CROSS-REFERNCE TO RELATED APPLICATIONS

[0001] This application claims priority to and the benefit of U.S. Provisional Patent Application No. 63/406,290, filed September 14, 2022, and is incorporated herein by reference in its entirety for all purposes.

FIELD OF INVENTION

[0002] The present application relates generally to oil and gas well drilling and well operations, including injection and waste wells, throughout the lifetime of the well. More specifically, to a method and a system using frequency domain analysis for rig state detection in well operations, such as drilling, completion or workover/intervention related operations insofar these operations are monitored by sensor data, and that the sensor data is available as time series data.

BACKGROUND

[0003] Drilling and well operations in oil and gas wells are expensive operations. The cost is typically several tens to several hundred thousand dollars per day, and a delayed or failed operation may ruin the well's production. These operations are also prone to a high percentage of non-productive time, often in the range of 10 to 20% of the total operations time. Some of this non-productive time also poses risk for injuries, loss of life and damage to the environment. Several steps are implemented to optimize operations, including making a careful account of the activities (e.g., rig states) executed on the rigs to benchmark rigs, crews and methodologies, and applying mathematical and physical models to the operations to either optimize operations or identify risks.

[0004] For both benchmarking and modeling, careful identification of the actual rig states is important. For benchmarking, an accurate identification of the rig state and its start and end time is essential. For operational modelling, the operational context given by the rig state is in many cases instrumental for correct estimations and decision making (whether real time or by the use of digital twins). SUMMARY

[0005] The invention consists of a method and system for identifying rig states and events using spectral analysis techniques. Time series data is transformed into a frequency domain representation and processed into a form compatible with automated detection methods, which could be rules-based or systems based on machine learning (statistical) classification models, such as logistic regression or tree -based models. Ensembles of gradient-boosted trees are one example of a default machine learning model used for detection. Machine learning models are trained to recognize specific rig states or events observed in the surface measurements of oil and gas wells, after processing these sensor measurements into appropriate spectral features based on moving windows sliced from the times series data.

[0006] The invention enhances situational awareness by providing an accurate estimate of current rig states based on the detection method’s outputs, and can support other software in automated decision-making by providing operational context in real-time.

[0007] Other aspects of the disclosure will become apparent by consideration of the detailed description and accompanying drawings.

DEFINITIONS

[0008] As used herein, the terms “processor” and “central processing unit” or “CPU” are used interchangeably and refer to a device that is able to read a program from a computer memory (e.g., ROM or other computer memory) and perform a set of steps according to the program. As used herein, the term “processor” (e.g., a microprocessor, a microcontroller, a processing unit, or other suitable programmable device) can include, among other things, a control unit, an arithmetic logic unit (“ALC”), and a plurality of registers, and can be implemented using a known computer architecture (e.g., a modified Harvard architecture, a von Neumann architecture, etc.). In some embodiments the processor is a microprocessor that can be configured to communicate in a stand-alone and/or a distributed environment, and can be configured to communicate via wired or wireless communications with other processors, where such one or more processor can be configured to operate on one or more processor-controlled devices that can be similar or different devices. [0009] As used herein, the term “memory” is any memory storage and is a non-transitory computer readable medium. The memory can include, for example, a program storage area and the data storage area. The program storage area and the data storage area can include combinations of different types of memory, such as a ROM, a RAM (e.g., DRAM, SDRAM, etc.), EEPROM, flash memory, a hard disk, a SD card, or other suitable magnetic, optical, physical, or electronic memory devices. The processor can be connected to the memory and execute software instructions that are capable of being stored in a RAM of the memory (e.g., during execution), a ROM of the memory (e.g., on a generally permanent bases), or another non- transitory computer readable medium such as another memory or a disc. In some embodiments, the memory includes one or more processor-readable and accessible memory elements and/or components that can be internal to the processor-controlled device, external to the processor- controlled device, and can be accessed via a wired or wireless network. Software included in the implementation of the methods disclosed herein can be stored in the memory. The software includes, for example, firmware, one or more applications, program data, filters, rules, one or more program modules, and other executable instructions. For example, the processor can be configured to retrieve from the memory and execute, among other things, instructions related to the processes and methods described herein.

[0010] As used herein, the term “computer readable medium” refers to any device or system for storing and providing information (e.g., data and instructions) to a computer processor. Examples of computer readable media include, but are not limited to, DVDs, CDs, hard disk drives, magnetic tape and servers for streaming media over networks, whether local or distant (e.g., cloud-based).

