SYSTEMS AND METHODS FOR CALIBRATING WELL-COMPLETION TECHNIQUES

STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

Not applicable.

BACKGROUND

Embodiments disclosed herein generally relate to wellbore designs and wellbore completion operations. More particularly, embodiments disclosed herein relate to systems and methods for executing a frac packing operation, including predicting a screen-out condition in an earthen formation through which a wellbore extends.

Wellbores are drilled into subterranean earthen formations to facilitate the recovery of hydrocarbons from reservoirs within the earthen formation. In some applications, a drilled wellbore may be “completed” to enhance fluid conductivity or permeability between the wellbore and the hydrocarbon bearing reservoir, and to improve extraction of resources from the reservoir. In some completion operations, a hydraulic fracturing system is employed to initiate and propagate hydraulic fractures in the subterranean formation extending from the wellbore to enhance fluid conductivity between the wellbore and the surrounding formation. For instance, a hydraulic fracturing fluid may be pumped down the wellbore against a desired location of the subterranean formation. The fluid is pressurized to a degree sufficient to initiate one or more fractures at the location along the formation.

Other completion techniques include gravel packing or “frac packing.” Frac packing is a completion technique that combines hydraulic fracturing and gravel packing. In a gravel packing or frac packing operation, gravel is used as a filter that prevents unwanted formation material from entering into the wellbore, while allowing hydrocarbons to pass through. By using the frac pack technique, users/operators can achieve the advantages of improved production from hydraulic fracturing, while also achieving the advantage of sand control that is provided by gravel packing.

SUMMARY

In an example of the present disclosure, a system is provided for a completion operation for a well extending through a subterranean earthen formation. The system includes a surface pump configured to pressurize a fluid to a downhole net pressure measurable by a sensor package, a fluid line extending between the surface pump and a wellhead positioned at an upper end of the well, where the fluid line is configured to flow the fluid into the well, and a monitoring system in signal communication with the sensor package and comprising a reinforcement learning (RL) frac packing module stored in a memory of the monitoring system, wherein the RL frac packing module is configured to a) perform an initial simulation based on geological data of the well and an input parameter, where the initial simulation provides simulated net pressure values as a function of time for the well; b) receive an indication of an actual net pressure value in the well; c) adjust the input parameter based on a difference between the actual net pressure value and a corresponding simulated net pressure value; d) perform an updated simulation based on the geological data of the well and the adjusted input parameter, where the updated simulation provides updated simulated net pressure values as a function of time for the well; e) iteratively adjust the input parameter by repeating step c) and step d), with the corresponding simulated net pressure value being from the updated simulated net pressure values; and f) provide an indication of an event at the well based on the actual net pressure value and the corresponding simulated net pressure value. The event at the well a tip screen-out event in some examples, and the indication is provided responsive to the difference between the actual net pressure value and the corresponding simulated net pressure value being less than a first threshold amount, and a slope of actual net pressure values deviating from a slope of simulated net pressure values by more than a second threshold amount.

In another example of the present disclosure, a method is provided for a completion operation of a well. The method includes a) performing, by a simulator, an initial simulation based on geological data of the well and an input parameter, where the initial simulation provides simulated net pressure values as a function of time for the well; b) receiving an indication of an actual net pressure value in the well; c) adjusting, by a reinforcement learning (RL) agent, the input parameter to the simulator based on a difference between the actual net pressure value and a corresponding simulated net pressure value; d) performing, by the simulator, an updated simulation based on the geological data of the well and the adjusted input parameter, where the updated simulation provides updated simulated net pressure values as a function of time for the well; e) iteratively adjusting the input parameter to the simulator by repeating step c) and step d), with the corresponding simulated net pressure value being from the updated simulated net pressure values; and f) providing an indication of an event at the well based on the actual net pressure value and the corresponding simulated net pressure value. The event at the well a tip screen-out event in some examples, and the indication is provided responsive to the difference between the actual net pressure value and the corresponding simulated net pressure value being less than a first threshold amount, and a slope of actual net pressure values deviating from a slope of simulated net pressure values by more than a second threshold amount.

