PREDICTIVE DOWNLINK AUTOMATION PROCESS AND ARTIFICIAL INTELLIGENCE POWERED SYSTEM

Description

FIELD

The present invention relates generally to oil field drilling. More particularly, the present invention relates to downlinking to downhole drilling & logging tools such as rotary steerable system.

BACKGROUND

Downlinking in the drilling industry refers to the process of communicating from the surface or from drilling rigs to a downhole tool, such as a rotary steering system (“RSS”). Downlinking allows an operator from the surface to send various commands, such as activating or deactivating downhole reamers, adjusting steering settings of the RSS, etc., to the downhole tool. By communicating with the downhole tool from the surface, operators have control over the downhole tool so that the tool follows an intended drilling trajectory.

Conventional downlinking generates a modulated signal according to a specific sequence, which is sometimes called a model, and a processor included within the downhole tool demodulates the signal to decode the commands transmitted from the surface. For example, the processor, which is connected to a flow meter, determines a downlink sequence sent from the surface by detecting variations in flow rate through the downhole tool (for example, changes in flow rate through a turbine within the downhole tool). In another embodiment, changes to rotation speed (“RPM”) of the turbine are measured rather than flow rate. The processor correlates the changes in flow rate or rotation rate to data sequences stored in memory of the downhole tool to determine which commands were sent from the surface. Once demodulated and decoded, the downhole tool implements the new commands and continues drilling the wellbore according to the newly received commands. For example, an RSS may change drilling trajectory to follow a dogleg in a drilling path or the like.

In an RSS system, an RSS controls both the direction and inclination by which a wellbore is drilled, and the RSS direction and inclination depends on parameters provided to the RSS through downlinking. If an RSS needs adjustment, an operator downlinks new parameters to the RSS, controlling and steering the RSS through a well trajectory. Operators thereby adjust RSS parameters in response to data generated from sensors included in the downhole tool, such as gyroscope sensors, magnetometers, accelerometers, flow rate sensors, etc., or through real-time surveys. Moreover, downlinking steers the downhole tool through a wellbore path following a desired wellbore path.

Generally, downlinking is a lengthy process, sometimes taking several minutes depending on a bit-period. Downlinking conventionally uses a modulation technique to represent flow rate or rotation rate changes as digital values. The processor can sample data at regular intervals to detect modulation changes, filter the results, and obtain a data sample. Subsequently, the processor correlates data samples with data sequences stored in memory. If the correlation exceeds a threshold, the tool determines that a downlink command has been sent, accepts the downlink command, and implements the command.

Not all downlinks successfully communicate commands to the downhole equipment. An unsuccessful downlink can result in the RSS deviating from a planned wellbore path, thereby causing non-productive time or extra directional demand and unnecessary dog leg requirement in well profile to correct trajectory toward desire path. Unsuccessful or challenging downlinks are particularly prevalent in smaller holes sizes (e.g., 6 inch and 8.5 inch holes). In addition, drilling environments, mud rheology, pump harmony, and incorrectly chosen downlink types all can cause unsuccessful downlinks. In view of the above, there is a continuing, ongoing need for improved downlinking techniques.

SUMMARY

In an embodiment, a method includes a processor receiving one or more data records representing previous drilling operations, the one or more data records including data representing one or more downlinks to a downhole tool; the processor determining, for each data record, whether each respective downlink of the one or more downlinks successfully communicated a command to the downhole tool; the processor identifying one or more variables to a machine learning algorithm, at least one of the one or more variables corresponding to a drilling condition; the processor training the machine learning algorithm using at least a first subset of the received data and data indicating downlink success, the machine learning algorithm training by identifying correlations between the one or more variables and the downlink success; and the machine learning algorithm receiving drilling condition data; the machine learning algorithm predicting downlink success probability for each of a plurality of downlink parameter combinations based on the drilling condition data; and the machine learning algorithm recommending one of the plurality of downlink parameter combinations for communication with the downhole tool based on the predicted downlink success probability.