[0011] “About” and “approximately” are used to provide flexibility to a numerical range endpoint by providing that a given value may be “slightly above” or “slightly below” the endpoint without affecting the desired result.

[0012] The term “coupled,” as used herein, is defined as “connected,” although not necessarily directly, and not necessarily mechanically. The term coupled is to be understood to mean physically, magnetically, chemically, fluidly, electrically, or otherwise coupled, connected or linked and does not exclude the presence of intermediate elements between the coupled elements absent specific contrary language. [0013] As used herein, the term “in electronic communication” refers to electrical devices (e.g., computers, processors, etc.) that are configured to communicate with one another through direct or indirect signaling. Likewise, a computer configured to transmit (e.g., through cables, wires, infrared signals, telephone lines, airwaves, etc.) information to another computer or device, is in electronic communication with the other computer or device. As used herein, the term “transmitting” refers to the movement of information (e.g., data) from one location to another (e.g., from one device to another) using any suitable means.

[0014] As used herein, the term “network” generally refers to any suitable electronic network including, but not limited to, a wide area network (“WAN”) (e.g., a TCP/IP based network), a local area network (“LAN”), a neighborhood area network (“NAN”), a home area network (“HAN”), or personal area network (“PAN”) employing any of a variety of communications protocols, such as Wi-Fi, Bluetooth, ZigBee, etc. In some embodiments, the network is a cellular network, such as, for example, a Global System for Mobile Communications (“GSM”) network, a General Packet Radio Service (“GPRS”) network, an Evolution-Data Optimized (“EV -DO”) network, an Enhanced Data Rates for GSM Evolution (“EDGE”) network, a 3GSM network, a 4GSM network, a 5G New Radio, a Digital Enhanced Cordless Telecommunications (“DECT”) network, a digital AMPS (“IS-136/TDMA”) network, or an Integrated Digital Enhanced Network (“iDEN”) network, etc.

[0015] As used herein, the term “machine learning” refers to any suitable technique for data analysis that leverage data to improve performance of a task. Examples of machine learning models include artificial neural networks, decision trees, support-vector machines, regression analysis, Bayesian networks, etc.

[0016] As used herein, the term “frequency domain conversion” refers to any suitable technique for converting time series data into frequency domain data. Examples of frequency domain conversion include Fourier Transform based methods (e.g., Periodogram method, Welch’s method), and methods using Wavelet Transforms.

BRIEF DESCRIPTION OF THE DRAWINGS

[0017] FIG. 1 is a schematic of a system including a drilling rig. [0018] FIG. 2 is a graph illustrating characteristics of a downlinking rig state.

[0019] FIG. 3 is a graph illustrating variable tension in the hookload sensor time series data.

[0020] FIG. 4 illustrates a method according to one aspect of the invention.

[0021] FIG. 5 illustrates a parallel process in which trigger events are detected in cleaned and filtered data.

[0022] FIG. 6 is a graph illustrating a time series data plot of pre-processed standpipe pressure, in which downlinks are present.

[0023] FIG. 7 is a graph illustrating a time series standpipe pressure curve illustrating the pressure oscillations of one of the downlinks.

[0024] FIG. 8 is the time series data of FIG. 7 transformed into a frequency spectrum.

[0025] FIG. 9 illustrates the data of FIG. 8 aggregated into frequency bins.

[0026] FIG. 10 illustrates the binary downlinking and not-downlinking binary statuses provided by the method of FIG. 4, corresponding to the pressure signal presented of FIG. 6.

[0027] FIG. 11 illustrates detected downlinks used in an operational hole cleaning model to suppress anomaly warnings caused by pressure fluctuations.

[0028] FIG. 12 is a schematic of a system according to one aspect of the disclosure.

[0029] FIG. 13 illustrates a method to use the detected rig state in conjunction with operational models, either in real time or as a digital twin.

[0030] FIG. 14 illustrates a hookload curve where the movement of pipe shows a high hookload, and periods when a new pipelength is connected show low hookloads and demonstrated effects of heave (e.g., large amplitude sea waves) on the measured hookload.

[0031] FIG. 15 illustrates heave status updated at end of each stand based on the data from FIG. 14.

