NEURON NETWORK MODELS FOR REAL-TIME OPTIMIZATION OF DRILLING PARAMETERS DURING DRILLING OPERATIONS

专利摘要:
The present invention relates to a system and methods for optimizing the parameters of drilling operations. Real-time data including values for input variables associated with a stage of the moment of a drilling operation along a planned well path are acquired. A neural network model is driven to produce an objective function defining a response value for at least one operating variable of the drilling operation. The response value for the operating variable is estimated based on the objective function produced by the trained neural network model. Stochastic optimization is applied to the estimated response value to produce an optimized response value for the operation variable. The controllable parameter values are estimated for a later stage of the drilling operation, based on the optimized response value of the operating variable. The subsequent stage of the drilling operation is implemented on the basis of the estimated values of the controllable parameters.
公开号:FR3070180A1
申请号:FR1856731
申请日:2018-07-19
公开日:2019-02-22
发明作者:Srinath Madasu；Keshava Prasad Rangarajan；Nishant Raizada
申请人:Landmark Graphics Corp；
IPC主号:

专利说明:

2017-IPM-101331-U1-FR 1
NEURON NETWORK MODELS FOR REAL-TIME OPTIMIZATION OF DRILLING PARAMETERS DURING DRILLING OPERATIONS
AREA OF DISCLOSURE
The present disclosure relates, in general, to the planning and control of a well during drilling operations and, more particularly, to a modeling and optimization in real time of the drilling parameters for the planning and control of a well during drilling operations.
CONTEXT
To obtain hydrocarbons, such as oil and gas, a borehole is drilled in a rock formation containing hydrocarbons by the rotation of a drill bit attached to a drill string. The drill bit is mounted on the lower end of the drill string as part of a downhole module (BHA) and is rotated by the rotation of the drill string at the surface, by actuation a downhole motor, or both. With a weight applied by the drill string, the rotating drill bit comes into contact with the formation and forms a borehole towards a target area. During the drilling process, drilling fluids are circulated to remove the cuttings while the drill bit enters the formation.
A number of sensors or measuring devices can be placed near the drill bit to measure downhole operating parameters associated with drilling and downhole conditions. The measurements captured by these sensors can be transmitted to a calculation device of a drilling operator on the surface of the borehole for the purposes of monitoring and controlling the drilling of the wellbore along a route planned for different stages of a drilling operation. When decisions need to be made to plan and effectively implement a drilling plan, the drilling operator may need to constantly monitor and adjust various parameters to account for changes in downhole conditions as and when as the wellbore is drilled through different layers of the formation. However, this can be difficult due to the complexity of the underlying physics and engineering aspects of the drilling process in addition to the inherent uncertainty of the data captured at the surface and downhole.
BRIEF DESCRIPTION OF THE FIGURES
2017-IPM-101331-U1-FR 2
Figure 1 shows a diagram of an offshore drilling system according to one or more embodiments of this disclosure.
Figure 2 shows a diagram of a shore drilling system according to one or more embodiments of this disclosure.
Figure 3 is a block diagram of a real-time analysis and optimization system for downhole parameters for planning and monitoring a well during a drilling operation.
Figure 4 is a diagram of an illustrative neural network model for optimizing the parameters of a drilling operation along a planned well path based on non-linear constraints applied to the models during different stages of the process. surgery.
Figure 5 is a diagram of a neural network model with real-time data inputs and Bayesian optimization to train and re-train the model.
Figure 6 is a diagram of a sliding window neural network (SWNN) for predicting the values of one or more operating variables of a drilling operation along a well path.
FIG. 7 is a diagram of a recurrent deep neural network (DNN) having one or more recurrent gate unit cells (GRU) for predicting the values of one or more operating variables of a drilling operation along a well path.
FIG. 8 is a diagram of an illustrative GRU cell of the DNN presented in FIG. 7.
Figure 9 is a diagram of an illustrative architecture of recurrent DNN for filtering noise from real-time data used to drive recurrent DNN.
FIG. 10 is a diagram of an illustrative method for optimizing downhole parameters with a real-time neural network model for planning and controlling a well during the various stages of a drilling operation .
Figures 11 A, 1 IB, 1 IC and 11D are graphs showing the penetration rate (ROP) values as predicted using a SWNN model versus actual ROP values along a well path.
Figures 12A and 12B are graphs showing a comparison between predicted and actual ROP values along a well path, where the predicted values are based on a recurring DNN with or without a noise filter, respectively.
Figure 13 is a block diagram of an illustrative computer system in which one or more embodiments can be implemented.
2017-IPM-101331-U1-FR 3
DESCRIPTION OF ILLUSTRATIVE EMBODIMENTS
The embodiments of this disclosure relate to the use of neural network models for real-time optimization of downhole parameters for drilling operations. Although the present disclosure is described here with reference to illustrative embodiments for particular applications, it is understood that the embodiments are not limited to these. Other embodiments are possible, and modifications may be made to the embodiments in the spirit and scope of these teachings and additional areas in which the embodiments would be of significant benefit.
In the present detailed description, references to "one of the embodiments", "an embodiment", "an exemplary embodiment", etc., indicate that the described embodiment may include a specificity, a structure , or a specific specificity, but all the embodiments do not necessarily include the specificity, the structure, or the particular characteristic. In addition, such expressions do not necessarily refer to the same embodiment. In addition, when a specificity, a structure, or a particular characteristic is described in connection with an embodiment, it is estimated that the person skilled in the art has the knowledge to implement such a specificity, structure, or characteristic in connection with other embodiments whether explicitly described or not.
It will also seem obvious to those skilled in the art that the embodiments, as described here, can be implemented in many different embodiments of software, hardware, firmware, and / or the entities illustrated in the FIGS. Any actual software code with specialized control of hardware to implement embodiments does not limit the detailed description. Therefore, the functional behavior of embodiments will be described with the understanding that modifications and variations of the embodiments are possible, given the level of detail presented here.
The terms "controllable parameter" and "input variable" can be used interchangeably in this document to refer to an input or a controllable parameter of a drilling operation that can be adjusted during the operation. and whose values can have an impact on the result of the operation. The drilling operation may involve drilling a wellbore along a path or planned path through different layers of an underground formation. Downhole operating conditions may change as the wellbore is drilled through the formation. Therefore, a drilling operator or an automated control system can continuously adjust
2017-IPM-101331-U1-EN 4 one or more controllable parameters in order to take these changes into account and thereby maintain or improve the efficiency of the drilling during the operation. Examples of such controllable parameters include, but are not limited to, the load on the drill bit (WOB), the rotational speed of the drill bit or the drill string (e.g., the rotational speed applied by the drilling mechanism). 'upper drive) in revolutions per minute (Rpm), and a speed of injection or pumping (Q) of a drilling fluid into the wellbore or a pipe arranged inside it. Although "RPM" is used in this document to refer to the rotation or rotational speed of the drill bit, it will be understood that such speed can be specified using any suitable unit of measurement and desired for a particular implementation.
In one or more embodiments, the controllable parameters can be used to control the values of an "operating variable" of the drilling operation when it is carried out at the bottom of the well during different stages along a planned path. from the wellbore through the formation. The operating variable can be selected by a user (e.g., a drilling operator) to monitor a particular downhole response when the drilling operation is performed along the wellpath, for example, based on instant values of controllable parameters or input variables. Therefore, the operating variable can also refer in this document to a "response variable" of the drilling operation. Examples of such operating / response variables include, but are not limited to, specific mechanical hydraulic energy (HMSE) and penetration rate (ROP). The controllable parameters (input variables) and the operating / response variables are collectively referred to in this document as "drilling parameters".
In one or more embodiments, a neural network model with stochastic optimization based on Bayesian optimization (BO) can be used to optimize one or more operating variables (or response variables) for each stage of the operation. drilling, for example, maximizing ROP and / or minimizing HMSE at different depths or at different points along the drilling path. As described in more detail below, real-time data, including the instantaneous values of one or more controllable parameters (for example, WOB, RPM, and / or Q) at different depths along the well path, can be applied as inputs to the neural network model to predict the values of the response variable or variables. The neural network model can be, for example, a sliding window neural network (SWNN). Alternatively, the neural network model may be a network of
2017-IPM-101331-U1-EN 5 deep recurrent neurons (DNN) with one or more recurrent unit-gate cells (GRU). However, it will be understood that any of the various neural network models, for example, short and long term memory (LSTM) deep neural network models, can also be used if desired for implementation. particular work.
Illustrative embodiments and related methodologies of this disclosure are described below with reference to Figures 1 to 15 as they could be employed, for example, in a computer system for real-time modeling and optimization of parameters drilling during different stages of a drilling operation along a planned well path. In some implementations, such a computer system may be part of an automated control system for orienting a drill bit along the well path by mathematically coupling the neural network model with the real-time drilling data and various non-linear discontinuous stresses associated with different stages of the drilling operation at various depths or points along the well path. For example, the drill bit can be steered iteratively as real-time data is acquired over a period of time during each stage of the drilling operation. At each iteration during the time period, the real-time data acquired for a stage of the time of the operation can be applied as inputs to train or re-train the neural network model in order to estimate or predict the response variable for a later stage along the well path. Bayesian optimization can be applied to optimize the predicted response variable. The predicted response variable can then be used to estimate or predict optimal values for one or more controllable parameters, and the later stage of the drilling operation can be achieved by directing the drill bit through the formation based on the values of controllable parameters estimated. In this way, the system can iteratively steer the drill bit and adjust the well path if necessary to optimize drilling efficiency, for example, by maximizing ROP and / or minimizing HMSE.
Other features and advantages of the disclosed embodiments will be or will become apparent to those skilled in the art after examining the following figures and the detailed description. It is intended that all of the additional specifics and advantages of this type are included within the scope of the disclosed embodiments. Furthermore, the illustrated figures are given by way of example only and are not intended to assert or imply any limitation relating to the environment, architecture, design or process in which the different embodiments can be implemented. Although the examples illustrated can be described in the context of predicting and optimizing the ROP and / or
2017-IPM-101331-U1-EN 6 of the HMSE, it should be noted that the embodiments are not intended to be limited to these and that the techniques for optimizing the disclosed parameters can be applied to any variable operating mode desired for a particular implementation. In addition, even if a figure may represent a horizontal wellbore or a vertical wellbore, unless otherwise indicated, those skilled in the art should understand that the apparatus according to the present disclosure is also well suited for use in wells drilling with other orientations, such as vertical drilling wells, inclined drilling wells, multilateral drilling wells or the like. In addition, unless otherwise indicated, even if a figure may represent a cased hole, those skilled in the art should understand that the apparatus according to the present disclosure is also well suited for use in open hole operations.
