the technology herein;

[0014] Figure 3 A flowchart is depicted illustrating a method for multi-sensor data interpretation and predictive analysis for well intervention according to one or more embodiments described herein; and

[0015] Figures 4 to 12 A diagram illustrating a method for optimizing multi-sensor data assimilation and predictive analysis for well intervention operations, according to one or more embodiments described herein; and

[0016] Figure 13 A method for processing downhole multisensor data representing well operations, according to one or more embodiments described herein, is described. Detailed Implementation

[0017] A modern bottom-of-well assembly (BHA) consists of several distributed components such as sensors and tools, each performing data acquisition and / or processing for a specific purpose. Some BHAs, such as those used in wireline logging and logging-while-drilling (LWD) operations, provide fluid analysis sampling and testing to obtain formation pressure and formation fluid samples during drilling.

[0018] Drilling wells below the surface to extract hydrocarbons and for other purposes. Specifically, Figure 1 A cross-sectional view of a wellbore operating system 100 according to various aspects of this disclosure is depicted. In conventional wellbore operations, LWD measurements are performed during drilling operations to determine the formation rock and fluid properties of formation 4. Those properties are then used for various purposes, such as estimating reserves based on saturation logging, defining completion settings, etc., as described herein.

[0019] Figure 1 The system and arrangement shown are an example illustrating a downhole environment. While this system can operate in any subsurface environment, Figure 1 A support 5 is shown installed in a borehole 2 penetrating the formation 4. A tool 7 is installed in the borehole 2 located at the distal end of the support 5, as shown. Figure 1 As shown.

[0020] like Figure 1 As shown, support 5 is a downhole instrument string including BHA 13. BHA 13 is part of the drilling rig or flexible tubing workover rig 8 (also referred to as the “surface assembly”) and may include drill collars, stabilizers, reamers, motors, turbines, reels, ejectors, flexible tubing, etc., as well as drill bit or flexible tubing tools 7. In this example, downhole tool 7 is positioned at the front end of BHA 13. BHA 13 also includes sensors (e.g., measurement tool 11) and electronics (e.g., downhole electronics 9). For example, measurements collected by measurement tool 11 may include measurements relevant to downhole operations during drilling, completion, or intervention wells. Surface assembly 8 also pumps drilling, completion, or intervention well fluids through the drill string or flexible tubing. According to one or more embodiments described herein, measurement tool 11 and downhole electronics 9 are configured to perform one or more types of measurements in embodiments referred to as logging while drilling (LWD) or measurement while drilling (MWD). In the example, the measuring tool 11 and the downhole electronics 9 are configured to perform downhole telemetry measurements during well intervention operations. This may include, for example, fluid sampling operations, production profiling, acid stimulation, hydraulic fracturing, search and rescue, milling, etc.

[0021] Data is collected by the measuring tool 11 and transmitted to the downhole electronics 9 for processing. Data can be transmitted between the measuring tool 11 and the downhole electronics 9 via wires 6 (such as power lines for transmitting power and / or data between the measuring tool 11 and the downhole electronics 9), and / or via a wireless link (not shown). Power is generated downhole via a turbine generator assembly (not shown), and communication with the surface 3 (e.g., with the processing system 12) is cableless (e.g., using mud pulse telemetry, electromagnetic telemetry, acoustic telemetry, etc.) and / or cable-bound (e.g., using a cable connected to the processing system 12). The data processed by the downhole electronics 9 can then be telemetryd to the surface 3 via wires 6, for example, by a telemetry system utilizing fluid pressure variations similar to a mud pulser, or by an electromagnetic telemetry system utilizing electromagnetic waves, using telemetry techniques for further processing or display by the processing system 12.

[0022] According to embodiments of this disclosure, downhole control signals may be generated and transmitted downhole by processing system 12 (e.g., based on raw data collected by measuring tool 11), or may be generated within downhole electronics 9, or a combination of both. Downhole electronics 9 and processing system 12 may each include one or more processors and one or more memory devices. In alternative embodiments, computing resources such as downhole electronics 9, sensors, and other tools may be positioned along support 5, rather than, for example, within BHA 13. The borehole 2 may be vertical as shown, or may be in other orientations / arrangements (see, for example...). Figure 3 ).

[0023] It should be understood that the embodiments of this disclosure can be implemented in conjunction with any other suitable type of computing environment now known or developed in the future. For example, Figure 2 Depicting Figure 1 A block diagram of a processing system 12 is provided, which can be used to implement the techniques described herein. In the example, the processing system 12 has one or more central processing units 21a, 21b, 21c, etc. (collectively or collectively referred to as processors and / or processing devices). In aspects of this disclosure, each processor 21 may include a Reduced Instruction Set Computer (RISC) microprocessor. The processor 21 is coupled to system memory (e.g., random access memory (RAM) 24) and various other components via a system bus 33. Read-only memory (ROM) 22 is coupled to the system bus 33 and may include a Basic Input / Output System (BIOS) that controls certain basic functions of the processing system 12.