[0032] Before any embodiments are explained in detail, it is to be understood that the invention is not limited in its application to the details of construction and the arrangement of components set forth in the following description or illustrated in the following drawings. The invention is capable of other embodiments and of being practiced or of being carried out in various ways.

DETAILED DESCRIPTION

[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. In case of conflict, the present document, including definitions, will control. Preferred methods and materials are described below, although methods and materials similar or equivalent to those described herein can be used in practice or testing of the present disclosure. All publications, patent applications, patents and other references mentioned herein are incorporated by reference in their entirety. The materials, methods, and examples disclosed herein are illustrative only and not intended to be limiting.

[0034] The terms “comprise(s),” “include(s),” “having,” “has,” “can,” “contain(s),” and variants thereof, as used herein, are intended to be open-ended transitional phrases, terms, or words that do not preclude the possibility of additional acts or structures. The singular forms “a,” “an” and “the” include plural references unless the context clearly dictates otherwise. The present disclosure also contemplates other embodiments “comprising,” “consisting of’ and “consisting essentially of,” the embodiments or elements presented herein, whether explicitly set forth or not.

[0035] For the recitation of numeric ranges herein, each intervening number there between with the same degree of precision is explicitly contemplated. For example, for the range of 6-9, the numbers 7 and 8 are contemplated in addition to 6 and 9, and for the range 6.0-7.0, the number 6.0, 6.1, 6.2, 6.3, 6.4, 6.5, 6.6, 6.7, 6.8, 6.9, and 7.0 are explicitly contemplated.

[0036] With reference to FIG. 1, a drilling rig 101 with data capture and monitoring of operations is illustrated. The operations conducted by the drilling rig 101 (or other well operating unit, such as workover rigs, coiled tubing units, wireline units or similar) are monitored by sensors attached to rig equipment 102, mud pump systems 111, 112, 113, downhole equipment/drill string 104, or near the drill bit 103. Such a system is detailed in U.S. Patent Application No. 17/205,063, filed March 18, 2021 and is incorporated herein by reference in its entirety.

[0037] The sensor data provide insights into the operations as well as the rocks 105 that are drilled through. These data are captured and stored in a computer storage 106 on the rig site. The sensor data are monitored by qualified personnel on data displays 107 in order to verify the quality of the ongoing operations, and to identify deviations or early warnings for undesired events. The data are frequently also distributed via the internet 108 to one or more locations outside the rig site 109 in oil company or contractor offices, where they are stored as time or depth-indexed series and displayed to other qualified personnel monitoring the operations.

METHOD

[0038] The methodology disclosed herein is applicable for a plurality of rig states where the impact of the rig states can be detected from a frequency domain representation but are difficult to reliably identify in the time domain. The method detailed herein is generic, and the generic nature of the methodology is exemplified by, but not limited to, two different cases: the rig state of downlinking, and the impact of wave-induced heave on floating offshore rig sensors during bad weather. These different occurrences give a practitioner of the field an understanding of the types of rig state/sensor behavior that are applicable to the invention disclosed herein. Specifically, the invention provides a method that enables the rig state to be uniquely identified and differentiated from other states.

[0039] With reference to FIG. 2, various characteristics of a downlinking rig state are illustrated. Downlinking is the rig state where communication occurs between the sensors in the downhole tools 104 exchange information with the surface computing system 106, typically using mud pulse telemetry. The communication is done by generating pulses measurable by a predefined sensor, typically pressure or torque. Examples of such pulses are illustrated in FIG. 2. Time series data is normally plotted in a diagram with multiple sensor values 201 displayed along a time axis 202 (e.g., a common time axis). The pre-defined communication sensor 203 (e.g., event standpipe pressure, green curve in FIG. 2) shows a distinctive oscillating pattern 204 in the time series plot during the downlinking rig state. However, due to natural noise and variation in the sensor values, this activity is not recognizable by the use of conventional methods for rig state detection. [0040] During well operations, the rig equipment 102 is connected to the drillstring 104.

With a floating rig, high waves cause the floating rig to heave and imposes a variable tension on the drillstring. FIG. 3 shows the manifestation of this variable tension in the hookload sensor time series data. The graph of FIG. 3 show various sensor values 301 on a time axis 302. The position of the rig equipment 102 relative to the rig 303 initially shows a normal activity before an operational standstill 304. The weight of the string fluctuates as a result of the normal operation 305. Whereas, the oscillating variation of the weight of the string at 306 during operational standstill 304 is the result of variable tension in the drillstring 104 caused by heave motion of the rig equipment 102. As such, the heave impact on the sensor data can obscure rig state detection and operational anomaly detection, and is therefore important to identify and potentially filter away.