Figure 1 illustrates a diagram showing an example of an offshore drilling system for an underwater drilling operation. In particular, Figure 1 shows a downhole assembly 100 for an underwater drilling operation, the downhole assembly 100 illustratively comprising a drill bit 102 at the distal end of the drilling train 104 Various logging during drilling (LWD) or measurement during drilling (MWD) tools can also be coupled inside the well bottom assembly 100. The distinction between LWD and MWD is sometimes blurred in industry, but within the scope of this specification and of the claims, the LWD tools measure properties of the surrounding formation (for example, resistivity, porosity, permeability), and the MWD tools measure properties associated with the well of drilling (for example, a tilt, and a direction). In the example system, a logging tool 106 can be coupled just above the drill bit, where the logging tool can read data associated with wellbore 108 (for example, an MWD tool), or the logging tool 106 can read data associated with the surrounding formation (for example, an LWD tool). In some cases, the downhole assembly 100 may include a mud motor 112. The mud motor 112 can divert energy from the drilling fluid flowing inside the drill string 104 and, from extreme energy, the mud motor 112 can rotate the drill bit 102 (and if present the logging tool 106) separately from the rotation imparted to the drill string by surface equipment . Additional logging tools may be located above the mud engine 112 in the drill string, such as the illustrative logging tool 114.
The well bottom assembly 100 is lowered from a drilling platform 116 using the drilling train 104. The drilling train 104 extends through an extension tube 118 and a well head 120. The drilling equipment supported inside and around the derrick 123 (illustrative drilling equipment described in more detail with respect to the
2017-IPM-101331-U1-EN 7 Figure 2) can rotate the drill string 104, and the rotational movement of the drill string 104 and / or the rotational movement created by the mud motor 112 bring the drill bit 102 forming the wellbore 108 through the formation material 122. The volume defined between the drill string 104 and the wellbore 108 is called annular space 125. The wellbore 108 penetrates into underground zones or reservoirs, like tank 110, capable of containing hydrocarbons in commercially viable quantities.
The well bottom assembly 100 may further comprise a communication subsystem comprising, for example, a telemetry module 124. The telemetry module 124 can couple the various logging tools 106 and 114 in communication and receive the data logs measured and / or recorded by logging tools 106 and 114. Telemetry module 124 can communicate logging data to the surface using any suitable communication channel (for example, pressure pulses at (inside the drilling fluid flowing in the drilling train 104, acoustic telemetry through the tubes of the drilling train 104, electromagnetic telemetry, optical fibers integrated in the drilling train 104, or combinations). Likewise, the telemetry module 124 can receive information from the surface on one or more of the communication channels.
FIG. 2 illustrates a diagram showing an example of a shore drilling system for implementing a shore drilling operation. In particular, FIG. 2 shows a drilling platform 200 equipped with a derrick 202 which supports a lifting device 204. The lifting device 204 suspends an upper drive mechanism 208, which rotates and lowers the undercarriage drilling 104 through the wellhead 210. A drilling fluid is pumped by a mud pump 214 through a flow line 216, a riser 218, a gooseneck 220, an upper drive 208, and to the low through the drill string 104 at high pressures and volumes to flow through nozzles or nozzles into the drill bit 102. The drilling fluid then again travels up the borehole through space annular 125, through a shutter block (not shown specifically), and in a mud basin 224 on the surface. On the surface of the well site, the drilling fluid is cleaned and then recirculated by a mud pump 214. The drilling fluid is used to cool the drill bit 102, to carry cuttings from the base of the borehole towards the surface, and to balance the hydrostatic pressure in rock formations.
In the illustrative case of telemetry mode 124 encoding the data into pressure pulses which propagate towards the surface, one or more transducers, such as one or more of the transducers 232, 234 and 236, convert the pressure signal into signals
2017-IPM-101331-U1-EN 8 electrical for a 238 signal digitizer (for example, an analog-to-digital converter). Although only transducers 232, 234 and 236 are illustrated, any number of transducers can be used as desired for a particular implementation. Digitizer 238 provides a digital form of pressure signals to a surface computer system 240 or another form of a data processing device located on the surface of the well site. The surface computer system 240 operates according to computer executable instructions (which may be stored on a computer readable storage medium) to monitor and control the drilling operation, as will be described in more detail below. Such instructions can be used, for example, to configure the surface computer system 240 to process and decode downhole signals received from telemetry mode 124 via digitizer 238.
In one or more embodiments, the time data collected at the well site, in particular the well bottom logging data coming from the telemetry module 124, can be displayed on a display device 241 coupled to the computer system 240. The representation of well site data can be displayed using a variety of display techniques, as will be described in more detail below. In some implementations, the surface computer system 240 may generate a two-dimensional (2D) or three-dimensional (3D) graphical representation of the well site data to display a graph on the display device 241. The graphical representation of well site data can be displayed along with a representation of the planned well path to allow a user of the computer system 240 to visually monitor or follow different stages of the drilling operation along the planned well path.
In one or more embodiments, the representations of the well site data and the planned well path can be displayed inside a graphical user interface (GUI) of a geoguiding or well engineering application. 280 executable at the surface computer system 240. A well engineering application 280 can provide, for example, a set of data analysis and visualization tools for planning and controlling a well. Such tools can allow the user to monitor different stages of the drilling operation and to adjust the planned well path if necessary, for example, by manually adjusting one or more parameters controllable via the GUI of the application d well engineering 280 to control the direction and / or orientation of the drill bit 102 and the well path. Alternatively, monitoring and control of the drilling operation can be performed automatically, without any user intervention, by the well engineering application 280.
2017-IPM-101331-U1-FR 9
For example, as each stage of the drilling operation is completed and a corresponding portion of the well is drilled along its planned path, the well engineering application 280 may receive indications of downhole operating conditions and controllable parameter values used to control drilling during the operation. Examples of such controllable parameters include, but are not limited to, WOB, flow rate and pressure of the drilling fluid (inside the drill pipe), the speed of rotation of the drill string and / or the drill bit (for example, the speed of rotation applied by the upper drive mechanism and / or a downhole motor), and the density and viscosity of the drilling fluid. In response to receiving downhole operation indications during a stage of the instant of the drilling operation, the surface computer system 240 may automatically send control signals to one or more downhole devices. (for example, a well bottom geoguiding tool) in order to adjust the planned path of the well for the later stages of the operation. The control signals may include, for example, optimized values of one or more controllable parameters to perform later stages of the drilling operation along the adjusted path of the well.
In one or more embodiments, some or all of the calculations and functions associated with manual or automatic monitoring and control of the drilling operation at the well site can be performed by a remote computer system 242 located away from the well site, for example, in an operations center of an oilfield service provider. In some implementations, the functions performed by the remote computer system 242 may be based on well site data received from the well site computer system 240 via a communication network. Such a network can be, for example, a local network, a medium network or a wide area network, for example, the Internet. As illustrated in the example of FIG. 2, the communication between the computer system 240 and the computer system 242 can pass through a satellite link 244. However, it should be understood that the embodiments are not limited to this one. and that any suitable form of communication can be used if desired for a particular implementation.
Although not shown in FIG. 2, the remote computer system 242 can run an application similar to the well engineering application 280 of the system 240 to implement all or part of the monitoring functionality and well site control described above. For example, such functionality can be implemented by using only the well engineering application 280 executable at the system level 240 or by using only the well engineering application executable at the level
2017-IPM-101331-U1-EN 10 of the remote computer system 242 or using a combination of the well engineering applications executable at the respective computer systems 240 and 242 so that all or part of the monitoring functionality and well site control can be propagated among the available computer systems.
In one or more embodiments, the well site monitoring and control functionality provided by the computer system 242 (and the computer system 240 or well engineering application 280 thereof) may include analysis and real-time optimization of parameters for different stages of the drilling operation along the planned well path, as described above and as will be described in more detail below with reference to Figures 3 to 15 Although the examples of Figures 1 and 2 are described in the context of a single well and well site, it should be understood that the embodiments are not intended to be by this and that the techniques of analysis and optimization methods disclosed in this document can be applied to multiple wells at various sites throughout a hydrocarbon production field. For example, the remote computer system 242 of Figure 2, as described above, can be coupled in communication via a communication network to corresponding well-site computer systems similar to the computer system 240 of Figure 2, as described above. The remote computer system 242 in the present example can be used to continuously monitor and control drilling operations at the various well sites by sending and receiving control signals and well site data to and from the computer systems. respective well site via the network.
FIG. 3 is a block diagram of a system 300 for real-time analysis and optimization of parameters for different stages of a drilling operation. The drilling operation may be, for example, an underwater drilling operation to drill a wellbore along a planned path through an underground formation at an offshore well site, as described below above with respect to Figure 1. Alternatively, the drilling operation may be an onshore drilling operation to drill the wellbore along a planned path through an underground formation at a site well to shore, as described above with respect to Figure 2. As shown in Figure 3, a system 300 includes a well planner 310, memory 320, a graphical user interface (GUI) 330, and a network interface 340. In one or more embodiments, the well planner 310 includes a data manager 312, a drilling optimizer 314, and a well controller 316. Although not shown in Figure 3, it must be understood that e system 300 may include additional components and subcomponents, which can be used
2017-IPM-101331-U1-FR 11 to provide the real-time analysis and optimization functionality described in this document.
The network interface 340 of the system 300 may include logic encoded in software, hardware, or a combination thereof to communicate with a network 304. For example, the network interface 340 may include software supporting one or more communication protocols such that equipment associated with the network interface 340 can communicate signals to other computer systems and devices via the network 304. The network 304 can be used, for example, to facilitate wireless communications or wired between system 300 and other computer systems and devices. In some implementations, the system 300 and other systems and devices may operate as separate components of a distributed computing environment in which the components are coupled in communication via the network 304. Although this is not shown in the drawing. Figure 3, it should be understood that these other systems and devices may include other local or remote computers such as, for example and without limitation, one or more client systems, servers, or other devices coupled in communication via the network 304.