In an embodiment, a system includes a storage device configured to store a machine learning algorithm; and a processor in communication with the storage device and configured to: receive one or more data records representing previous drilling operations, the one or more data records including data representing one or more downlinks to a downhole tool; determine, for each data record, whether each respective downlink of the one or more downlinks successfully communicated a command to the downhole tool; identify one or more variables to a machine learning algorithm, at least one of the one or more variables corresponding to a drilling condition; train the machine learning algorithm using at least a first subset of the received data and data indicating downlink success, the machine learning algorithm training by identifying correlations between the one or more variables and the downlink success; and execute the machine learning algorithm, where the machine learning algorithm is configured to: receive drilling condition data; predict downlink success probability for each of a plurality of downlink parameter combinations based on the drilling condition data; and recommend one of the plurality of downlink parameter combinations for communication with the downhole tool based on the predicted downlink success probability.

In an embodiment, a non-transitory machine-readable medium includes instructions, which, when executed by one or more processors, cause the one or more processors to perform the following operations: receive one or more data records representing previous drilling operations, the one or more data records including data representing one or more downlinks to a downhole tool; determine, for each data record, whether each respective downlink of the one or more downlinks successfully communicated a command to the downhole tool; identify one or more variables to a machine learning algorithm, at least one of the one or more variables corresponding to a drilling condition; train the machine learning algorithm using at least a first subset of the received data and data indicating downlink success, the machine learning algorithm training by identifying correlations between the one or more variables and the downlink success; and receive, by the machine learning algorithm, drilling condition data; predict, by the machine learning algorithm, downlink success probability for each of a plurality of downlink parameter combinations based on the drilling condition data; and recommend, by the machine learning algorithm, one of the plurality of downlink parameter combinations for communication with the downhole tool based on the predicted downlink success probability.

This summary is provided to introduce a selection of concepts that are further described below in the detailed description. This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used as an aid in limiting the scope of the claimed subject matter.

BRIEF DESCRIPTION OF THE DRAWINGS

The subject disclosure is further described in the detailed description which follows, in reference to the noted plurality of drawings by way of non-limiting examples of the subject disclosure, in which like reference numerals represent similar parts throughout the several views of the drawings, and wherein:

FIG. 1 illustrates a data processing system in which an illustrative embodiment of the present invention may be implemented;

FIG. 2 illustrates a plot representing a wellbore drill path according to an embodiment;

FIG. 3 illustrates a computing system according to an embodiment;

FIG. 4 illustrates a flow chart for training and modeling a machine learning algorithm according to an embodiment;

FIG. 5 illustrates correlations between drilling condition variables and downlink success according to an embodiment;

FIG. 6 illustrates downlink success predictions for various downlink types according to an embodiment;

FIG. 7 illustrates a flow chart for deploying a machine learning algorithm during a drilling planning phase according to an embodiment;

FIGS. 8A-8C illustrate displayed outputs of the machine learning algorithms according to an embodiment; and

FIG. 9 illustrates a flow chart for deploying a machine learning algorithm during a drilling phase according to an embodiment.

DETAILED DESCRIPTION

The particulars shown herein are by way of example and for purposes of illustrative discussion of the embodiments of the subject disclosure only and are presented in the cause of providing what is believed to be the most useful and readily understood description of the principles and conceptual aspects of the subject disclosure. In this regard, no attempt is made to show structural details in more detail than is necessary for the fundamental understanding of the subject disclosure, the description taken with the drawings making apparent to those skilled in the art how the several forms of the subject disclosure may be embodied in practice. Furthermore, like reference numbers and designations in the various drawings indicate like elements.

While this invention is susceptible of an embodiment in many different forms, there are shown in the drawings and will be described herein in detail specific embodiments thereof with the understanding that the present disclosure is to be considered as an exemplification of the principles of the invention. It is not intended to limit the invention to the specific illustrated embodiments.

Embodiments disclosed herein can include a downlink automation portfolio configured to recommend downlinking type based on one or more drilling conditions to increase downlinking success rate. The downlink automation system can provide recommendations based on one or more drilling conditions variables, which can include hole section, geographic area, wellbore diameter, mud rheology, mud density, type of bottom hole assembly, rate of penetration, flow amplitude, RPM amplitude, bit period, shock and vibrations, true vertical depth, measured depth, stick/slip vibration, etc. These variables can impact downlink success, and the system can recommend downlink parameters (bit period length, flow vs. RPM) that account for these variables.