[0041] With reference to FIG. 4, a method of the invention is illustrated. In particular, the method of FIG. 4 is based on the measurement of wellbore parameters from sensors 401 on the rig and/or in the wellbore of FIG. 1. The sensor data is conventionally stored in a database as time series data at step 402. In the method of the present invention, the time series data are retrieved from the data store and cleaned and filtered at step 403. Such cleaning typically involve removing zero or unfeasibly high or low valued observations, anomalous data such as sensor spikes or other malfunctions, and/or data when no activity takes place. The disclosed method also allows the implementation of different triggering mechanisms at step 404 using methodologies initiating the transformation of the time series data to frequency data. In the event a triggering mechanism is implemented in the invention, the cleaned and filtered data 501 of the method step 403 are evaluated by the triggering mechanism step 502. For downlinking rig state detection, the triggering event may be the onset of drilling data, for example.

[0042] With continued reference to FIG. 4, spectral analysis is used to pre-process temporal signals into a format which can be used as an input for classification models or rules-based decision logic. The transformation of time series signals into frequency domain representations at step 405 (“frequency domain conversion”) is, in some embodiments done by slicing signals into time windows, and then spectra are calculated via Fourier Transforms or similar methodologies, for example using Welch’s method. In some embodiments, window lengths are increased to improve low-frequency spectral resolution. Additional step 406 includes aggregation of spectral power densities into discrete bins, log-transforms, or normalization relative to the highest spectral component, can also be optionally applied if they improve performance on the chosen detection task. The main detection step 407 requires an appropriate discriminative system for detecting the rig state based on the processed spectral features; this system could be rules-based or utilizing machine learning techniques for classification, such as logistic regression, tree-based models or neural networks. The method also includes step 408 the supports a post-processing step to increase the rig state identification accuracy, for example by requiring multiple positive statuses before updating the rig state estimate. In the method disclosed herein, the rig states and any associated data are converted into time series data 409 before they are stored at step 411 and/or viewed in a viewer application 410.

[0043] With reference to FIG. 5, the invention also supports parallel processes, in which the trigger event 502 is detected in cleaned and filtered data 501, and initiates a multitude of trigger methodologies 503, each supporting a unique transformation model 504, following the methodology described in steps 405-407.

[0044] The methodology of the invention is exemplified by downlinking rig state detection, wherein a customized step 403 of data cleaning and filtering is implemented. In this instance of the invention, the data sampling frequency is checked and any inconsistencies and/or missing values are filtered out, cleaned or corrected, for example via interpolation. FIG. 6 illustrates a typical example of a time series data plot of pre-processed standpipe pressure 601, in which downlinks 602 are present. FIG. 7 shows an example of a time series standpipe pressure curve 701 illustrating the pressure oscillations 702 of one of the downlinks 602. The transformation trigger event at step 404 in this instance of the invention is the onset of drilling activities. For the application of the methodology of the invention for recognizing downlinking, Fourier transformation using Welch’s method has been applied to transform time series data to frequency domain data.

[0045] FIG. 8 illustrates the time series data of FIG. 7 transformed into a frequency spectrum 801. Further, the frequency domain data is processed at step 406 by aggregating them into a set of frequency intervals/bins. FIG. 9 shows data aggregated into frequency bins 901. A gradient- boosted tree model is used in the rig state detection logic at step 407 for classifying the frequency data into a binary downlinking status (1 = downlinking, 0 = not downlinking) returned by the model at each timestep. FIG. 10 shows the binary downlinking 1001 and not-downlinking binary statuses provided by the invention in this instance, corresponding to the pressure signal presented in FIG. 6. FIG. 11 illustrates an instance of the invention, wherein output from a downlinking detection system shows the binary downlinking status as a time series annotation text 1103, where the upper plot 1101 shows the standpipe pressure signal varying in time, with downlinking rig states 1102.