Network 304 can be any or any combination of networks such as, but not limited to, a local area network, a medium network, or a wide area network, for example, the Internet. This or these networks can be all or part of a corporate or secure network. In some cases, part of network 304 may be a virtual private network (VPN) between, for example, a system 300 and other computers or other electronic devices. In addition, all or part of the network 304 may include a wired or wireless link. Examples of such wireless links include, but are not limited to, 802.1 la / b / g / n, 802.20, WiMax, and / or any other suitable wireless link. Network 304 can include any number of internal (private) or external (public) networks, subnets, or a combination of these to facilitate communications between various computer components including the system 300.
In one or more embodiments, the system 300 can use the network 304 to communicate with a database 350. The database 350 can be used to store data accessible to the system 300 to implement the modeling functionality and real-time guidance described in this document. The database 350 can be associated with or located at the level of the operations center of an oil field service provider, as described above with respect to the computer system 242 of FIG. 2. The stored data can include, for example, historical well data and parameters associated with drilling operations at various well sites, for example, other well sites within the same hydrocarbon production field as the well site
2017-IPM-101331-U1-FR 12 in this example. In addition or alternatively, the data may include data collected in real time from the well site during the various stages of the drilling operation. This real-time data can be retrieved from the database 350 via the network 304 and stored in a memory 320 as well site data 322, for example, to be retrieved and applied as input data to implement the modeling and optimization techniques disclosed in the document. In some implementations, the data may be broadcast from the database 350 as real-time data entered into a designated buffer or storage area corresponding to the well site data 322 in the memory 320.
In one or more embodiments, the well site data 322 may include data transmitted via the network 304 directly from a surface control system (for example, the surface computer system 240 of Figure 2, as described herein above) at the level of a drilling platform or an offshore platform using an industrial format such as the standard well site information transfer markup language (WITSML). WITSML is known to facilitate the free flow of technical data across networks between oil companies, service companies, drilling companies, application sellers and regulators for drilling, completion and interventions by the upstream petroleum and natural gas industry. However, it should be understood that well site data 322 can be transmitted and stored using any type of data format, standard, or structure desired for a particular implementation.
The stored well data 322 may include the values of the instant of controllable parameters, for example, a flow rate (Q), a load on the drill bit (WOB), and a bit rotation speed (RPM). However, it should be understood that well site data 322 may also include various measurements or other data collected at the well site. Examples of these other data include, but are not limited to, depth (the vertical depth within the formation and / or the measured depth of the wellbore, whether vertical or deviated), size of the drill bit, the length of the drill collar, the torque and the resistance on the train, the plastic viscosity, the elastic deformation limit, the weight of the mud, the resistance of the gel, the pressure at the bottom of the well, and the temperature.
In one or more embodiments, the data manager 312 of the well planner 310 can preprocess the stored well site data 322 or the real-time data received via the network 304 from the database 350 or a computer system of well site. Preprocessing can include, for example, filtering data in one
2017-IPM-101331-U1-EN 13 time series of sampling rate or drilling rate. In some implementations, the data manager 312 may include one or more data filters to reduce or cancel noise from data in real time. Examples of such filters include, but are not limited to, a convolutional neural network, a bandpass filter, a Kalman filter, a high pass filter, a low pass filter, a mean filter, a noise reduction filter, delay filter, summation filter, format conversion filter, and any other type of digital or analog data filter. The preprocessed data can then be classified for use in predicting and optimizing one or more operating variables and one or more controllable parameters for different stages of the drilling operation, such as will be described in more detail below.
In one or more embodiments, at least one operating variable of interest can be selected by a user 302 via the GUI 330. The operating variable selected by the user 302 can be, for example, at least one among ROP and HMSE. Fa or the user selected operating variables 302 in this example can be used to monitor drilling efficiency and trends in performance of the drilling operation as the wellbore is drilled. through training. In one or more embodiments, a display of the estimated values of the operating variable and / or of the controllable parameters affecting the operating variable can be presented to the user 302 via a display window or a content display area of the GUI 330. The GUI 330 can be displayed using any type of display device (not shown) coupled to the system 300. Such a display device can be, for example and without limitation, a cathode ray tube monitor (CRT), a liquid crystal display (LCD), or a light emitting diode (LED) monitor. The user 302 can interact with the GUI 330 using the input device (not shown) coupled to the system 300. The user input device can be, for example and without limitation, a mouse, a QWERTY or T9 keyboard, a touch screen, stylus or other pointer device, graphics tablet, or microphone. In some implementations, user 302 can use the information displayed via the GUI 330 to assess drilling performance at each stage of the operation and make any manual adjustments to the planned path of the well, for example, by entering appropriate commands into a drilling operations control module used to control drilling operations at the well site. However, it should be understood that such adjustments can be made automatically by an automatic control system for the well site.
2017-IPM-101331-U1-FR 14
During the drilling operation, drilling fluids are pumped into the wellbore to remove the cuttings produced while the drill bit enters the underground rock layers and forms the wellbore within the underground formation. The main physical and engineering aspects of the drilling process can be very complex, and the well site data collected as the well is drilled often includes a significant amount of noise and uncertainty. Therefore, the response surface for operating variables, such as ROP and HMSE, tends to be non-linear and discontinuous.
In one or more embodiments, a drilling optimizer 314 can use a neural network model with stochastic optimization to estimate or predict optimal values for both the operating variable (s) and the controllable parameters selected from the drilling operation which affect the operating variable (s) during the operation. Such a stochastic approach can provide the level of precision and speed necessary to implement real-time applications, for example, real-time modeling and geoguiding, in a relatively short period of time to optimize the path of the well. drilling while it is being drilled in a localized region of the formation during each stage of the drilling operation. An example of a neural network model with stochastic optimization is presented in Figure 4.
In FIG. 4, a neural network model 400 can use a stochastic optimization to optimize at least one operating variable (for example, maximize the ROP and / or minimize the HMSE) at each stage of a plurality of stages 402a, 402b, and 402c of a drilling operation along a well path 402. Each stage may correspond to an interval or a section of path from 402 along which part of a wellbore is drilled through an underground formation. Although three stages are shown in Figure 4, it should be understood that the drilling operation can include any number of stages. It should also be understood that each stage of the operation can be of any length or size and that the overall spacing of the stages along the well path 402 can be customized or configured as desired for a particular implementation . For example, in some implementations, each stage of the drilling operation may be carried out over a predetermined length or depth interval (for example, 30 feet) along the well path and the stages may be adjacent to each other. to others.
Although the drilling operation is carried out along the well path 402, a drilling operator or an automatic control system at the well site can adjust the values of one or more controllable parameters, for example, WOB , RPM, and Q,
2017-IPM-101331-U1-EN 15 to take into account changes in drilling conditions. The value of the operating variable can also change in response to changes made to the controllable parameters. Consequently, the operating variable in this context can be designated by response variable and a value of the operating variable by response value. In one or more embodiments, real-time data including time values of controllable parameters can be collected at the well site during each stage 402a, 402b and 402c. The real-time data can be multidimensional time data, for example, samples of drilling data captured with depth during a time series, which may correspond to the speed of drilling. The neural network model 400 can be used to couple depth data with non-linear constraints to resolve the temporal and spatial variation of the response variable during the drilling operation.
In one or more embodiments, the values of the controllable parameters associated with a stage of the instant (for example, 402a) of the drilling operation can be applied as input variables to drive the network model. neurons 400 in order to produce an objective function defining a response value for the operating variable to be optimized for a later stage (for example, 402b and / or 402c) of the operation. For example, the objective function can define a response value for the ROP in terms of WOB, RPM, and Q, as expressed using equation (1):
ROP = f (W0B, RPM, Q) (1)
The objective function in this context can be a cost function, which can be maximized or minimized depending on the operating variable of particular interest, for example, maximized for the ROP and minimized for the HMSE.
To account for any high level of non-linearity and / or noise in real-time or time series drilling speed data, the objective function produced by the neural network model 400 to define the value of response of the operating variable may be subject to a set of non-linear constraints 410. The non-linear constraints 410 may be derived from data models representing different aspects of the drilling operation which may be associated with certain values of the controllable parameters and which can influence the response value of the operating variable to be changed during the drilling operation. The data models in this example may include, but are not limited to, a torque and resistance model ("T&D") 412, a vortex model 414, and a drilling fluid model ("DMF") 416 .
2017-IPM-101331-U1-FR 16
In one or more embodiments, simulations to determine the appropriate constraints can be performed by applying the real-time data acquired during the operation as inputs to each of these models. For example, the torque and resistance model 412 can be used to simulate the forces exerted on the drill bit by friction on the underground formation in which the wellbore is formed. The torque and resistance model 412 can therefore provide a threshold on the WOB in order to avoid excessive wear which could lead to the failure of the drill bit or of another component of the drilling assembly fixed to the end of the train. drilling. The 414 vortex model can be used to simulate vibration forces in the drill string that can cause damage to certain RPM values. As the RPM values can change with the length and depth of the drill string, the 414 vortex model can be used to constrain RPM within ranges of safety values to avoid excessive vibration at a WOB given. Drilling fluid model 416 can be used to simulate the injection of drilling fluid (for example, mud) used to remove cuttings or debris from the wellbore during the drilling operation. The ROP of the drill bit may be limited by the maximum amount of debris that can be removed from the wellbore by injecting or pumping a fluid during a given period of time. Therefore, drill fluid model 416 can provide a maximum fluid injection or pumping speed at which debris-laden fluid can be removed from the wellbore.