In yet another example of the present disclosure, a non-transitory machine-readable medium contains instructions that, when executed by a processor, cause the processor to a) perform an initial simulation based on geological data of a well extending through a subterranean earthen formation and based on an input parameter, where the initial simulation provides simulated net pressure values as a function of time for the well; b) receive an indication of an actual net pressure value in the well; c) adjust the input parameter based on a difference between the actual net pressure value and a corresponding simulated net pressure value; d) perform an updated simulation based on the geological data of the well and the adjusted input parameter, where the updated simulation provides updated simulated net pressure values as a function of time for the well; e) iteratively adjust the input parameter by repeating step c) and step d), with the corresponding simulated net pressure value being from the updated simulated net pressure values; and f) provide an indication of an event at the well based on the actual net pressure value and the corresponding simulated net pressure value. The event at the well a tip screen-out event in some examples, and the indication is provided responsive to the difference between the actual net pressure value and the corresponding simulated net pressure value being less than a first threshold amount, and a slope of actual net pressure values deviating from a slope of simulated net pressure values by more than a second threshold amount.

Embodiments described herein comprise a combination of features and characteristics intended to address various shortcomings associated with certain prior devices, systems, and methods. The foregoing has outlined rather broadly the features and technical characteristics of the disclosed embodiments in order that the detailed description that follows may be better understood. The various characteristics and features described above, as well as others, will be readily apparent to those skilled in the art upon reading the following detailed description, and by referring to the accompanying drawings. It should be appreciated that the conception and the specific embodiments disclosed may be readily utilized as a basis for modifying or designing other structures for carrying out the same purposes as the disclosed embodiments. It should also be realized that such equivalent constructions do not depart from the spirit and scope of the principles disclosed herein.

BRIEF DESCRIPTION OF THE DRAWINGS

For a detailed description of various exemplary embodiments, reference will now be made to the accompanying drawings in which:

FIG. 1 is a schematic view of a system for enhancing fluid connectivity between a well and a subterranean earthen formation in accordance with the principles disclosed herein;

FIG. 2 is a schematic view of an embodiment of a dendritic fracture system formed by the system of FIG. 1 and extending from the well of FIG. 1;

FIG. 3 is a schematic view of an example of training a reinforcement learning frac packing module in accordance with the principles disclosed herein;

FIG. 4 is a flowchart of an embodiment of a method for calibrating a frac pack simulator during a completion operation of a well extending through a subterranean formation in accordance with the principles disclosed herein;

FIG. 5 is a graph of exemplary simulated net pressure values and actual net pressure values provided by one or more embodiments in accordance with the principles disclosed herein; and

FIG. 6 is a block diagram of an embodiment of a computer system in accordance with the principles disclosed herein.

DETAILED DESCRIPTION

The following discussion is directed to various exemplary embodiments. However, one skilled in the art will understand that the examples disclosed herein have broad application, and that the discussion of any embodiment is meant only to be exemplary of that embodiment, and not intended to suggest that the scope of the disclosure, including the claims, is limited to that embodiment.

Certain terms are used throughout the following description and claims to refer to particular features or components. As one skilled in the art will appreciate, different persons may refer to the same feature or component by different names. This document does not intend to distinguish between components or features that differ in name but not function. The drawing figures are not necessarily to scale. Certain features and components herein may be shown exaggerated in scale or in somewhat schematic form and some details of conventional elements may not be shown in interest of clarity and conciseness.

In the following discussion and in the claims, the terms “including” and “comprising” are used in an open-ended fashion, and thus should be interpreted to mean “including, but not limited to . . . .” Also, the term “couple” or “couples” is intended to mean either an indirect or direct connection. Thus, if a first device couples to a second device, that connection may be through a direct connection of the two devices, or through an indirect connection that is established via other devices, components, nodes, and connections. In addition, as used herein, the terms “axial” and “axially” generally mean along or parallel to a particular axis (e.g., central axis of a body or a port), while the terms “radial” and “radially” generally mean perpendicular to a particular axis. For instance, an axial distance refers to a distance measured along or parallel to the axis, and a radial distance means a distance measured perpendicular to the axis. Any reference to up or down in the description and the claims is made for purposes of clarity, with “up”, “upper”, “upwardly”, “uphole”, or “upstream” meaning toward the surface of the borehole and with “down”, “lower”, “downwardly”, “downhole”, or “downstream” meaning toward the terminal end of the borehole, regardless of the borehole orientation. As used herein, the terms “approximately,” “about,” “substantially,” and the like mean within 10% (i.e., plus or minus 10%) of the recited value. Thus, for example, a recited angle of “about 80 degrees” refers to an angle ranging from 72 degrees to 88 degrees.

The completion operations of this disclosure are generally described with reference to hydrocarbon wells. However, such completion operations may also be applied to geothermal energy extraction examples, as well as carbon-capture-utilization-storage (CCUS) well examples. The scope of the present disclosure is not intended to be limited to a particular type of well unless explicitly stated.