[0024] Also shown are input / output (I / O) adapter 27 and network adapter 26 coupled to system bus 33. I / O adapter 27 may be a Small Computer System Interface (SCSI) adapter that communicates with hard disk 23 and / or tape storage device 25 or any other similar component. I / O adapter 27, hard disk 23, and tape storage device 25 are collectively referred to herein as mass storage device 34. Operating system 40 for execution on processing system 12 may be stored in mass storage device 34. Network adapter 26 interconnects system bus 33 with external network 36, enabling processing system 12 to communicate with other such systems.

[0025] A display (e.g., a display monitor) 35 is connected to the system bus 33 via a display adapter 32, which may include a graphics adapter and a video controller for improving the performance of graphics-intensive applications. In one aspect of this disclosure, adapters 26, 27, and / or 32 may be connected to one or more I / O buses, which are connected to the system bus 33 via an intermediate bus bridge (not shown). Suitable I / O buses for connecting peripheral devices such as hard disk controllers, network adapters, and graphics adapters typically include common protocols such as Peripheral Component Interconnect (PCI). Additional input / output devices are shown connected to the system bus 33 via a user interface adapter 28 and a display adapter 32. A keyboard 29, a mouse 30, and a speaker 31 may be interconnected to the system bus 33 via a user interface adapter 28, which may include, for example, a super I / O chip integrating multiple device adapters into a single integrated circuit.

[0026] In some aspects of this disclosure, the processing system 12 includes a graphics processing unit 37. The graphics processing unit 37 is specialized electronic circuitry designed to manipulate and modify memory to accelerate the creation of images in a buffer intended for output to a display. Generally, the graphics processing unit 37 is highly efficient in manipulating computer graphics and image processing, and has a highly parallel architecture, making it more efficient than a general-purpose CPU for algorithms that perform parallel processing of large data blocks.

[0027] Therefore, as configured herein, the processing system 12 includes processing power in the form of a processor 21, storage capacity including system memory (e.g., RAM 24) and mass storage device 34, input devices such as a keyboard 29 and a mouse 30, and output capacity including a speaker 31 and a display 35. In some aspects of this disclosure, a portion of the system memory (e.g., RAM 24) and mass storage device 34 jointly stores an operating system to coordinate the functions of the various components shown in the processing system 12. The system memory (e.g., RAM 24) may also store computer-readable instructions for performing the various operations described herein.

[0028] Based on the examples described herein, techniques for multi-sensor data assimilation and predictive analytics for optimizing well operations are provided. Available downhole sensors (both single and distributed sensors) generate a vast amount of data about the downhole environment during well operations. The knowledge of real-time (or near real-time) estimations and predictions in well operations provides useful input for monitoring and controlling well operations such as physical petroleum engineering systems within broader enterprise operations, such as artificial hoisting systems, smart completion systems, chemical injection systems, well enhancement systems (e.g., matrix acidizing and hydraulic fracturing), enhanced oil recovery processes (e.g., water, polymer, and chemical flooding), rock physics estimation, geological carbon storage operations, well interventions, etc. This type of downhole data can be analyzed and validated using this technique to provide near-instantaneous decision-making in well operations.

[0029] Specifically, according to one or more embodiments described herein, downhole temperature sensors (DTS) and downhole acoustic sensors (DAS) generate data. Each dataset (from DTS and DAS) can be used individually to transform the sensed data into an output stream (i.e., a flow distribution along the well) in real-time or near real-time using various models. The output streams from the DTS and DAS data can be compared to validate the model. If the output streams are consistent (i.e., match), a confidence level in the accuracy of the results can be assumed. However, if the output streams are inconsistent, data science and data analytics can be used to iteratively adjust the input parameters to the model until the output streams match. Having two different and independent models and comparing the output streams provides higher accuracy than having results from only one dataset / model combination that has not been validated against something else (i.e., another dataset / model combination).

[0030] Downhole single-point monitoring (e.g., permanent downhole gauges (PDG), downhole flow meters (DHFM), etc.) and distributed sensor systems (e.g., permanently / temporarily installed fiber optic cables) provide qualitative online well monitoring. However, improved quantitative predictions can be achieved according to one or more embodiments described herein to optimize multi-sensor data utilization and enhance state estimation, model predictive control, and efficient operation of petroleum engineering systems. Existing methods are not well-suited for real-time assimilation of multi-scale downhole multi-sensor data (such as PDG, DHFM, DTS, DAS), distributed strain sensing data, and imaging data.

[0031] The techniques described in this paper provide a predictive analytics framework that assimilates heterogeneous downhole multi-scale, multi-sensor data and provides actionable information for real-time monitoring, optimization, and control of physical petroleum engineering systems as described herein. One or more implementations described herein efficiently extract actionable information from multi-length / time / frequency scale data at the well and reservoir levels and utilize predictive analytics and big data monitoring and control to build effective decision support systems for early detection, diagnosis, and prognosis (mitigation).