The neural network model 400 with the constraints applied to the objective function, as described above, can then be used to estimate or predict a response value for the operating variable to be optimized for a later stage of the operation. drilling along the well path 402. In one or more embodiments, stochastic optimization, for example, Bayesian optimization, can be applied to the response value to produce an optimized response value.
As shown in Figure 5, Bayesian optimization (BO) can be applied iteratively to restrict a neural network model if necessary to satisfy a predetermined criterion. Such a criterion can be, for example, an error tolerance threshold, and the neural network model can be re-trained whenever it is determined that a difference between the estimated response value and a value value of the operating variable exceeds the threshold. The actual value of the operating variable can be based on additional real-time data acquired during a later stage of the drilling operation. In one or more embodiments, the neural network model can be re-trained by the application of Bayesian optimization to one or more hyperparameters of the
2017-IPM-101331-U1-FR 17 model. Examples of such hyperparameters include, but are not limited to, the number of layers in the neural network, the number of nodes in each layer, the rate of learning decay, and any other behavioral parameter and / or the capacity of the model.
Returning Figure 4, the optimized response value produced by the neural network model 400 can then be used to predict or estimate optimal values of controllable parameters 420. The controllable parameters 420 in this example may include, but are not limit it, WOB 422, the speed of the drill bit (or RPM) 424, and the flow rate (Q) 426. The flow rate 426 can be the speed at which a fluid (for example, a mud) is pumped into the wellbore.
Returning to the system 300 of Figure 3, the modeling and simulation operations described above to optimize the response value and the controllable parameter values using the neural network model 400 can be performed by a drilling optimizer 314, based on real time data acquired and preprocessed by the data manager 312. The value and / or response values of the controllable parameters can be stored as output data 324 in memory 320.
In one or more embodiments, the drilling optimizer 314 can provide the estimated values of the controllable parameters to a well controller 316 of the well planner 310 to perform one or more stages of the drilling operation at the level well site. The well controller 316 can provide the parameter values as control inputs to a downhole geoguiding tool (not shown), which can be used to orient the drill bit and the wellbore. along a planned or adjusted route through training. For example, the well control device 316 can be coupled in communication with the downhole geoguiding tool via a wireless or wired communication interface (not shown) (for example, a wired line) of the system 300. Such a communication interface can be used by the well control device 316 to transmit the values of controllable parameters in the form of control signals to the downhole geoguiding tool. The control signals may allow the well control device 316 to control, for example, the direction and orientation of the geo-guidance tool and thereby adjust the planned path of the well during the drilling operation.
As the operation is performed along the planned well path, additional well site data may be collected by a downhole sensor (for example, inside the downhole tool well 106 of Figures 1 and 2, as
2017-IPM-101331-U1-FR 18 described above), measuring devices on the surface of the wellbore or a combination of both. This data may include, for example and not be limited to, values of the instant of controllable parameters, for example WOB, RPM and Q. However, it should be understood that the data collected may also include measurements of formation properties and other data regarding the current downhole operation. As described above, this well site data can be obtained directly or indirectly by the system 300 via the network 304. In one or more embodiments, the drilling optimizer 314 can use this additional data to automatically update and further optimize the response value of the selected operating variable (s) (for example, ROP and / or HMSE) for later stages of the operation along the drilling path.
In one or more embodiments, the neural network model used by the drilling optimizer 314 to estimate the response value of the operating variable and the values of the controllable parameters, as described above, can be at least one from a sliding window neural network (SWNN) or a recurrent deep neural network (DNN).
Figure 6 is a diagram of a sliding window neural network (SWNN) 600 for predicting the values of one or more operating variables of a drilling operation along a well path. The SWNN 600 can be used to predict a response value for at least one operating variable, for example ROP and / or HMSE. The SWNN 600 can be trained using multivariate time series data acquired during a drilling operation along the well path. This drilling data can include real-time data sampled on a sliding window of the SWNN 600, as shown in Figure 6. The sliding window of the SWNN 600 can be a sampling interval of any size or length along from the well path. For example, the sliding window can correspond to a predetermined depth interval (for example, 30 feet) that the SWNN 600 can use to incrementally sample real-time data by dragging the window along the depth of the well. drilling, for example, by moving the window position between different depth increments along the well path. The size of the sliding window or the sampling interval can correspond to a stage of the drilling operation or to a part of an entire stage. In one or more embodiments, the real-time drilling data acquired on a first part of the sliding window (for example, the first 24 feet) along the well path can be used to train the SWNN 600 and the data acquired on the remaining part (for example, 6 feet) can
2017-IPM-101331-U1-EN 19 be used to test or validate the trained model to determine if retraining is necessary.
FIG. 7 is a diagram of a recurrent deep neural network (DNN) 700 having one or more recurrent gate unit cells (GRU) for predicting the values of one or more operating variables of an operation of drilling along a well path. However, it should be understood that the disclosed embodiments are not intended to be limited to GRU cells and that the disclosed techniques can be applied using recurrent DNN with other cell types, for example, memory cells short and long term (LSTM). Like the SWNN 600 in Figure 6, the DNN 700 can be trained using multivariate time series drilling data acquired along the well path. Part of the acquired data can be used to train the DNN 700, while the rest can be used to test and validate the trained model. As shown in Figure 7, the DNN 700 in the present example may include multiple GRU cells in a stacked configuration, where each GRU cell in the stack may represent a layer of the DNN 700 in which the DNN 700 can be driven iteratively during a series of time steps.
Figure 8 is a diagram of a GRU 800 cell illustrating the DNN 700 shown in Figure 7. The rectangular boxes indicate the layers in the GRU 800 cell, which have weights and biases associated with it. Circular and elliptical shapes indicate mathematical operations. In the schematic representation of the GRU 800 cell as shown in FIG. 8, h _t .i can be the state of the cell or the output ROP coming from an earlier time step M, also expressed by ROPt-i. The term xt can represent the multivariate input for the time step of the instant, which includes WOB (r _wo b, t), _RPM (r _rpm , f) and Q (r _q , t) coming from the temporal stages of the instant and earlier in a predefined search window of the GRU 800 cell.
The GRU 800 cell in the present example may have four layers, each of which may have weights and biases associated therewith. These weights and biases can be trained during the training process to provide optimal predictions of treatment pressure in the time series. The variables / z, and o can correspond to values for "forget", "enter", and "exit" doors. These gates may involve calculations based on a sigmoid function (σ), where the resulting values fall in the range [0, 1]. The resulting values can define the amount of information to be transmitted from the previous time step to the next time step.
2017-IPM-101331-U1-FR 20
In one or more embodiments, a set of mathematical operations, expressed by equations (2) to (6) below, can be performed to calculate a state of the GRU 800 cell or an output ROP at each step ζ namely, ht (ROPt
): At lfwob, t Frpm, t ’^ q, t] (2) ^z t = σ (^ ζ 'k-iAi]) (3) ^r t = · [Vi, uD (4)(5) ^h t = + ^z t * h _t (6) whereROPt indicates the ROP predicted in time step Z; x is the entry for each
time step, which can include the values of WOB (r _wo b, t Tr / min (r _rp m, f) and Q (r _q , t) which are shared by all the stacked layers; z _t is the update gate vector; W _z represents the weights of the update gate; r _t is the reset gate vector; W _r represents the weights of the reset gate; is the output of equation (5) and serves as an intermediate value used in equation (6) to calculate the final output ht and W represents the weights for the final output.
Returning to DNN 700 in Figure 7, equations (2) to (6) above can be used to calculate a state of each output GRU or ROP cell (for example, a predicted response value for the ROP) at each time step. In one or more embodiments, the cell state and the output ROP from an individual layer of the DNN 700 can be transmitted from one time step to the next in the same layer and thus provide the basis for an input formulation to the next time step. A final predicted ROP can be obtained by combining the predicted ROP values from all the layers stacked at a given time step. The stacked GRU configuration and other variants of the DNN 700 can help capture highly non-linear variations in the time series data acquired during the drilling operation. This makes these recurring DNNs an ideal choice for reducing ROP during drilling, especially given the highly non-linear nature of ROP time series.
In some implementations, the DNN 700 can incorporate a quadratic error loss and backpropagation over time (BPTT) architecture. An example of such a recurrent DNN architecture is shown in Figure 9. In Figure 9, the input data for the recurrent DNN is passed through a convolutional neural network (CNN) to filter noise. A previous DNN output, for example, a response value of the variable
2017-IPM-101331-U1-EN 21 operation which was estimated during an earlier stage of the drilling operation or which was acquired from an external source, such as a boundary sink, is passed into a noise filter, for example, a Kalman filter or an autoencoder, to remove noise from the data in order to eliminate or at least reduce noise before the data is entrained inside the DNN and the stochastic optimization, for example, the Bayesian optimization (BO) described above, be applied.
Figure 10 is a diagram of an illustrative method 1000 of optimizing drilling parameters using a neural network model for real-time planning and control during different stages of a drilling operation along a path well. For discussion purposes, the method 1000 will be described using the system 300 of Figure 3, as described above. However, it is not intended that the method 1000 is limited to this. In addition, for discussion purposes, a method will be described using the drilling systems 100 and 200 of Figures 1 and 2, respectively, but is not intended to be limited thereto. The operations of blocks 1002, 1004, 1006, 1008, 1010, 1012, 1014, 1016, and 1018 of method 1000 can be carried out, for example, by one or more components of the well planner 310 of system 300, as described below. above.
Method 1000 begins at block 1002, which includes acquiring real-time data, including values for a plurality of input variables, for a point in the time of a drilling operation along a path planned a borehole within an underground formation.
In block 1004, a neural network model is trained to produce an objective function defining a response value of at least one operating variable to be optimized during the drilling operation along the planned path, based on the real-time data acquired in block 1002. In one or more embodiments, the operating variable can be selected by the user, for example, the user 302 of the system 300 of FIG. 3, as described above . Therefore, although not shown in Figure 10, block 1002 or 1004 can also include receiving input from the user (for example, via the GUI 330 in Figure 3) with the selection of the user operating variable. The selected operating variable can be, for example and not limited to, ROP and / or HMSE.
In block 1006, a response value for the at least one operating variable is estimated, based on the objective function produced by the trained neural network model.
2017-IPM-101331-U1-FR 22
Method 1000 then proceeds to block 1008, which includes applying Bayesian optimization to the response value defined by the objective function so that the trained neural network model produces an optimized response value for at least an operating variable.
In block 1010, the optimized response value of the operating variable is used to estimate the values of a plurality of controllable parameters for a later stage of the drilling operation.
In block 1012, the later stage of the drilling operation is implemented based on the estimated values of the plurality of controllable parameters.
In block 1014, a real value of the operating variable can be determined at the later stage (on the basis of additional data acquired during this stage) and the method 1000 can pass to block 1016.