The present disclosure relates generally to calibrating well-completion techniques and, more specifically, to calibrating fracturing-and-gravel packing (or frac packing) models by using reinforcement learning.

During a frac packing process, a viscous fracturing fluid is pumped into the well to create fractures. Subsequently, a mixture of fracturing fluid and proppant, such as sand, is pumped into the well to prevent the hydraulic fractures from closing following the conclusion of the hydraulic fracturing operation. During the frac packing process, a pressure signal is referred to as net pressure, which is an additional pressure over a fracture closure pressure that is useful to propagate and pack a fracture with proppant. As proppant enters the periphery of the fracture, the proppant bridges off and inhibits additional fracture growth. A tip screen-out (TSO) event refers to the point at which the proppant inhibits additional fracture growth, and such TSO event is characterized by a distinct change (e.g., an increase) in the net pressure slope as a function of time. Following the TSO event, additional proppant that enters the fracture may inflate the fracture and results in a net pressure gain. The design and implementation of a frac packing plan involves managing an amount of net pressure gain after the TSO event, through decisions such as adding or subtracting proppant at the wellbore surface, reducing a pump rate for packing inside the well bore, and the like, which improves the likelihood of a relatively high-quality frac packing job, or frac pack.

Currently, users attempt to achieve a high-quality frac pack (e.g., attempt to achieve an appropriate net pressure gain after the TSO event) by manually performing the aforementioned management decisions in order to arrive at the desired pressure signature. In order to manually perform these management decisions, the operators/users generally need to rely on their intuition and upon the output readings of a third-party simulator. Specifically, currently, the timing of when and how to stop filling in the proppant is performed manually by experienced operators/engineers. These experienced operators/engineers discuss and come to an agreement on an action depending on personal prior experience. The net pressure gain that is needed to maintain good connectivity between the earthen formation and the wellbore is dependent on many factors.

Accordingly, it is difficult to properly control a frac packing job. First, human operators relying on intuition can be inconsistent from one job to the next. Second, frac packing jobs may occur over the course of a 36-hour timeline, or longer in some cases, which subjects the human operators to fatigue that may further degrade their decision-making abilities. Finally, human operators with expertise in these well-completion techniques are relatively scarce.

Embodiments disclosed herein address the foregoing by providing a frac packing tool or module that enables operators to make improved decisions at appropriate times, which facilitates efficient, consistent, and improved frac packing operations. As described further below, the frac packing module includes a reinforcement learning (RL) agent. The RL agent refers to computer-implemented functionality, such as a software program, and may be considered similar or equivalent to a computer-implemented RL system, a computer-implemented RL engine, a computer-implemented RL “brain”, and the like. The RL agent is not as affected by human biases, and thus enables faster, more efficient frac packing operations. The RL agent may also increase or maintain safety levels, while reducing operator fatigue.

The frac packing module also includes a simulator that operates in conjunction with the RL agent. Accordingly, the combination of the simulator and the RL agent is referred to herein as an RL frac packing module, for relative brevity. In some embodiments, the RL agent applies a model-free, deep RL approach in order to improve and/or manage calibration of the simulator. In some embodiments, the calibration of the simulator occurs in real time, or in near-real time.

In some embodiments, an impending TSO event may be predicted or identified using actual net pressure data (e.g., from a downhole gauge, or determined from a surface pressure gauge based on the hydrostatic pressure of the fluid column in the wellbore) and simulated net pressure data from the simulator. For example, the simulator is configured to determine (and display) simulations of net pressure as a function of time for a given set of configurable input parameters, and for a geological profile of a given well (e.g., from logging data for the given well). The simulator is thus configured to provide simulated net pressure data that is predicted for the well. The RL agent is configured to receive actual net pressure data, as well as the simulated net pressure data from the simulator. Based on the actual net pressure data and the simulated net pressure data, the RL agent may iteratively adjust the configurable input parameter(s) of the simulator (e.g., in real time) in an attempt to improve the accuracy of the simulated net pressure data provided by the simulator. The simulator then updates the simulated net pressure data based on the adjusted configurable input parameter(s) from the RL agent, and thus may display updated simulations of net pressure that more closely reflect (e.g., more closely match) the actual net pressure data for the well. The updated simulations can then be analyzed to determine whether a TSO event has occurred.