[0046] In some embodiments, a machine learning model is used for detecting events or rig states of interest, such as downlinking. In other embodiments, a machine learning model is not required. Heave-induced influence on hookload is another example of an application of the method disclosed herein. Figure 14 shows an example hookload curve where the movement of pipe shows a high hookload 1401, and periods when a new pipelength is connected show low hookloads 1402. During normal operations at 1403, the hookload is oscillating with different frequencies and magnitudes. Heave impact at 1404, adds an additional oscillatory component on top of normal hookload variation. It should be noted that active heave compensation may be used on rigs equipped with it, and thus the effects of heave may only be seen strongly at the start and end of stand 1404 near connection periods 1402, where the active compensation is switched off. [0047] In the method of the invention, a transformation of the time series data to spectral data, and aggregation of the frequencies into discrete bins, allows heave impact to be distinguished from normal hookload value oscillations, indicated by different status values 1501 and 1502 in Figure 15. The system based on the method of the invention first detects the extent of stands 1405 by monitoring hookload measurements; once a stand is completed, a windowed time-series of hookload measurements corresponding to the period without active heave compensation (e.g., the time period after 1406, right-most part of 1404 in FIG. 14) is processed into aggregated spectral information using the technique described in the previous paragraph for downlinking detection. For heave detection, a simple rules-based system applied to the aggregated spectral bins calculated from windowed hookload time-domain signals was sufficient for identifying strong effects of wave motion on floating rigs. As the spectra of oceanic waves fall within a particular frequency band for most cases relevant to offshore drilling, heave can be detected by a relative comparison of the aggregated spectral power density bins produced by the aforementioned pre-processing steps. In some embodiments, an output of the method is the binary rig state as a time series. The onset and end time of the detected rig state at 1503, enables the calculation of duration of the detected rig state, which enables benchmarking of the performance of the rig state. [0048] The method disclosed herein is not limited to downlinking or heave detection applications and can be applied more generally to scenarios with complex periodic signals, if the signatures of the rig states are recorded within the sensor data. This assumes the sampling rates of the sensors are sufficiently high to properly resolve the frequency components of interest without aliasing issues - sampling rates should be at least twice the frequency of the highest frequency component under consideration, according to the Nyquist theorem.

[0049] The invention also incorporates a methodology to use the detected rig state in conjunction with Operational Models, whether this operates in real time or as a Digital Twin. In the methodology of the invention output of the rig state detection model 409/1302 is consumed by the operational model 1301, wherein an evaluation of the relevance of the detected rig state 1302 is determined at step 1303. In the example of the downlinking, this is a rule -based evaluation; in the event that the downlinking state is detected, the rig state is relevant for further processing, in the event that downlinking state is “not detected”, the operational model ignores this state until a new evaluation is made for the next data point in the time series data output from the rig state detection model 1302.

[0050] In the event that the identified rig state is relevant, a rule based decision is made on the relevant action to take step 1304. Such an action may be to change the configuration or methodology of the operational model 1305. An example of such use will be changing the methodology for calculating hookload in the event that heave is detected. Another action can be to use the detected rig state to suppress an anomaly warning for operational anomaly detection models 1306. An example of such use is illustrated in FIG. 11 , where detected downlinks are used in an operational hole cleaning model to suppress anomaly warnings caused by pressure fluctuations. Another action may be to filter the data associated with the rig state away from the input data of the operational model 1307. An example of this is for operational models using breakover hookload as an input. In the event that the rig state of heave is detected by the invention, the breakover hookload will be heavily influenced by the load oscillations created by the heave and thus its “true” value obscured; in the event that rig state is heave, these data are filtered away from the input data to the Operational Model. An action may also be to provide information on the rig state to the Operational Model for information purposes, as exemplified in FIG. 11. SYSTEM

[0051] Sensor data will be pulled via a data transfer protocol from the Real Time Data Acquisition, storage and distribution computer 1201 to the computer of the invention 1202, where the different modules of the invention, 403-409 are installed. The computer of the invention 1202, may be cloud hosted, and can run (micro)service-oriented or monolithic applications. The detection systems running on 1202 utilize the information from the data streams (possibly published to a remote data store 1204), which may be provided by Message Queues, Publisher-Subscriber systems, or systems providing equivalent functionality. The status output (relating to particular rig states) of the invention can then be published and consumed by other software applications, which could include user interfaces or monitoring systems 1203 associated with the computer of the invention, or applications or monitoring systems 1205 associated with a remote data store 1204 where knowledge of particular rig states is relevant. An example user-interface visualizing a downlinking detection system’s outputs is shown in Fig 11, where messages 1103 are provided when downlinking events 1102 are detected from the standpipe pressure time-series data 1101.

[0052] Various features and advantages are set forth in the following claims.