Block 1016 may include determining whether a difference between the actual value of the at least one operating variable and the response value, as estimated and optimized in blocks 1006 and 1008, respectively, exceeds a threshold of error tolerance. When it is determined that the difference exceeds the error threshold, method 1000 proceeds to block 1018, which includes retraining the neural network model by applying stochastic optimization (for example, Bayesian optimization ) to one or more hyperparameters of the model, as described above. Method 1000 then returns to block 1006 and the operations described above in blocks 1006, 1008, 1010, 1012, 1014 and 1016 can be repeated for the next stage of the drilling operation using the neural network model retrained. Otherwise, process 1000 can return to block 1006 directly from block 1016 so that the operations in the blocks described above can be repeated for the next stage of the drilling operation using the previously trained model.
Figures 11 A, 1 IB, 1 IC and 11D are graphs representing the penetration rate (ROP) values as predicted using a SWNN model versus actual ROP values along a well path as a function of the depth. The size of the SWNN sliding window used for predictions in the example shown in Figures 11A and 1 IB, as well as for the example shown in Figures 1 IC and 11D, is assumed to be 30 feet, where the values of the first 24 feet of each window (for example, as shown in each of Figures 11A and 11C) are used to train the SWNN and values for the next or last 6 feet of the window (as shown in each of Figures 1 IB and 11D) are used to test the model's predictions and re-train the model if necessary. It can also be assumed in this example that no SWNN retraining has
2017-IPM-101331-U1-EN 23 been necessary, for example, because the predictions produced by the SWNN meet the retraining criterion or the error tolerance threshold. For example, the retraining criterion or the error threshold can be a specified mean square error value (for example, 0.2) and the difference between an actual value of the operating variable and the response value predicted using SWNN may be less than this mean square error value.
Figures 12A and 12B are graphs showing a comparison between predicted ROP values and normalized actual ROP values along a well path over a depth (e.g., in feet), where the predicted values are based on a DNN recurring with or without a noise filter, respectively. As shown in the graph in each of Figures 12A and 12B, the predicted values tend to be closer to the actual values when a noise filter, for example, a Kalman filter or a denoising autoencoder, is used. Using such a filter to produce more accurate predictions of ROP in this example may also indicate that a number of input variables may be unknown or missing. Therefore, the accuracy and / or efficiency of the DNN model to produce the ROP response in the present example can be further improved by increasing the number of input variables that are used to train or re-train the model appropriately. . For example, the model can be re-trained using additional input variables, for example, reservoir properties or other information regarding the characteristics of the underground formation, which may affect the ROP during a mining operation. drilling.
Although the various embodiments are described in this document in the context of surface computer systems, it should be noted that the disclosed parameter modeling and optimization techniques are not intended to be limited to these. In one or more embodiments, some or all of the calculations concerning the operating variable and / or the controllable parameters can be performed by a processor inside a downhole tool disposed inside the wellbore near the drill bit. For example, the telemetry module 124 of Figures 1 and 2, as described above, may include a computer system for performing such well bottom calculations. The telemetry module 124 may include an encoding system, such as a mud pulse generator, to communicate (for example, by telemetry) some or all of the results of the calculations to surface computer systems. In cases where the control of the operational parameter is automated, the telemetry module 124 or another downhole computer system (for example, a downhole geoguiding tool) coupled to it can be used to control or change one or more controllable parameters (by
2017-IPM-101331-U1-EN 24 example, the RPM or the speed of the mud motor 112, the WOB, and / or the injection / flow of a fluid).
Figure 13 is a block diagram of an illustrative computer system 1300 in which embodiments of this disclosure can be implemented. For example, the method 1000 in FIG. 10 and the functions implemented by the system 300 (in particular the well planner 310) in FIG. 3, as described above, can be implemented using a system 1300 A 1300 system can be a computer, telephone, PD A, or any other type of electronic device. Such an electronic device includes various types of computer-readable media and interfaces for various other types of computer-readable media. As shown in Figure 13, the system 1300 includes a permanent storage device 1302, a system memory 1304, an output device interface 1306, a system communication bus 1308, a read only memory (ROM) 1310, one or more units 1312, an input device interface 1314, and a network interface 1316.
A bus 1308 collectively represents all the system, peripheral and chipset buses which connect the numerous internal devices of the system 1300 in a communicative manner. For example, a bus 1308 connects one or more processing units 1312 with a ROM 1310 , a system memory 1304 and a permanent storage device 1302.
From these various memory units, one or more processing units 1312 extract instructions to be executed and data to be processed in order to execute the processes of the present disclosure. One or more processing units can be a single processor or a multi-core processor in different implementations.
A ROM 1310 stores data and instructions that are necessary for one or more processing units 1312 and other modules of a system 1300. Furthermore, a permanent storage device 1302 is a read and write memory device. This device is a non-volatile memory unit that stores instructions and data even when a 1300 system is powered down. Some implementations of the present disclosure use a mass storage device (such as a magnetic or optical disc and its corresponding disc drive) as a permanent storage device 1302.
Other implementations use a removable storage device (such as a floppy disk, a flash drive and its corresponding disk drive) as a permanent storage device 1302. Just like a permanent storage device 1302, system memory 1304 is a read and write memory device. However,
2017-IPM-101331-U1-EN 25 unlike the storage device 1302, the system memory 1304 is a volatile memory for reading and writing, such as a random access memory. A system memory 1304 stores some of the instructions and data that the processor needs at run time. In some implementations, the processes of this disclosure are stored in system memory 1304, permanent storage device 1302 and / or ROM 1310. For example, the various memory units include instructions for a train design. computer-aided pipes based on existing train designs according to certain implementations. From these various memory units, one or more processing units 1312 extract instructions to be executed and data to be processed in order to execute the processes of certain implementations.
A bus 1308 also connects to input and output device interfaces 1314 and 1306. An input device interface 1314 allows the user to communicate information and select commands to the 1300 system. Inputs used with an input device interface 1314 include, for example, an alphanumeric keyboard, QWERTY or T9, microphones and pointing devices (also called "cursor control devices"). Output device interfaces 1306 allow, for example, the display of images generated by the 1300 system. Output devices used with an output device interface 1306 include, for example, printers and display devices , such as cathode ray tube (CRT) or liquid crystal (ECD) screens. Some implementations include devices, such as a touch screen, that act as both input and output devices. It will be understood that embodiments of the present disclosure can be implemented using a computer including any of various types of input and output devices to allow interaction with a user. Such interaction may include feedback to or from the user in various forms of sensory feedback such as, but not limited to, visual feedback, auditory feedback, or tactile feedback. In addition, user input can be received in any form such as, but not limited to, acoustic, voice, or touch input. In addition, interaction with the user may include transmission and reception of different types of information, for example in the form of documents, to or from the user through the interfaces described above.
In addition, as shown in Figure 13, a bus 1308 also couples a system 1300 to a public or private network (not shown) or to a combination of networks via a network interface 1316. Such a network may include , for example, a network
2017-IPM-101331-U1-EN 26 local (“LAN”), such as an intranet, or a wide area network (“WAN”), such as the Internet. All or part of the components of a 1300 system can be used in conjunction with this disclosure.
These functions described above can be implemented in a digital electronic circuit, in computer software, firmware or hardware. The techniques can be implemented using one or more computer program products. Programmable processors and computers can be included in mobile devices or packaged as mobile devices. Logic processes and flows can be performed by one or more programmable processors and by one or more programmable logic circuits. Computing devices and storage devices for general or specific applications can be interconnected via communication networks.
Some implementations include electronic components, such as microprocessors, storage, and memory that store computer program instructions on machine-readable media (otherwise known as computer-readable storage media, machine, or machine-readable storage medium). Some examples of such computer-readable media include RAM, ROM, read-only compact discs (CD-ROM), recordable compact discs (CD-R), rewritable compact discs (CD-RW), digital discs versatile read-only (for example, DVD-ROM, double-layer DVD-ROM), various recordable / rewritable DVDs (for example, DVD-RAM, DVD-RW, DVD + RW, etc.), flash memory (e.g. SD cards, mini SD cards, micro SD cards, etc.), magnetic and / or solid state hard drives, read-only and recordable Blu-Ray® discs , ultra-density optical discs, any other optical or magnetic medium, and floppy disks. The computer-readable medium can store a computer program which is executable by at least one processing unit and which includes sets of instructions for performing various operations. Examples of computer programs or computer codes include machine code, as produced by a compiler, and files comprising higher level code which are executed by a computer, electronic component, or a microprocessor using an interpreter.
Although the above discussion primarily refers to a microprocessor or multi-core processors that run software, some implementations are performed by one or more integrated circuits, such as Application Specific Integrated Circuits (ASICs) or programmable pre-broadcast integrated circuits (FPGA). In
2017-IPM-101331-U1-EN 27 certain implementations, such integrated circuits execute instructions which are stored on the circuit itself. Accordingly, the method 1000 of Figure 10 and the functions and operations performed by the system 300 of Figure 3, as described above, can be implemented using a 1300 system or any computer system having a processing circuit or a computer program product including instructions stored thereon, which, when executed by at least one processor, cause the processor to perform functions related to these methods.
As used herein and in any one of the claims of this application, the terms "computer", "server", "processor", and "memory" all refer to electronic devices or other technological devices. These terms exclude individuals or groups of individuals. As used herein, the terms "computer readable media" and "computer readable media" generally refer to tangible, physical, non-transient electronic storage media that store information in a form that can be read by a computer.
Embodiments of the subject described herein may be implemented in a computer system which includes a background component, eg, a data server, or which includes a middleware component, eg, an application server, or one that includes a front-end component, eg, a client computer having a graphical user interface or a web browser through which a user can interact with an implementation of the subject described herein, or n any combination of one or more of these middleware, background, or front-end components. System components can be interconnected by any form or any digital data communication medium, for example a communication network. Examples of communication networks include a local area network ("LAN") and a wide area network ("WAN"), internetwork (for example, Internet) and peer-to-peer networks (for example, peer-to-peer networks -pair ad hoc).
The computer system can include clients and servers. A client and a server are generally distant from each other and traditionally interact through a communication network. The client-server relationship stems from computer programs running on the respective computers and having a client-server relationship to each other. In some embodiments, a server transmits data (e.g., a web page) to a client device (e.g., for purposes of displaying data to and receiving user input from a user interacting with the