[0032] In some examples, machine learning is applied to extract streaming features from heterogeneous, multi-sensor big data within downhole single-point and distributed sensing data architectures. Numerical simulation and machine learning can be coupled to develop probabilistic surrogate models using simulation analysis. By leveraging real-time data collected from actual physical petroleum engineering systems, a feedback loop can be established between the actual petroleum engineering system and the simulation model. By assimilating this data from the actual physical petroleum engineering system, the simulation system can continuously adjust / improve itself, for example, to estimate state variables and corresponding model parameters based on model / observation discrepancies, in order to achieve more accurate predictions. For example, distributed and cloud computing can be used to perform short- and medium-range probabilistic predictions. Mathematical and network infrastructure tools can be integrated with physical multi-scale sensing data platforms for proactive information gathering and subsequent decision support for optimization strategies.

[0033] According to one or more embodiments described herein, this technology enables the real-time monitoring, optimization, and control of physical petroleum engineering systems (e.g., surface components) using real-time distributed multi-sensor data fusion and assimilation methods. This technology can be recommended to alleviate decision-making workload and achieve efficiency.

[0034] Figure 3 A flowchart is depicted of a method 300 for multi-sensor data interpretation and predictive analysis for well intervention according to one or more embodiments described herein. Method 300 may be performed by any suitable processing system (e.g., processing system 12), any suitable processing device (e.g., one of processors 21), and / or combinations thereof, or another suitable system or device.

[0035] At box 302, processing system 12 analyzes the first dataset by applying a first dataset to a first model to generate a first result. At box 304, processing system 12 analyzes the second dataset by applying a second dataset to a second model to generate a second result. The first and second datasets may include data from one or more downhole sensors (e.g., Figure 1The measurement tool 11) collects data about the downhole environment, which is collected by one or more downhole sensors. The first and / or second datasets may be collected by optical sensors and / or single-point sensors. According to one or more embodiments described herein, the first dataset may be received from one or more temperature sensors associated with a surface component, and the second dataset may be received from one or more acoustic sensors associated with a surface component. It should be understood that other types of data may be used. For example, the dataset may include one of force data, pressure data, temperature data, acoustic data, and nuclear data. In some examples, the first and second datasets are collected from the same wellbore; however, in other examples, the first and second datasets may be collected from different wellbores. Figure 4 Examples of how to analyze data such as a first dataset and / or a second dataset are described. Figure 4 This paper illustrates a computer-implemented technique for assimilating observed multi-sensor data with a computer-implemented physical and data-driven model or algorithm representing well operations (smart completion and wellbore intervention operations).

[0036] Specifically, Figure 4 A flowchart depicts a method 400 for multi-sensor data interpretation and predictive analysis for well intervention according to one or more embodiments described herein. This example refers to the analysis of a first dataset (e.g., Figure 3 This example is described using box 302, but it is also applicable to analyzing other datasets, such as analyzing a second dataset (e.g., Figure 3 (Box 304). Data is received as system input u into the Intelligent Completion Wellbore Intervention (I-CWI) system and operation box 402, which continuously collects information about wellbore operations (e.g., measurements taken by physical sensors 403a-403n, such as measuring tool 11) via measurements taken by physical sensors 403a-403n. Figure 1 The data comes from the wellbore operating system 100. At box 404, multi-sensor data fusion and preprocessing are performed on the measurements from physical sensors 403a-403n, and a filtered pressure and temperature output (referred to as downhole flowmeter (DHFM) data) is output. Specifically, box 404 provides a method for combining / fusion sensor data from physical sensors 403a-403n, where preprocessing is performed in box 404 to clean the data. Data cleaning may include filtering out unwanted data, such as noise or outliers. Preprocessing may also include determining acoustic emission (AE) (e.g., AE root mean square (AE-RMS), sound velocity profile, feature extraction, etc.) and temperature and pressure (TP) information.

[0037] As shown in the figure, this type of data is iteratively input into the observed sensor data frames 405a, 405b, ..., 405n, which have noise.

[0038] In one example, at the first iteration, the observed sensor data box 405a with noise outputs pressure, temperature, and / or DHFM information to the production mismatch determination box 406, where it is determined whether the data from box 405a indicates a production mismatch. If so, the pressure and temperature data are fed into the stochastic acoustic modeling box 442, and a calculated flow rate Q is generated and input into the performance evaluator box 408. The performance evaluator box 408 determines whether the output of box 406 matches the expected output from the model. If the output of box 442 matches the expected output from the model, an output stream 410 is generated based on the output of box 442 and the injection logging (ILT) and production logging (PLT) data 412. The output stream 410 may include multiple phase values ​​Q. This is the flow rate in the wellbore (or any material cutting into the wellbore).

[0039] If the output of box 442 does not match the expected output from the model as determined by performance evaluator box 408, or if a production mismatch is determined at decision box 406, the model can be updated by using statistical and / or machine learning techniques in adaptive optimization box 414 to update the model parameters. Figure 5 Further details of the adaptive optimization box 414 are shown in more detail in section 12.