The downlink automation system can operate during a planning phase before drilling commences and in real-time during drilling. When used in the planning phase, the system can recommend downlink parameters having the highest chance of success for all sections of the drilling path based on analysis of historical data. When used in real-time, the downlink automation system can receive data from downhole sensors (e.g., magnetometer, gyroscope, accelerometer, etc.) and determine whether the received data from the sensors resembles or corresponds to data generated during the planning phase. If the sensor data deviates significantly from the planning phase data, the downlink automation system may provide updated or new recommendations for downlinking parameters that account for the newly received data. For example, during a planning phase, the downlink automation system may expect mud density of approximately 1100, but the actual mud density may be greater than 1200, resulting in a different drilling operation than expected. As a result, different downlink parameters (e.g., longer bit period, changing from flow rate variation to RPM rate variation) may be necessary to increase a chance of a successful downlink with a downhole tool.

In a typical drilling operation, several well bores are located strategically around a field in order to maximize production of the oil or gas contained within the field. A wellbore is drilled into formation rock that contains oil, natural gas, and water. Maximizing the speed to complete the drilling procedure may be a primary goal. The drilling procedure should follow an intended drilling path so that oil or gas production can be maximized after drilling. Careful control of the downhole tool, controlled via downlinking, can steer the downhole tool through an intended drilling path. If the downhole tool deviates from the intended drilling path, for example, due to unsuccessful downlinks, the result can be non-productive time (“NPT”), which can delay drilling operations. The systems and methods described herein can increase the probability of a successful downlink by accounting for one or more drilling conditions by analyzing data using machine learning or other artificial intelligence algorithms.

Referring to FIG. 1, a pictorial representation of a data processing system is depicted in which an illustrative embodiment of the present invention may be implemented. In this example, data processing system 100 is one or more computing devices in which different embodiments of the present invention may be implemented.

In this depicted example, well site 102 includes or is connected to downlink hardware 103. The downlink hardware 103 can include at least a valve and a pump that control downhole liquid flow into the well site 102. In some embodiments, the valve and pump are controlled manually by a drilling engineer and flow variations may occur based on the engineer's manual actions. In another embodiment, the downlink hardware 103 may include more sophisticated computer-controlled hardware, such as software that controls the valve and pump to generate flow or RPM variations in fluid sent downhole. In yet another embodiment, the downlink hardware 103 can include a downlink modulation kit to control the valve, pump, and an amount or percentage of fluid sent down the well site 102.

The downlink hardware 103 can connect to a computing device 104 that produces data regarding the well site 102. This information may be used in the management of drilling operations at the well site 102. For example, the information may be used to direct drilling operations for the well site. Moreover, the computing device 104 can display data to a drilling engineer or other operator regarding a drilling operation. The operator can interact with the computing device 104 to oversee drilling operations, including generating commands provided to downhole drilling equipment through downlinking. In the embodiment where the downlink hardware 103 is controlled manually, the computer 104 may not electrically connect to the downlink hardware 103, but instead, the computer 104 may display information to the drilling engineer providing guidance and recommendations for the manual control. In these examples, well site 102 is located in a geographic region 112. This geographic region 112 is a single reservoir, as described in these examples. While FIG. 1 illustrates only a single well site 102, the geographic region 112 may include a plurality of well sites at or near the reservoir strategically positioned to maximize oil or gas production.

The computing device 104 may be connected to an analysis center 120 at which data processing systems are located to assist with wellbore drill path planning and real-time analysis of wellbore drilling. Depending on the particular implementation, multiple analysis centers may be present. These analysis centers may be, for example, at an office or on-site in the geographic location 112 depending on the particular implementation. In these illustrative embodiments, analysis center 120 analyzes data from computing device 102 using processes for different embodiments of the present invention. Analysis center 120 can also connect with other computing devices associated with other well sites (not illustrated). Moreover, the analysis center 120 can implement one or more machine learning algorithms, store past drilling data collected during previously completed drilling operations, store present drilling data collected during an ongoing drilling operation, and provide recommendations for downlinking parameters to increase the probability of a successful downlink with downhole drilling tools. In some embodiments, the analysis center 120 is omitted, and the computing device 104 performs the functions described herein as being performed by the analysis center 120.