2017-IPM-101331-U1-FR 28 client device). Data generated at the client device (eg, a result of user interaction) can be received from the client device at the server.
It is understood that any specific order or hierarchy of steps in the disclosed processes is an illustration of examples of approaches. Based on design preferences, it is understood that the specific order or hierarchy of steps in the processes can be rearranged, or that all of the illustrated steps can be performed. Some of these steps can be done simultaneously. For example, in certain circumstances, multitasking and parallel processing may be advantageous. In addition, it should be understood that the separation of the various components of the systems in the embodiments described above is not necessary in all embodiments, and it should be understood that the program components and the systems described can generally be integrated together into a single software product or packaged into multiple software products.
In addition, the example methodologies described here can be implemented by a system including a processing circuit or a program product including instructions which, when executed by at least one processor, cause the processor to perform any of the methodologies described here.
As described above, embodiments of the present disclosure are particularly useful for real-time optimization of parameters during drilling operations. In one embodiment of the present disclosure, a computer implemented method of optimizing drilling operation parameters includes: real-time data acquisition including values for a plurality of input variables associated with a stage of the time of a drilling operation along a planned path of a wellbore within an underground formation; training a neural network model to produce an objective function defining a response value for at least one operating variable to be optimized during the drilling operation along the planned path, based on the time data real acquired; estimating the response value for at least one operating variable, based on the objective function produced by the trained neural network model; applying stochastic optimization to the estimated response value in order to produce an optimized response value for the at least one operating variable; estimating the values of a plurality of controllable parameters for a later stage of the drilling operation, based on the optimized response value of the at least one operating variable; and implementing the later stage of the drilling operation based on the estimated values of the plurality of controllable parameters. In another embodiment of the present disclosure, a computer-readable storage medium having instructions
2017-IPM-101331-U1-EN 29 stored therein is disclosed, in which the instructions, when executed by a computer, cause the computer to perform a plurality of functions, including functions for: acquiring data by real-time comprising values for a plurality of input variables associated with a stage in the instant of a drilling operation along a planned path of a wellbore within an underground formation; training a neural network model to produce an objective function defining a response value for at least one operating variable to be optimized during the drilling operation along the planned path, based on the acquired real time data; estimate the response value for at least one operating variable, based on the objective function produced by the trained neural network model; applying a stochastic optimization to the estimated response value in order to produce an optimized response value for the at least one operating variable; estimating the values of a plurality of controllable parameters for a later stage of the drilling operation, based on the optimized response value of the at least one operating variable; and implementing the later stage of the drilling operation based on the estimated values of the plurality of controllable parameters.
In one or more embodiments of the preceding method and / or computer-readable storage medium: the at least one operating variable is at least one of a penetration rate (ROP) or a specific mechanical hydraulic energy (HMSE) ), the stochastic optimization is a Bayesian optimization, and the optimized response value is at least one of a maximum value for the ROP or a minimum value for the HMSE; the values for the plurality of input variables are initial values for the plurality of controllable parameters associated with the instant stage of the drilling operation; the plurality of controllable parameters includes the load on the drill bit (WOB), a speed of rotation of a drill bit, and a pumping speed of drilling fluid; the neural network model is a sliding window neural network (SWNN); the neural network model is a recurrent deep neural network (DNN); and the recurring DNN includes at least one of a gateway recurring unit cell (GRU) or a short and long term memory cell (LSTM). In addition, one or more embodiments of the foregoing computer readable storage method and / or medium may include any or any combination of the following additional elements, functions or operations: determining an actual value of the at least an operating variable during the later stage of the drilling operation; determining whether a difference between the actual value and the estimated value of the at least one operating variable exceeds an error tolerance, and retraining the neural network model when it is determined that the difference exceeds tolerance
2017-IPM-101331-U1-FR 30 error; re-training by applying a Bayesian optimization to one or more hyperparameters of the neural network model; noise filtering training from real-time data and recurrent DNN training based on filtered real-time data; and filtering by applying the real-time data as input to a convolutional neural network and passing the output data including a previously estimated response value of the at least one operating variable through a Kalman filter, and entrainment of recurrent DNN by entrainment of recurrent DNN on the basis of an output from the convolutional neural network and an output from the Kalman filter.
In yet another embodiment of the present disclosure, a system includes at least one processor and a memory coupled to the processor which has instructions stored therein, which when executed by the processor, cause the processor to activate performs functions comprising functions for: acquiring real-time data comprising values for a plurality of input variables associated with a stage of the instant of a drilling operation along a planned path of a well drilling in an underground formation; training a neural network model to produce an objective function defining a response value for at least one operating variable to be optimized during the drilling operation along the planned path, based on the acquired real time data; estimate the response value for at least one operating variable, based on the objective function produced by the trained neural network model; applying a stochastic optimization to the estimated response value in order to produce an optimized response value for the at least one operating variable; estimating the values of a plurality of controllable parameters for a later stage of the drilling operation, based on the optimized response value of the at least one operating variable; and implementing the later stage of the drilling operation based on the estimated values of the plurality of controllable parameters.
In one or more embodiments of the previous system: the at least one operating variable is at least one of a penetration rate (ROP) or a specific mechanical hydraulic energy (HMSE), the stochastic optimization is an optimization Bayesian, and the optimized response value is at least one of a maximum value for the ROP or a minimum value for the HMSE; the values for the plurality of input variables are initial values for the plurality of controllable parameters associated with the instant stage of the drilling operation; the plurality of controllable parameters includes the load on the drill bit (WOB), a speed of rotation of a drill bit, and a pumping speed of drilling fluid; the neural network model is a windowed neural network
2017-IPM-101331-U1-FR 31 sliding (SWNN); the neural network model is a recurrent deep neural network (DNN); and the recurring DNN includes at least one of a gateway recurring unit cell (GRU) or a short and long term memory cell (LSTM). In addition, in one or more embodiments of the previous system, the functions implemented by the processor can further comprise functions for: determining a real value of the at least one operating variable during the subsequent stage of the drilling operation; determining if a difference between the actual value and the estimated value of the at least one operating variable exceeds an error tolerance; re-train the neural network model when it is determined that the difference exceeds the error tolerance; apply Bayesian optimization to one or more hyperparameters of the neural network model; filter noise from real-time data; train recurring DNN based on filtered real-time data; apply the data in real time as input to a convolutional neural network; passing the output data comprising a previously estimated response value of the at least one operating variable through a Kalman filter; and train the recurrent DNN based on an output from the convolutional neural network and an output from the Kalman filter.
Although specific details on the above-mentioned embodiments have been described, the preceding descriptions of hardware and software are merely exemplary embodiments and are not intended to limit the structure or implementation of the disclosed embodiments . For example, although many other internal components of the 1300 system are not shown, the ordinary specialist in the field will understand that these components and their interconnection are well known.
In addition, certain aspects of the disclosed embodiments, as noted above, can be implemented in software which is executed using one or more processing units / components. Program aspects of the technology can be thought of as "products" or "articles of manufacture" traditionally in the form of executable code and / or associated data which is carried or implemented in a readable medium type per machine. Tangible non-transient “storage” media include all or part of the memory or other storage device for computers, processors or the like, or associated modules thereof, such as various semiconductor memories , tape drives, disc drives, optical or magnetic discs, and the like, which can provide storage at all times for software programming.
In addition, the process diagram and the functional diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments.
2017-IPM-101331-U1-EN 32 of this disclosure. It should also be noted that, in certain alternative implementations, the functions noted in the block may occur in a different order from that noted in the figures. For example, two blocks shown in succession can, in reality, be executed substantially at the same time, or the blocks can sometimes be executed in reverse order, depending on the functionality concerned. It should also be noted that each block of block diagrams and / or illustrative process diagrams, and combinations of blocks in block diagrams and / or illustrative process diagrams, can be implemented by hardware-based systems specialist who performs specialized functions or actions, or combinations of specialized hardware and computer instructions.
The examples of specific embodiments above are not intended to limit the scope of the claims. The exemplary embodiments can be modified by inclusion, exclusion, or combination of one or more specificities or functions described in the disclosure.
As used here, the singular forms "one", "one", "the" and "the" are also intended to include plural forms unless the context clearly indicates otherwise. It will further be understood that the terms "comprises" and / or "comprising", when used in the present specification and / or the claims, indicate the presence of specific features, integers, steps, operations, d 'elements and / or components indicated, but do not prevent the presence or addition of one or more other specificities, integer, steps, operations, elements, components, and / or groups thereof. The structures, materials, corresponding actions, and equivalents of any means or steps plus functional elements in the claims below are intended to include any structure, material, or action to perform the function in combination with other claimed elements as specifically claimed. The description of this disclosure has been presented for illustrative and descriptive purposes, but it is not intended to be exhaustive or to be limited to the embodiments in the disclosed form. Many modifications and variations will be apparent to the ordinary specialist in the field without departing from the scope and spirit of the disclosure. The illustrative embodiments described herein are provided to explain the principles of disclosure and the practical application thereof, and to allow other persons skilled in the art to understand that the disclosed embodiments can be modified as is desired for a particular implementation or use. The scope of the claims is intended to broadly cover the disclosed embodiments and any such modifications.