[0040] Figure 5 A block diagram of a system 500 for optimizing well intervention operations using multi-sensor data assimilation and predictive analysis according to one or more embodiments described herein is depicted. The I-CWI operation block uses multiple sensors (i.e., multiple sensors) 504 to collect data on wellbore operations. Data is passed from the multiple sensors 504 to a probabilistic sensor and signal processor 506, which performs preprocessing to configure noise. That is, the probabilistic sensor and signal processor 506 determines how much noise has been added. The system 500 then generates an output 508, such as flow rate or another suitable output. The system 500 also provides the output 508 to an I-CWI control and decision support system 510 for making drilling or completion decisions. Therefore, the I-CWI control and decision support system uses the output 508 of the system 500 for multi-sensor data assimilation and predictive analysis to optimize and improve well intervention operations.

[0041] Figure 6A flowchart is depicted for a method 600 for optimizing well intervention operations using multisensor data assimilation and predictive analysis according to one or more embodiments described herein. At box 602, the initial state and model parameter probability density function (PDF) are received and applied to the I-CWI dynamic state-space model at box 604. PDF is a measure of the noise added to the model parameters. At box 606, the multisensor data and the I-CWI dynamic model are integrated using timed multisensor data 607 with spatial and temporal aspects. I-CWI diagnostics and predictions are performed accordingly at boxes 610 and 612 to evaluate and analyze current data (diagnostics) and future data (predictions). The results are output at box 614.

[0042] Figure 7 A block diagram of a system 700 for optimizing well intervention operations using multi-sensor data assimilation and predictive analysis, according to one or more embodiments described herein, is depicted. I-CWI state parameters 702 (e.g., temperature, pressure, flow rate, etc.) and I-CWI model parameters 704 (e.g., reservoir, rock physics, heat, fluid, and design characteristics) are input into an I-CWI dynamic state-space model 710 (see [link to documentation]). Figure 6 (See box 604). The I-CWI state-space dynamic model 710 also receives I-CWI process noise (or interference) 706 and I-CWI observation noise 708. The I-CWI state-space dynamic model 710 uses a reservoir model 712, a smart completion system model 714, a well intervention operation model 716, a permanent downhole instrument (PDG) / DHFM sensor noise model 718, and a distributed sensor signal noise model 720 to process the received information to generate one or more I-CWI prior probability density functions 722 in the form of flow rates.

[0043] Figure 8 A block diagram of a system 800 for optimizing well intervention operations using multi-sensor data assimilation and predictive analysis, according to one or more embodiments described herein, is depicted. System 800 receives I-CWI downhole multi-sensor data 802 as described herein via a joint I-CWI state and parameter estimator 806. The joint I-CWI state and parameter estimator 806 updates the data based on an I-CWI dynamic state-space model 804 using a state variable updater 808 and a model parameter updater 810 (see box 802). The results of the joint I-CWI state and parameter estimator 806 are fed into a probabilistic sampler 812, which detects the future behavior of the state variables and model parameters. The results of the probabilistic sampler 812 can be used to update / adjust the I-CWI dynamic state-space model 804.

[0044] Figure 9A block diagram of a system 900 for optimizing well intervention operations using multi-sensor data assimilation and predictive analysis, according to one or more embodiments described herein, is depicted. A state variable updater 902 updates the state variables to generate a posterior probability density function and a new probability density function. An I-CWI model parameter updater 904 uses the estimated probability density function to generate an updated probability density function. These processes are now described further.

[0045] An initial probability density function 910 is created for I-CWI parameter sampling and used by the I-CWI dynamic state-space model 912 to generate a prior probability density function based on past data. A state variable updater 902 receives timing sensor data 914 at the sensor data integrator. The timing sensor data 914 is fused by the sensor data integrator 916, and a posterior probability density function is generated by the sensor data integrator 916 based on the prior probability density function generated by the I-CWI dynamic state-space model 912. The state variable updater 902 also uses the current I-CWI model parameters 918 to generate the posterior probability density function using the prior probability density function. The state variable updater is then fed into the performance evaluator 408 (see...). Figure 4 The performance of the posterior probability density function is evaluated at box 408. If there is no data mismatch at box 408, the state variable updater 902 resamples the probability density function at box 920 to generate a new probability density function. If there is a data mismatch at box 408, the I-CWI model parameter updater 904 uses the estimated probability density function (as flow rate) to generate an updated probability density function using the model parameter updater 922 and the timing multi-sensor data 924.

[0046] Figure 10A A fused multi-sensor data 1010 from the shown downhole sensors 1001, 1002, 1003, and 1004 is depicted. For example, Figure 10B A fused multi-sensor 1010 is depicted, generated from a PDF / DHFM sensor 1011, a downhole temperature sensor (DTS) 1012, a downhole acoustic sensor (DAS) 1013, a downhole strain sensor (DSS) 1014, and / or other downhole sensors 1015.

[0047] Figure 11 A block diagram of a system 1100 for optimizing well intervention operations using multi-sensor data assimilation and predictive analysis, according to one or more embodiments described herein, is depicted. System 1100 is similar to... Figure 9 System 900. System 1100 uses fused multi-sensor data 1102 instead of timing sensor data 914. The fused multi-sensor data 1102 can be measured by downhole sensors 1, 2, ... N.