The computing device 104 may connect to the analysis center 120 via a networked connection. In an example, the networked connection can include the Internet, which includes a worldwide collection of networks and gateways that use the Transmission Control Protocol/Internet Protocol (TCP/IP) suite of protocols to communicate with one another. Of course, the networked connection also may be implemented as a number of different types of networks, such as, for example, an intranet, a local area network (LAN), or a wide area network (WAN). FIG. 1 is intended as an example, and not as an architectural limitation for different embodiments.

FIG. 2 illustrates a plot showing a vertical section length (x-axis) plotted against true vertical depth (TVD) of a wellbore. In other words, FIG. 2 represents a drilling path for a well. As shown in plot 200, different downlink parameters increase the probability to successfully downlink new commands at different sections of the well. For example, in section 202, a first bit-period and a first downlink type may have the highest probability to succeed. Continuing this example, in section 204, a second bit-period and the first downlink type may have the highest probability to succeed; in section 206, a third bit-period and the first downlink type or the second bit-period and a second downlink type may have the highest probability to succeed, and in section 208, the third bit-period and the second downlink type or the first downlink type and an off-bottom downlink (i.e., a calmer drilling consideration by stopping drilling process, pick up off bottom and reduced (or stop) rotary speed of drill string (RPM) and then send a downlink to tool (RPM or flow rate variation); once a downlink is identified and been accepted by downhole tool, tag bottom and drilling would continue) may have the highest probability to succeed. As an example, the downlink automation system may recommend that, in section 202, in approximately the first 700 m of the wellbore's vertical section, having a TVD between Om and approximately 1300 m, an 18 second bit-period, flow rate variation downlink may have the highest probability to succeed. Continuing this example, in section 204, from approximately 700 m to approximately 2100 m of the wellbore's vertical section, having a TVD between approximately 1200 m and approximately 1500 m, a 36 second bit-period, flow rate variation downlink may have the highest probability to succeed; in section 206, from approximately 2100 m to approximately 3500 m of the wellbore's vertical section, having a TVD of approximately 1500 m, a 60 second bit-period, flow rate variation or a 36 second, RPM variation downlink may have the highest probability to succeed, and in section 208, from approximately 3500 m and beyond of the wellbore's vertical section, having a TVD of approximately 1500 m, a 60 second RPM or an off-bottom downlink (i.e., a calmer drilling consideration by stopping drilling process, pick up off bottom and reduced (or stop) rotary speed of drill string (RPM) and then send a downlink to tool (RPM or flow rate variation); once a downlink is identified and been accepted by downhole tool, tag bottom and drilling would continue). FIG. 2 is exemplary and may not account for additional drilling parameters. However, FIG. 2 illustrates that different sections of the wellbore require different types of downlinks to provide the best probability of successfully downlinking new commands. FIG. 2 also illustrates that the different downlink parameters can include different bit-periods (e.g., 9 s, 18 s, 36 s, 60 s) and either flow rate or rotation rate (RPM) variations to generate the modulated signal.

Turning now to FIG. 3, a diagram of a computing system is depicted in accordance with an illustrative embodiment. In this illustrative example, the computing system 300 includes a communications bus 302, which provides communications between a processor unit 304, memory 306, persistent storage 308, a communications unit 310, an input/output (I/O) unit 312, and a display 314. The computing system 300 of FIG. 3 can depict the components of either the computing system 104 or the analysis center 120 in FIG. 1.

The processor unit 304 serves to execute instructions for software that may be loaded into memory 306. The processor unit 304 may comprise one or more processors or may be a multi-processor core, depending on the particular implementation. Further, the processor unit 304 may be implemented using one or more heterogeneous processor systems in which a main processor is present with secondary processors on a single chip. In another illustrative example, the processor unit 304 may be a symmetric multi-processor system containing multiple processors of the same type.

The memory 306, in these examples, may be, for example, a random access memory or any other suitable volatile or non-volatile storage device. The persistent storage 308 may take various forms depending on the particular implementation. For example, the persistent storage 308 may contain one or more components or devices. For example, the persistent storage 308 may be a hard drive, a flash memory, a rewritable optical disk, a rewritable magnetic tape, or some combination of the above. The media used by the persistent storage 308 also may be removable. For example, a removable hard drive may be used for the persistent storage 308. In another embodiment, the persistent storage 308 may comprise a database.