权利要求:
Claims (15)
[1" id="c-fr-0001]
THE CLAIMS ARE AS FOLLOWS:
1. Method implemented by computer of an optimization of parameters for drilling operations, the method comprising:
real-time data acquisition including values for a plurality of input variables associated with a point in time in a drilling operation along a planned path from a wellbore within underground formation;
training a neural network model to produce an objective function defining a response value for at least one operating variable to be optimized during the drilling operation along the planned path, based on the time data real acquired;
estimating the response value for at least one operating variable, based on the objective function produced by the trained neural network model;
applying stochastic optimization to the estimated response value in order to produce an optimized response value for the at least one operating variable;
estimating the values of a plurality of controllable parameters for a later stage of the drilling operation, based on the optimized response value of the at least one operating variable; and implementing the later stage of the drilling operation based on the estimated values of the plurality of controllable parameters.
[2" id="c-fr-0002]
2. Method according to claim 1, in which the at least one operating variable is at least one of a penetration rate (ROP) or a specific mechanical hydraulic energy (HMSE), the stochastic optimization is a Bayesian optimization , and the optimized response value is at least one of a maximum value for the ROP or a minimum value for the HMSE; or wherein the values for the plurality of input variables are initial values for the plurality of controllable parameters associated with the stage of the instant of the drilling operation; or wherein the plurality of controllable parameters includes the load on the drill bit (WOB), a speed of rotation of a drill bit, and a pumping speed of drilling fluid.
2017-IPM-101331-U1-EN
[3" id="c-fr-0003]
3. Method according to claim 1 or 2, further comprising:
determining an actual value of the at least one operating variable during the later stage of the drilling operation;
determining whether a difference between the actual value and the estimated value of the at least one operating variable exceeds an error tolerance; and when it is determined that the difference exceeds the error tolerance, retraining the neural network model.
[4" id="c-fr-0004]
4. Method according to claim 3, in which the retraining comprises:
applying a Bayesian optimization to one or more hyperparameters of the neural network model.
[5" id="c-fr-0005]
5. Method according to any preceding claim, in which the neural network model is at least one of a sliding window neural network (SWNN) or a recurrent deep neural network (DNN); and optionally, wherein the recurrent DNN includes at least one of a gateway recurring unit cell (GRU) or a short and long term memory cell (LSTM).
[6" id="c-fr-0006]
6. Method according to claim 5, in which the training comprises:
filtering noise from real-time data; and training recurring DNN based on filtered real-time data.
[7" id="c-fr-0007]
7. Method according to claim 6, in which the filtering comprises:
applying real-time data as input to a convolutional neural network; and passing the output data comprising a previously estimated response value of the at least one operating variable through a Kalman filter; and in which recurrent DNN training includes:
2017-IPM-101331-U1-FR 35 training of recurrent DNN on the basis of an output from the convolutional neural network and an output from the Kalman filter.
[8" id="c-fr-0008]
8. System comprising:
at least one processor; and a memory coupled to the processor having instructions stored therein, which when executed by the processor, cause the processor to implement a plurality of functions, in particular functions for:
acquire real-time data comprising values for a plurality of input variables associated with a stage of the instant of a drilling operation along a planned path of a wellbore within a formation underground;
training a neural network model to produce an objective function defining a response value for at least one operating variable to be optimized during the drilling operation along the planned path, based on the acquired real time data;
estimate the response value for at least one operating variable, based on the objective function produced by the trained neural network model;
applying a stochastic optimization to the estimated response value in order to produce an optimized response value for the at least one operating variable;
estimating the values of a plurality of controllable parameters for a later stage of the drilling operation, based on the optimized response value of the at least one operating variable; and implementing the later stage of the drilling operation based on the estimated values of the plurality of controllable parameters.
[9" id="c-fr-0009]
9. The system as claimed in claim 8, in which the at least one operating variable is at least one of a penetration rate (ROP) or a specific mechanical hydraulic energy (HMSE), the stochastic optimization is a Bayesian optimization , and the optimized response value is at least one of a maximum value for the ROP or a minimum value for the HMSE; or wherein the values for the plurality of input variables are initial values for the plurality of controllable parameters associated with the instant stage of the drilling operation, and the plurality of controllable parameters includes the load on the drill bit ( WOB), a speed of rotation of a drill bit, and a speed of pumping of drilling fluid.
2017-IPM-101331-U1-EN
[10" id="c-fr-0010]
10. The system as claimed in claim 8 or 9, in which the functions implemented by the processor further comprise functions for:
determining an actual value of the at least one operating variable during the later stage of the drilling operation;
determining if a difference between the actual value and the estimated value of the at least one operating variable exceeds an error tolerance; and re-train the neural network model when it is determined that the difference exceeds the error tolerance.
[11" id="c-fr-0011]
11. The system as claimed in claim 10, in which the neural network model is re-trained by the application of a Bayesian optimization to one or more hyperparameters of the neural network model, and the one or more hyperparameters are selected in the group consisting of: a number of layers of the neural network model; a number of nodes in each layer of the neural network model; and a learning decay rate of the neural network model.
[12" id="c-fr-0012]
The system of any of claims 8 to 11, wherein the neural network model is at least one of a sliding window neural network (SWNN) or a recurrent deep neural network (DNN); and optionally, wherein the recurrent DNN includes at least one of a gateway recurring unit cell (GRU) or a short and long term memory cell (LSTM).
[13" id="c-fr-0013]
The system of claim 12, wherein the functions implemented by the processor further include functions for filtering noise from the real-time data, and the recurring DNN is trained based on the filtered real-time data.
[14" id="c-fr-0014]
14. The system as claimed in claim 13, in which the functions implemented by the processor further comprise functions for:
apply the data in real time as input to a convolutional neural network;
passing the output data comprising a previously estimated response value of the at least one operating variable through a Kalman filter; and
2017-IPM-101331-U1-FR 37 cause recurrent DNN based on an output from the convolutional neural network and an output from the Kalman filter.
[15" id="c-fr-0015]
15. Computer readable storage medium having instructions stored therein, which when executed by a computer cause the computer to implement a plurality of functions, including functions for:
acquire real-time data comprising values for a plurality of input variables associated with a stage of the instant of a drilling operation along a planned path of a wellbore within a formation underground;
training a neural network model to produce an objective function defining a response value for at least one operating variable to be optimized during the drilling operation along the planned path, based on the acquired real time data;
estimate the response value for at least one operating variable, based on the objective function produced by the trained neural network model;
applying a stochastic optimization to the estimated response value in order to produce an optimized response value for the at least one operating variable;
estimating the values of a plurality of controllable parameters for a later stage of the drilling operation, based on the optimized response value of the at least one operating variable; and implementing the later stage of the drilling operation based on the estimated values of the plurality of controllable parameters.