[0048] Figure 12 A block diagram of a system 1200 for optimizing well intervention operations using multi-sensor data assimilation and predictive analysis, according to one or more embodiments described herein, is depicted. In this example, the fusion prior probability density function 1202 is generated by an I-CWI dynamic state-space model 1204. The I-CWI dynamic state-space model 1204 uses an I-CWI model 1204 and an I-CWI observation model 1208 to generate the fusion prior probability density function 1202 using I-CWI state parameters 1210, I-CWI model parameters 1212, I-CWI process noise 1214, and I-CWI observation noise 1216.

[0049] Continue to refer to Figure 4 Once the adaptive optimization at box 414 has been performed, the parameter P to be updated can be identified, for example, based on the model to be updated. In some examples, one parameter in parameter P is updated at a time. In some examples, one or more parameters in parameter P are updated iteratively. The model is updated for the identified parameter P at box 416. At box 418, inputs, such as those from the I-CWI system 402 (e.g., reservoir, completion, and wellbore parameters), are used to perform the experimental design at box 420 using, for example, Monte Carlo simulation and Latin hypercube sampling. An accurate simulator 422 performs a simulation at box 424 to build an alternative model via a database including input and output data. Adaptive sampling 426 is performed on the alternative model built at box 424 to perform a computationally fast and accurate simulation at box 428, thereby identifying the optimal input configuration at decision box 430. Specifically, at decision box 430, it is determined whether model convergence has occurred. If not, the database is updated (box 424); but if so, the entire alternative I-CWS model is updated at box 431. The model output is then combined with the output from the fast physics model 432 to perform a multi-I-CWI system simulation model 440, which is used to perform production mismatch determination at decision box 406, as described herein.

[0050] Fast physics model 432 takes the system input data u and the identified parameters P as inputs and generates an initial I-CWI fast physics model at box 434. Uncertainty properties are generated at box 436 and used to perform multiple I-MSW system model generation at box 438. The results of fast physics model 432 are then input into the multi-I-CWI system simulation model 440.

[0051] According to one or more implementation schemes described herein, the first dataset and the second dataset can be analyzed simultaneously.

[0052] Continue to refer to Figure 3At box 306, the processing system 12 performs validation on the first and second models by comparing a first result (e.g., output stream 410 generated from the first dataset using the first model) with a second result (e.g., output stream 410 generated from the second dataset using the second model). This comparison determines whether the first and second results match. In some examples, a match can be determined by the first result being within a threshold tolerance of the second match (e.g., 1%, 3%, 5%, 10%, etc.). In other examples, the match is based on an exact match between the models.

[0053] In response to a match between the first and second results, the processing system 12 modifies the surface component (e.g., at block 308) based on at least one of the first or second results. Figure 1 The operation of the drilling rig 8). Modifying the operation may include modifying the trajectory of tool 7, the weight on the head of tool 7, the rotation speed of tool 7, etc. The operation can be modified in real time or near real time according to one or more embodiments described herein.

[0054] In response to a mismatch between the first and second results, processing system 12 updates at least one of the first or second models at box 310, such as... Figure 4 -As shown in Figures 1 to 12 and as described with respect to the figures. For example, one or more of machine learning, optimization, rapid physics modeling, etc., can be used to update the first model and / or the second model.

[0055] In some examples, after updating the first model and / or the second model, the processing system 12 analyzes the first dataset by applying the first dataset to the first model to generate a third result. Similarly, in some examples, after updating the first model and / or the second model, the processing system 12 analyzes the second dataset by applying the second dataset to the second model to generate a fourth result. In such examples, the first and second datasets are used to generate new results (i.e., the third and fourth results) using the updated model. The processing system 12 can then use the new results to revalidate the first and second models by comparing the new results (i.e., comparing the third and fourth results). If the third result matches the fourth result, the operational action can be modified based on at least one of the third or fourth results. This results in improved drilling operations by making drilling decisions using the updated model. If the third result does not match the fourth result, the model can be further updated, such as iteratively, until the model matches.

[0056] Additional processes may also be included, and it should be understood that... Figure 3 The processes depicted are illustrative, and other processes may be added or existing processes may be removed, modified or rearranged without departing from the scope of this disclosure.

[0057] Exemplary embodiments of this disclosure include or produce various technical features, effects, and / or improvements to the technology. Exemplary embodiments of this disclosure provide technical solutions for multi-sensor data interpretation and predictive analysis for well interventions. These solutions provide data and model validation and use data analysis results from different models to modify operational actions, for example, thereby improving the efficiency of surface components. The accuracy of the results is determined by comparing the results of the data analysis. More accurate, validated results improve downhole exploration and production operations and increase hydrocarbon recovery rates in hydrocarbon reservoirs compared to conventional techniques.