The communications unit 310, in these examples, provides for communications with other data processing systems or devices. In these examples, the communications unit 310 is a network interface card. The communications unit 310 may provide communications through either or both physical and wireless communications links.

The input/output unit 312 allows for input and output of data with other devices that may be connected to the computing system 300. For example, the input/output unit 312 may provide a connection for user input through a keyboard and mouse. Further, the input/output unit 312 may send output to a printer. The display 314 provides a mechanism to display information to a user. In one example, the display 314 may display downlink recommendations generated by the processing unit 304.

Instructions for the operating system and applications or programs are located on the persistent storage 308. These instructions may be loaded into the memory 306 for execution by the processor unit 304. The processes of the different embodiments may be performed by the processor unit 304 using computer implemented instructions, which may be located in a memory, such as the memory 306. These instructions are referred to as, program code, computer usable program code, or computer readable program code that may be read and executed by a processor in the processor unit 304. The program code in the different embodiments may be embodied on different physical or tangible computer readable media, such as the memory 306 or the persistent storage 308.

The program code 316 is located in a functional form on computer readable media 318 and may be loaded onto or transferred to the computing system 300 for execution by the processor unit 304. The program code 316 and the computer readable media 318 form computer program product 320 in these examples. In one example, the computer readable media 318 may be in a tangible form, such as, for example, an optical or magnetic disc that is inserted or placed into a drive or other device that is part of the persistent storage 308 for transfer onto a storage device, such as a hard drive that is part of the persistent storage 308. In a tangible form, the computer readable media 318 also may take the form of a persistent storage, such as a hard drive or a flash memory that is connected to the computing system 300. The tangible form of the computer readable media 318 is also referred to as computer recordable storage media.

Alternatively, the program code 316 may be transferred to the computing processing system 300 from the computer readable media 318 through a communications link to the communications unit 310 and/or through a connection to the input/output unit 312. The communications link and/or the connection may be physical or wireless in the illustrative examples. The computer readable media also may take the form of non-tangible media, such as communications links or wireless transmissions containing the program code. In one particular embodiment, the program code 316 can comprise one or more machine learning algorithms or other artificial intelligence algorithms configured to train using data stored on the persistent storage 308, such as past drilling information, which can include previous downlink data. Once trained, the trained machine learning algorithms can be stored in the memory 306 or the persistent storage 308, and the processor unit 304 can call the trained machine learning algorithms to provide downlink recommendations for a planned drilling run or real-time recommendations for a presently ongoing drilling operation.

The different components illustrated for the data processing system 300 are not meant to provide architectural limitations to the manner in which different embodiments may be implemented. The different illustrative embodiments may be implemented in a data processing system including components in addition to or in place of those illustrated for the data processing system 300. Other components shown in FIG. 3 can be varied from the illustrative examples shown.

For example, a bus system may be used to implement communications bus 302 and may be comprised of one or more buses, such as a system bus or an input/output bus. Of course, the bus system may be implemented using any suitable type of architecture that provides for a transfer of data between different components or devices attached to the bus system. Additionally, a communications unit may include one or more devices used to transmit and receive data, such as a modem or a network adapter. Further, a memory may be, for example, memory 306 or a cache, such as found in an interface and memory controller hub that may be present in communications bus 302.

The systems and methods described herein can implement machine learning or artificial intelligence to recommend downlink parameters (bit period, flow vs. RPM) at numerous points along a drilling operation based on one or more drilling condition variables.

Recommendations for downlink parameters (i.e., downlink type) can be generated by a machine learning algorithm after training the machine learning algorithm using a training dataset. In a preferred embodiment, the training dataset can include historical drilling data, which can include downlink data implemented during previously conducted drilling operations. The variables considered by the machine learning algorithm can include hole section, geographic area, wellbore diameter, mud rheology, mud density, type of bottom hole assembly, rate of penetration, flow amplitude, RPM amplitude, bit period, shock and vibrations, true vertical depth, measured depth, stick/slip vibration, etc.