类似技术:

---

公开号 | 公开日 | 专利标题

---

FR3070180A1|2019-02-22|NEURON NETWORK MODELS FOR REAL-TIME OPTIMIZATION OF DRILLING PARAMETERS DURING DRILLING OPERATIONS

---

FR3027339A1|2016-04-22|

---

FR3027049A1|2016-04-15|PREDICTION OF DOWNHOLE TOOL DEFECT INDUCED BY TEMPERATURE CYCLE

---

FR3081026A1|2019-11-15|LEARNING-BASED BAYESIAN Optimization to Optimize CONTROLLED Drilling Parameters

---

Payette et al.2015|Mitigating drilling dysfunction using a drilling advisory system: results from recent field applications

---

FR3067745A1|2018-12-21|IMPROVED WAVEFORM ANALYSIS FOR TARGET BOREHOLE WAVE MODES

---

FR3085053A1|2020-02-21|PREDICTION OF OIL TANK BEHAVIOR USING A PROXY FLOW MODEL TECHNICAL AREA

---

FR3038408A1|2017-01-06|

---

FR3077142A1|2019-07-26|OPTIMIZING A REFERENCE LENGTH FOR SIGNAL PRESERVATION AND PROCESSING A REFERENCE LENGTH FOR DISTRIBUTED VIBRATION DETECTION PURPOSES

---

FR3069929A1|2019-02-08|PREDICTION IN FRONT OF TREPAN USING VERTICAL SEISMIC PROFILE AND GLOBAL INVERSION

---

FR3031131A1|2016-07-01|REAL-TIME PERFORMANCE ANALYZER FOR DRILLING OPERATIONS

---

FR3036428A1|2016-11-25|

---

FR3057604B1|2019-11-22|IMPROVED MCI DIAGRAPHY FOR THE TREATMENT OF WELL BOTTOM MEASUREMENTS

---

FR3085404A1|2020-03-06|AUTOMATED OPTIMIZATION OF THE PENETRATION RATE FOR DRILLING

---

FR3058456A1|2018-05-11|AUTOMATED DOWNHOLE FLUID CLASSIFICATION USING MAIN SPECTROSCOPY COMPONENT DATA

---

FR3086779A1|2020-04-03|MATCHING AN AUTOMATED PRODUCTION HISTORY USING BAYESIAN OPTIMIZATION

---

FR3062674A1|2018-08-10|DISTANCE UP TO MULTI-LAYER BASE LIMIT | WITH MULTIPLE INITIAL SUPPOSED VALUES

---

FR3036429A1|2016-11-25|

---

FR3059108A1|2018-05-25|AUTOMATIC MULTI-ZONE HORIZON TRACKING

---

FR3058455A1|2018-05-11|UNIVERSAL DOWNHOLE FLUID ANALYZER WITH GENERIC INPUTS

---

FR3040509A1|2017-03-03|

---

US20220003108A1|2022-01-06|Event prediction using state-space mapping during drilling operations

---

FR3076316A1|2019-07-05|EFFECTIVE REPRESENTATION OF COMPLEX THREE-DIMENSIONAL SIMULATION RESULTS FOR REAL-TIME OPERATIONS

---

FR3034232A1|2016-09-30|

---

FR3085077A1|2020-02-21|HYBRID MODELS BASED ON PHYSICS AND AUTOMATIC LEARNING FOR TANK SIMULATIONS

同族专利:

---

公开号 | 公开日

---

CA3069299A1|2019-02-28|

---

GB2578700A|2020-05-20|

---

US20210148213A1|2021-05-20|

---

NO20200053A1|2020-01-16|

---

AU2017428353A1|2020-01-30|

---

WO2019040091A1|2019-02-28|

---

GB202000648D0|2020-03-04|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

US20040256152A1|2003-03-31|2004-12-23|Baker Hughes Incorporated|Real-time drilling optimization based on MWD dynamic measurements|

---

US9587478B2|2011-06-07|2017-03-07|Smith International, Inc.|Optimization of dynamically changing downhole tool settings|

---

US20150081222A1|2013-09-19|2015-03-19|Sas Institute Inc.|Control variable determination to maximize a drilling rate of penetration|

---

US20170017883A1|2015-07-13|2017-01-19|Conocophillips Company|Ensemble based decision making|

---

US9022140B2|2012-10-31|2015-05-05|Resource Energy Solutions Inc.|Methods and systems for improved drilling operations using real-time and historical drilling data|

---

WO2015060810A1|2013-10-21|2015-04-30|Halliburton Energy Services, Inc.|Drilling automation using stochastic optimal control|

---

US10577876B2|2015-07-13|2020-03-03|Halliburton Energy Services, Inc.|Estimating drilling fluid properties and the uncertainties thereof|WO2019147297A1|2018-01-29|2019-08-01|Landmark Graphics Corporation|Controlling range constraints for real-time drilling|

---

WO2021040829A1|2019-08-23|2021-03-04|Landmark Graphics Corporation|Active reinforcement learning for drilling optimization and automation|

---

WO2021040774A1|2019-08-23|2021-03-04|Landmark Graphics Corporation|Wellbore trajectory control using reservoir property projection and optimization|

---

CA3102561A1|2020-02-03|2021-08-03|Landmark Graphics Corporation|Event prediction using state-space mapping during drilling operations|

---

GB2595549A|2020-03-26|2021-12-01|Landmark Graphics Corp|Physical parameter projection for wellbore drilling|

---

US11149510B1|2020-06-03|2021-10-19|Saudi Arabian Oil Company|Freeing a stuck pipe from a wellbore|

法律状态:
2019-07-30| PLFP| Fee payment|Year of fee payment: 2 |

---

2019-10-18| PLSC| Search report ready|Effective date: 20191018 |

---

2021-03-26| RX| Complete rejection|Effective date: 20210215 |

优先权:

---

申请号 | 申请日 | 专利标题

---

US201762548274P| true| 2017-08-21|2017-08-21|

---

US62548274|2017-08-21|

---

PCT/US2017/061849|WO2019040091A1|2017-08-21|2017-11-15|Neural network models for real-time optimization of drilling parameters during drilling operations|

---