[0058] Additional implementation schemes are also provided. For example, a method 1300 for processing downhole multisensor data representing well operations is provided. A physical notification model representing a function of well operations is integrated into one or more computer processing systems (box 1302). The computer processing system uses the physical notification model to fit constants associated with at least one of reservoir rock, fluid, heat, completion, and wellbore intervention characteristics. The computer processing system uses the physical notification model to generate a prior probability distribution function. The computer processing system uses the physical notification model to iteratively fuse input data from at least a first downhole sensor and a second downhole sensor with the prior probability distribution function to generate a posterior probability distribution function. The computer processing system uses the physical notification model to process the posterior probability distribution function to generate an output in real time. The output can be one or more of a zone pressure, temperature, and strain profile prediction. The output can also be, or alternatively, an estimate of zone flow allocation. At least one dynamic model associated with zone flow allocation can be applied to the probability distribution function.

[0059] In the example, method 1300 may further include determining noise artifact events through analysis of input data from a first downhole sensor, and using the noise artifact events to filter input data from a second sensor. The method may also include an iterative process for developing a dynamic completion wellbore intervention (CWI) model and a dynamic downhole multisensor measurement model. The iterative process is performed by: defining estimates of state variables and uncertain parameters, which become elements of a state vector; defining nominal time propagation of the estimated variables and uncertain parameters; defining a nominal downhole multisensor measurement model; defining the relationships between the estimated state variables, uncertain parameters, and downhole multisensor measurement results; and validating the nominal dynamic physical drive model and the nominal downhole multisensor measurement results.

[0060] In this example, the database of downhole multi-sensor measurement results is used to empirically validate the nominal dynamic physical notification model and the nominal downhole multi-sensor measurement model. In this example, the database contains multiple downhole sensor measurement results and their corresponding direct measurement results.

[0061] In some examples, method 1300 also includes combining input signals from a first downhole sensor and a second downhole sensor. This can result in one or more of the generated measurements of phase volumetric flow rate measurement and / or phase volumetric flow rate variability measurement. Input data from the first downhole sensor may include permanent downhole instrumentation measurements. Method 1300 may also include determining noise artifact events through analysis of the input data from the first downhole sensor, and using the noise artifact events to filter the input data from the second sensor.

[0062] Another embodiment is provided, which includes a real-time (or near-real-time) state variable estimator. The real-time (or near-real-time) state variable estimator includes multiple downhole sensor measurement inputs. At least the measurement inputs are configured to receive inputs indicating zone pressure and temperature values. The state variable estimator also includes multiple control inputs, a particle filter, and an optimization algorithm. The particle filter and optimization algorithm are configured to receive the multiple measurement inputs and multiple control inputs to provide optimal estimates of the state variables and model parameters of the wellbore intervention operation in real-time or near real-time. The state variable estimator also includes a system for receiving dynamic model parameter values ​​adjusted by a particle filter-based optimizer. In the example, the state variable estimator also includes a decision support system configured to use the state variable estimator to determine performance degradation and derive a decision. In the example, the state variable estimator also includes an intelligent control system configured to update the operation of the completion and wellbore intervention system based on the decision. In such examples, the decision can predict future completion and wellbore intervention operation conditions and perform dynamic optimization when performance degrades below a predetermined threshold.

[0063] The following are some of the aforementioned publicly disclosed implementation schemes:

[0064] Implementation Scheme 1: A method comprising: analyzing the first dataset by applying it to a first model to generate a first result; analyzing the second dataset by applying it to a second model to generate a second result; performing validation on the first model and the second model by comparing the first result with the second result; modifying the operational action of a surface component based on at least one of the first result or the second result in response to determining that the first result and the second result match; and updating at least one of the first model or the second model in response to determining that the first result and the second result do not match.

[0065] Implementation Scheme 2: According to any previous implementation scheme, the method further includes, after updating at least one of the first model or the second model: analyzing the first dataset by applying the first dataset to the first model to generate a third result.

[0066] Implementation Scheme 3: The method according to any prior implementation scheme, the method further comprising, after updating at least one of the first model or the second model: analyzing the second dataset by applying the second dataset to the second model to generate a fourth result.

[0067] Implementation Scheme 4: The method according to any prior implementation scheme, the method further comprising: after updating at least one of the first model or the second model: re-performing the validation of the first model and the second model by comparing the third result with the fourth result.

[0068] Implementation Scheme 5: The method according to any prior implementation scheme, the method further comprising: after updating at least one of the first model or the second model: in response to determining that the third result and the fourth result match, modifying the operation action based on at least one of the third result or the fourth result.

[0069] Implementation Scheme 6: The method according to any prior implementation scheme, the method further comprising: after updating at least one of the first model or the second model: in response to determining that the third result and the fourth result do not match, updating at least one of the first model or the second model.

[0070] Implementation Scheme 7: The method according to any of the previous implementation schemes, wherein an update to at least one of the first model or the second model is performed iteratively.

[0071] Implementation Scheme 8: The method according to any of the previous implementation schemes, wherein machine learning is used to perform an update on at least one of the first model or the second model.

[0072] Implementation Scheme 9: The method according to any of the previous implementation schemes, wherein optimization techniques are used to perform an update on at least one of the first model or the second model.

[0073] Implementation Scheme 10: The method according to any previous implementation scheme, wherein a fast physics technique is used to perform an update on at least one of the first model or the second model.