FIG. 4 illustrates a method of training a machine learning algorithm to generate downlink parameter recommendations for drilling operations according to an embodiment. As shown in FIG. 4, a processor (e.g., processing unit 304 of FIG. 3) can extract and ingest past downlinking data stored in the persistent storage 308 in step 402. The past downlinking data can include a plurality of downlinks stored as data included in data generated from a plurality of drilling runs. In other words, the processor can receive one or more data records representing previous drilling operations, the one or more data records including data representing one or more downlinks. Preferably, the past downlinking data can correspond to drilling runs from a plurality of geographical locations, representing a plurality of wellbore conditions (e.g. mud density, mud rheology, etc.). In some embodiments, the past downlinking data can include vertical section length and TVD data at a plurality of points along a drill path. Alternatively, in some embodiments, some variables can be calculated based on data stored as past downlinking data. For example, rate of penetration (“ROP”) data may not be stored as data in the persistent storage 308, but ROP can be calculated or estimated using data from drilling runs and downlink data included therein. In addition, fantom downlinks can be removed from consideration before ingesting the data into the machine learning algorithm.

The data ingested by the machine learning algorithm can include drilling data and downlink data. The data stored can include a location where the drilling occurred (e.g., country where drilling occurred, a field where drilling occurred, or a more general geographic location (e.g., offshore Atlantic Ocean, North America, etc.). The data stored can further include downhole tool type (e.g. RSS, logging while drilling, measurement while drilling), wellbore diameter, drill model name, and whether the drilling was on-bottom or off-bottom drilling. The data can further include drilling condition data, such as mud density and mud type, or this data can be stored elsewhere and may be correlated with the geographic location data. The downlink data can further include bit-period data, RPM or flow rate amplitude, and whether the downlink used flow rate variation or RPM variation to send the command.

After the data is ingested, the processor can analyze and preprocess the ingested data in step 404. Preprocessing data, as would be understood by those of ordinary skill in the art, can be an important step in training a machine learning algorithm. Preprocessing can include, for example, handling null values, standardizing the data, handling categorical variables, multicollinearity, or running exploratory data analysis. In addition, preprocessing can include removing a predetermined number of records from the ingested data, which can be subsequently used as unseen data for validation purposes.

In some embodiments, the data indicates whether the downlink was successful or unsuccessful. In an embodiment where the data is not indicating whether the downlink was successful or unsuccessful, the processor can calculate whether each downlink record was accepted or was unsuccessful in step 406. The processor can determine whether each respective downlink was accepted and successful by reviewing successive downlinks to determine whether the downlinks indicate corrective actions were necessary due to an unsuccessful downlink. Alternatively, the processor can determine whether each respective downlink was accepted and successful by comparing downlink signal parameters to a set of acceptance thresholds.

After calculating the acceptance rate of reach downlink, the processor can perform another preprocessing step on the data in step 408. The additional preprocessing step can include appending the data to include a new column in the downlink data indicating whether each downlink was successful or unsuccessful. Other preprocessing steps can occur here, including some or all of the preprocessing steps described above with reference to step 404. Step 408 can also include feature engineering, which can include selecting relevant features or variables for a machine learning algorithm to consider while finding associations between variables and downlink success. Step 408 can also include encoding of categorical values and scaling numerical features and other preprocessing steps performed before training a machine learning algorithm. A machine learning algorithm may also be selected at this step or in previous steps. The machine learning algorithm may be selected from the group including logical regression, random forest, or gradient boosting, but the systems and methods described herein are not limited to the above-listed machine learning algorithms. When selecting a machine learning algorithm, a designer can consider model interpretability and explainability. In some embodiments, each of the machine learning algorithms may train and test data, and a user may select which algorithm best predicts downlink accuracy by considering accuracy, precision, and recall of each algorithm.

Further still, the processor can calculate or estimate additional values or data not stored as drilling or downlink data in steps 410-414. For example, in step 410, the processor can estimate an ROP for each downlink record. The processor can calculate ROP for each downlink record by calculating the time duration of each power-up session. Assuming 90 feet drilling in each power-up session, the ROP can be calculated by dividing 90 feet by the duration of each power-up session.