[0074] Implementation Scheme 11: The method according to any prior embodiment, the method further comprising: receiving the first dataset from one or more temperature sensors associated with the surface component; and receiving the second dataset from one or more acoustic sensors associated with the surface component.

[0075] Implementation Scheme 12: A system comprising: a surface component; and a processing system comprising: a memory including computer-readable instructions; and a processing device for executing the computer-readable instructions, the computer-readable instructions controlling the processing device to perform operations, the operations comprising: analyzing the first dataset by applying a first dataset to a first model to generate a first result; analyzing the second dataset by applying a second dataset to a second model to generate a second result; performing validation on the first model and the second model by comparing the first result with the second result; modifying the operational action of the surface component based on at least one of the first result or the second result in response to determining that the first result and the second result match; and updating at least one of the first model or the second model in response to determining that the first result and the second result do not match.

[0076] Implementation Scheme 13: The system according to any of the previous implementation schemes, wherein the first output is a first flow rate, and wherein the second output is a second flow rate.

[0077] Implementation Scheme 14: The system according to any of the previous embodiments, wherein the operation further includes: receiving the first dataset from one or more temperature sensors associated with the surface component; and receiving the second dataset from one or more acoustic sensors associated with the surface component.

[0078] Implementation Scheme 15: A system according to any of the previous embodiments, wherein the first dataset includes one of force data, pressure data, temperature data, acoustic data, and nuclear data, and wherein the second data includes another of the force data, the pressure data, the temperature data, the acoustic data, and the nuclear data.

[0079] Implementation Scheme 16: The system according to any prior implementation scheme, wherein the operation further includes: analyzing the third dataset by applying the third dataset to the third model to generate a third result; and performing verification on the third model by comparing the third result with one or more of the first result and the second result.

[0080] Implementation Scheme 17: The system according to any of the previous implementation schemes, wherein the operation is performed in real time.

[0081] Implementation Scheme 18: The system according to any of the previous embodiments, wherein the first dataset includes data collected from a first wellbore, and wherein the second dataset includes data collected from a second wellbore.

[0082] Implementation Scheme 19: A method for processing downhole multi-sensor data representing well operations, the method comprising: using a physical notification model representing a well operation encoded into one or more computer systems; using a fitting constant associated with at least one of: reservoir rock, fluid, heat, completion, and wellbore intervention characteristics; using the physical notification model to generate a prior probability distribution function; repeatedly fusing input data from at least a first downhole sensor and a second downhole sensor with the prior probability distribution function to generate a posterior probability distribution function; and processing the posterior probability distribution function using one or more computer systems to generate an output in real time.

[0083] Implementation Scheme 20: A method according to any prior embodiment, further comprising determining noise artifact events by analyzing the input data from a first downhole sensor; and filtering the input data from a second sensor using the noise artifact events.

[0084] In the context of describing this disclosure (particularly in the context of the appended claims), the terms “an,” “a,” and “the,” and similar designations, should be interpreted to cover both singular and plural forms, unless otherwise specified herein or clearly contradicted by the context. Furthermore, it should be noted that the terms “first,” “second,” etc., used herein do not indicate any order, quantity, or importance, but are used to distinguish one element from another. The modifier “about,” used in conjunction with quantity, includes the stated value and has a meaning determined by the context (e.g., it includes the degree of error associated with a particular quantity of measurement).

[0085] The teachings of this disclosure can be applied to a variety of well operations. These operations may involve treating a formation, fluids residing in the formation, the wellbore, and / or equipment within the wellbore, such as production tubing, with one or more treatment agents. Treatment agents can be in the form of liquids, gases, solids, semi-solids, and mixtures thereof. Exemplary treatment agents include, but are not limited to, fracturing fluids, acids, steam, water, brine, corrosion inhibitors, binders, permeability modifiers, drilling mud, emulsifiers, demulsifiers, tracers, flow improvers, etc. Exemplary well operations include, but are not limited to, hydraulic fracturing, production enhancement, tracer injection, cleaning, acidizing, steam injection, water injection, cementing, etc.

[0086] While this disclosure has been described with reference to one or more exemplary embodiments, those skilled in the art will understand that various changes may be made and equivalents may be substituted for elements therein without departing from the scope of this disclosure. Furthermore, many modifications may be made to adapt particular situations or materials to the teachings of this disclosure without departing from the basic scope of this disclosure. Therefore, it is contemplated that this disclosure is not limited to the specific embodiments disclosed as the best mode contemplated for carrying out this disclosure, but rather that this disclosure will include all embodiments falling within the scope of the claims. Additionally, exemplary embodiments of this disclosure have been disclosed in the drawings and detailed descriptions, and although specific terminology has been used, it is used in a general and descriptive sense only, and not for limiting purposes, unless otherwise specified, and the scope of this disclosure is therefore not limited thereto.