Once the processor determines the ROP for each downlink record, the processor can use the determined ROP to calculate or estimate other values, including TVD and measured depth (MD), in step 412. The processor can estimate MD by multiplying ROP by the time duration of each powerup session, and the processor can estimate the TVD by multiplying MD for each powerup session and a cosine value of an inclination angle stored in the data records. In some embodiments, the TVD and MD values are stored as data in the persistent storage 308 as part of a run summary data record.

Further still, the processor can calculate stick/slip values in step 414. The processor can calculate stick/slip data based on an estimation. The stick/slip value can be computed by calculating the difference between a maximum turbine RPM value and a minimum turbine RPM value. The difference can be divided by the mean turbine RPM value to generate a quotient. The quotient can be multiplied by 100 to generate an estimated stick/slip value. In some embodiments, the processor can omit one of or all of steps 410-414, however, omitting ROP, TVD, MD, and stick/slip from the list of variables considered by the machine learning algorithm may negatively impact the accuracy of the machine learning algorithm's predictions and recommendations.

Subsequently, in step 416, the selected machine learning algorithm can train using the ingested data and any other data calculated in steps 406 and 410-414. In some embodiments, the processor can divide the data records into a training subset and a test subset. The training subset can comprise 70% of the records in the past downlinking data, and the test subset can comprise 30% of the records in the past downlinking data. However, other allocations of the past downlinking data into training and testing subsets are contemplated. The machine learning algorithm can train itself using the training subset. While training, the machine learning algorithm can analyze a plurality of features/variables to determine a correlation with downlink success, and the machine learning algorithm can assign weights to the various features/variables analyzed. Training the machine learning algorithm may also include tuning hyperparameters using grid search or random search.

During training, the machine learning algorithm can find a correlation between a drilling condition and downlink success. FIG. 5 illustrates a results table 500 of one machine learning algorithm's determination of the correlation for each drilling condition. As can be seen in the example of FIG. 5, downlink success can correlate strongly with MD, TVD, and bit-period. While FIG. 5 illustrates ten variables that correlate to some degree with computed downlink acceptance, other variables can impact downhole acceptance other than those illustrated in FIG. 5.

Referring again to FIG. 4, after training, the machine learning algorithm can predict success of a downlink under a given set of parameters using the test subset in step 418. The machine learning algorithm can predict downlink success rate for a plurality of bit periods and for each of flow rate variation and RPM variation.

FIG. 6 illustrates prediction results 600 for each record in the test data set, a set that includes, for example, 1515 records. As shown, the machine learning algorithm predicts the probability of a successful downlink for each record using one of three bit-periods (18 s, 36 s, and 60 s) for flow rate variation and one of three bit periods (18 s, 36 s, and 60 s) for RPM rate variation, resulting in, for example, six combinations of downlink parameters (i.e., six downlink types). As a result of this calculation, the machine learning algorithm can recommend one of the six downlink types having the highest probability of success. For example, for the first entry in the table shown in FIG. 6, the downlink type having the highest probability of success is an 18 s bit-rate, flow rate variation downlink. This prediction can be validated by the test data to determine the machine learning algorithm's accuracy. The first entry in the table can correspond to a specific TVD and MD, as well as several other factors which are illustrated in FIG. 6. After trained and tested, the machine learning algorithm may be implemented during a planning phase of a drilling run.

FIG. 7 illustrates a method for using a trained machine learning algorithm to recommend downlink parameters for all phases of a drilling run. In step 702, the processor can receive a planned drill path, the planned drill path can include the TVD and MD at multiple points along the planned drill path. In addition, the processor may receive other drilling condition variables, such as mud density, wellbore diameter, etc., in step 704. The processor can calculate downlink prediction success rate for each of a plurality of downlink parameter combinations (see FIG. 6) at each of the multiple points in step 706, and recommend a downlink type based on the combination of downlink parameters having the highest predicted success rate at each point in 708. The method 700 can continue for each point along the drill path.

The method 700 can display a recommendation in the form of a plot showing recommended downlink type for multiple points along a drill path. Exemplary recommendation plots are shown in FIGS. 8A-8C, which illustrate the recommendations as graphs. In some embodiments, the method 700 can further include sending control commands to downlink hardware to implement the downlink types and bit periods generated by the method 700, such as by sending commands to software controlling the valves and/or pumps of the downlink hardware 103.

For example, referring to FIG. 8A, during a first segment 802 of a drill path, the processor may recommend a 36 s bit-rate, flow variation downlink type; during a second segment 804 of a drill path, the processor may recommend a 18 s bit period, flow variation downlink type; during a third segment 806 of a drill path, the processor may recommend a 36 s bit period, flow variation downlink type; and during a fourth segment 808 of a drill path, the processor may recommend a 60 s bit period, flow variation downlink type. The graph in FIG. 8A can assume ROP, mud density, drilling location, mud type, and stick/slip.

For example, referring to FIG. 8B, during a first segment 810 of a drill path, the processor may recommend a 36 s bit-rate, flow variation downlink type; during a second segment 812 of a drill path, the processor may recommend a 18 s bit period, flow variation downlink type; and during a third segment 814 of a drill path, the processor may recommend a 36 s bit period, RPM variation downlink type. The graph in FIG. 8B can assume ROP, mud density, drilling location, mud type, and stick/slip.

For example, referring to FIG. 8C, during a first segment 820 of a drill path, the processor may recommend a 36 s bit-rate, flow variation downlink type; during a second segment 822 of a drill path, the processor may recommend a 18 s bit period, flow variation downlink type; and during a third segment 824 of a drill path, the processor may recommend a 36 s bit period, flow variation downlink type. The graph in FIG. 8C can assume ROP, mud density, drilling location, mud type, and stick/slip.

The trained machine learning algorithm can provide recommendations during the planning phase, as shown in FIGS. 7 and 8A-8C, However, the trained machine learning algorithm may also assist during real-time drilling and provide downlinking recommendations for downlinking in real-time as a downhole tool drills the wellbore.

FIG. 9 illustrates a method 900 for using a trained machine learning algorithm to assist with real-time drilling. As shown in FIG. 9, the method 900 can include the processor receiving data from sensors included in the downhole tool in step 902. The sensors can include a magnetometer, an accelerometer, a gyroscope, a flowmeter, and other sensors that indicate a trajectory or path of the downhole tool. The processor can also receive other data input, such as real-time survey data or other data about drilling conditions, such as observed conditions from an operator.

Subsequently, the processor can compare the data received from the downhole tool sensors or other inputs to expected data generated during the planning phase in step 904. For example, a drill may have deviated from a drill path, resulting in changes to the drill path in terms of MD and/or TVD. As a result, the processor may determine that a different downlink type, different from a downlink type recommended during the planning phase, has a better probability of resulting in a successful downlink. If a higher probability downlink type exists for current drilling conditions, the processor can recommend a new downlink type, different from a downlink type generated during the planning phase, and display that recommendation to a drill operator in step 906. The operator can accept the processors recommendation and send downlinks according to the recommended downlink parameters. In some embodiments, the method 900 can further include sending control commands to downlink hardware to implement the downlink types and bit periods generated by the method 900, such as by sending commands to software controlling the valves and/or pumps of the downlink hardware 103.

The downlink automation system solves the problems over the prior art because the recommended downlink parameters provided by the downlink automation system increase the success rate of all downlinks performed during a drilling run. The downlink recommendations provided by the system dramatically increase downlink success and acceptance rate using a significant amount of past data. The past data is analyzed by machine learning tools to find the highest accuracy of downlink parameters at various depths and accounting for numerous other variables. As a result of more successful downlink acceptance, non-productive time is minimized, drill path accuracy is increased, and rate of penetration is increased.

Although only a few embodiments have been described in detail above, other modifications are possible. For example, the steps described above do not require the particular order described or sequential order to achieve desirable results. Other steps may be provided, steps may be eliminated from the described flows, and other components may be added to or removed from the described systems. Other embodiments may be within the scope of the invention.

From the foregoing, it will be observed that numerous variations and modifications may be effected without departing from the spirit and scope of the invention. It is to be understood that no limitation with respect to the specific system or method described herein is intended or should be inferred. It is, of course, intended to cover all such modifications as fall within the spirit and scope of the invention.