Claims

1. A method (300), the method comprising: The first dataset is analyzed by applying it in real time to a first model to generate a first result related to continuous wellbore operations occurring within the wellbore. The first model includes first model parameters. The second dataset is analyzed by applying a second dataset, which is different from the first dataset, to the second model in real time to generate a second result related to continuous wellbore operations occurring within the wellbore. The second model includes second model parameters. Validation of the first model and the second model is performed by comparing the first result with the second result; In response to determining that the first result and the second result match, control signals are generated and transmitted based on at least one of the first result or the second result to modify continuous operation actions of at least one of the surface components (8) or downhole tools in relation to continuous wellbore operations occurring in real time within the wellbore. as well as In response to determining that the first result and the second result do not match, at least one of the first model parameters or the second model parameters is updated by constructing an alternative model, wherein the alternative model provides at least one of the updated first model parameters or the updated second model parameters; The analysis of the first dataset, the analysis of the second dataset, the performance of verification, modification, and updates are all performed in real time. At least one of the first dataset and the second dataset is received from a plurality of downhole sensors, the plurality of downhole sensors including at least a first downhole sensor and a second downhole sensor, and the analysis of at least one of the first dataset and the second dataset further includes: Noise artifact events are identified by analyzing data from the first downhole sensor, and these noise artifact events are used to filter data from the second downhole sensor. The method further includes: Multi-sensor data fusion and preprocessing are performed on the first dataset or the second dataset to obtain a filtered first dataset or a filtered second dataset; Perform production mismatch determination on the filtered first dataset or the filtered second dataset; If either the filtered first dataset or the filtered second dataset indicates a production mismatch, then update the corresponding model in the first model and the second model.

2. The method (300) of claim 1, the method further comprising: After updating at least one of the first model or the second model: The first dataset is analyzed in real time by applying it to the first model to generate a third result.

3. The method (300) according to claim 2, further comprising: After updating at least one of the first model or the second model: The second dataset is analyzed in real time by applying it to the second model to generate a fourth result.

4. The method (300) according to claim 3, further comprising: After updating at least one of the first model or the second model: The first model and the second model are re-validated in real time by comparing the third result with the fourth result.

5. The method (300) according to claim 4, further comprising: After updating at least one of the first model or the second model: In response to determining that the third result and the fourth result match, the continuous operation action is modified in real time based on at least one of the third result or the fourth result.

6. The method (300) according to claim 4, further comprising: After updating at least one of the first model or the second model: In response to determining that the third result and the fourth result do not match, at least one of the first model or the second model is updated.

7. The method (300) of claim 1, wherein an update to at least one of the first model or the second model is performed iteratively.

8. The method (300) of claim 1, wherein machine learning is used to perform an update on at least one of the first model or the second model.

9. The method (300) of claim 1, wherein optimization techniques are used to perform an update on at least one of the first model or the second model.

10. The method (300) of claim 1, wherein a fast physics technique is used to perform an update on at least one of the first model or the second model.

11. The method (300) according to claim 1, further comprising: The first dataset is received from multiple temperature sensors (1012) associated with the surface component (8); as well as The second dataset is received from multiple acoustic sensors (1013) associated with the surface component (8).

12. A system (100) comprising: At least one of the surface components (8) or downhole tools is associated with continuous wellbore operations occurring within the wellbore; and Processing system (12), the processing system comprising: Memory (24), the memory including computer-readable instructions; and Processing device (21) for executing the computer-readable instructions, the computer-readable instructions controlling the processing device (21) to perform operations, the operations including: The first dataset is analyzed by applying it in real time to a first model to generate a first result related to continuous wellbore operations occurring within the wellbore. The first model includes first model parameters. The second dataset is analyzed by applying a second dataset, which is different from the first dataset, to the second model to generate a second result related to continuous wellbore operations occurring within the wellbore. The second model includes second model parameters. Validation of the first model and the second model is performed by comparing the first result with the second result; In response to determining that the first result and the second result match, a control signal is generated and transmitted based on at least one of the first result or the second result to modify the continuous operation actions of the surface component (8) related to continuous wellbore operations occurring in real time within the wellbore; and In response to determining that the first result and the second result do not match, at least one of the first model parameters or the second model parameters is updated by constructing an alternative model, wherein the alternative model provides at least one of the updated first model parameters or updated second model parameters. At least one of the first dataset and the second dataset is received from a plurality of downhole sensors, the plurality of downhole sensors including at least a first downhole sensor and a second downhole sensor, and the analysis of at least one of the first dataset and the second dataset further includes: Noise artifact events are identified by analyzing data from the first downhole sensor, and these noise artifact events are used to filter data from the second downhole sensor. The computer-readable instructions also control the processing device (21) to perform operations including: Multi-sensor data fusion and preprocessing are performed on the first dataset or the second dataset to obtain a filtered first dataset or a filtered second dataset; Perform production mismatch determination on the filtered first dataset or the filtered second dataset; If either the filtered first dataset or the filtered second dataset indicates a production mismatch, then update the corresponding model in the first model and the second model.

13. The system (100) of claim 12, wherein the first result is a first flow rate, and wherein the second result is a second flow rate.

14. The system (100) of claim 12, wherein the operation further comprises: The first dataset is received from multiple temperature sensors (1012) associated with the surface component (8); as well as The second dataset is received from multiple acoustic sensors (1013) associated with the surface component (8).

Images