In some embodiments, the RL frac packing module is configured to provide an alarm indicating the impending TSO event to an operator of a system for a well completion operation, or of a fracturing system performing part of the well completion operation, allowing the operator to perform one or more actions to achieve a desired net pressure gain after the TSO event, which may result in an improved frac pack. For example, in response to receiving an alarm indicating a predicted TSO event (or otherwise identifying the TSO event), the operator may adjust one or more properties of the fluid (e.g., fracturing fluid), such as a fluid flow rate, a surface pressure, introducing or reducing proppant, etc., of the fluid to better achieve the desired net pressure gain following the TSO event. Although the description below specifically refers to frac packing, one of ordinary skill will understand that certain embodiments of the present disclosure may be applied to various types of sand- or proppant-laden pumping processes during completion operations. For example, the simulator tool can be used to simulate the evaluations of net pressure traces including a tip screen-out, a magnitude of the net pressure increase, and the like. These and other examples are described more fully below, with reference made to the accompanying figures.

Referring now to FIGS. 1 and 2, an embodiment of a hydraulic fracturing system 100 is shown. Fracturing system 100 is generally configured to enhance fluid communication or connectivity between a wellbore or well 102 and a hydrocarbon bearing reservoir disposed in a subterranean earthen formation 10 (shown in FIG. 2) through which well 102 extends. In other words, the earthen formation 10 is disposed about the well 102. As will be discussed in more detail below, fracturing system 100 is used to initiate and propagate fractures in the earthen formation 10 extending radially from well 102 in response to the communication of pressurized hydraulic fracturing fluids into well 102. Such fractures in earthen formation 10 increase fluid connectivity between the well 102 and the targeted subterranean reservoir within earthen formation 10. Thus, fracturing system 100 is employed to perform hydraulic fracturing operations, including frac packing operations and in particular, fracturing operations designed to increase the complexity of fractures formed in the earthen formation 10 to enhance fluid connectivity between the well 102 and the targeted reservoir within earthen formation 10.

As best shown in FIG. 1, fracturing system 100 may generally include a fracturing fluid supply conduit or manifold 104, a plurality of fluid storage tanks 106, a hydration unit 108, a blending unit 110, a fracturing manifold 114, a plurality of high pressure surface pumps 116, a high pressure fracturing line 118, a frac tree 120 (including a flow cross), a wellhead 122, a flowback line 124, and a flowback tank 126. Storage tanks 106 and units 108 and 110 are in series fluid communication with conduit 104. In the embodiment shown in FIG. 1, fluid storage tanks 106 store a base fluid (e.g., water) that is routed through the hydration unit 108 where chemicals, such as e.g., gelling agents, are added. After sufficient hydration time to allow for adequate mixing in hydration unit 108, the mixed fluid may travel to blending unit 110 and be blended with predetermined quantities of chemicals (e.g., acids, surfactants, gels, emulsifiers, viscosity reducers, friction reducers, etc.) and/or diverting agents provided by blending unit 110. In this embodiment, blending unit 110 supplies hydraulic fracturing system 100 with a permanent or non-dissolvable/non-biodegradable diverting agent, such as proppant (e.g., sand, etc.). Examples of suitable permanent diverting agents include, without limitation, 100 mesh proppant, walnut hulls, large grain size proppants (e.g., 8-16 to 40-70 mesh proppants), or any other particulate which remains in the reservoir. The permanent diverting agent(s) provided by blending unit 110 to the mixed fluid are designed to assist in preventing fracturing fluid from leaking into the earthen formation 10 from a fracture formed therein, and, as will be discussed further herein.

Referring still to FIG. 1, the blended hydraulic fracturing fluid provided by blending unit 110 is supplied to fracturing manifold 114, where the fracturing fluid may be routed through the plurality of surface pumps 116 to pressurize the fracturing fluid to a fracturing pressure sufficient to initiate or form one or more fractures in the earthen formation 10 through which the well 102 extends. In this embodiment, each fracturing pump 116 comprises a positive displacement pump powered by a power source such as a diesel engine, a gas turbine, or other devices known in the art; however, in other embodiments, the configuration of each fracturing pump 116 may vary. Accordingly, in at least this embodiment, shutting down or de-energizing the power source of each pump 116 holds each pump 116 stationary such that back pressure may be held within the well 102.

Following pressurization via surface pumps 116, the fracturing fluid may be routed from manifold 114 through high pressure fracturing line 118 into frac tree 120. Thus, high pressure fracturing line 118 extends and provides fluid communication between manifold 114 and frac tree 120. Frac tree 120 manages the flow and communication of fluid between well 102 and the components (e.g., high pressure fracturing line 118, flowback line 124, etc.) of fracturing system 100. In this embodiment, frac tree 120 comprises a flow cross and is coupled with high pressure fracturing line 118, flowback line 124, and wellhead 122; however, in other embodiments, the configuration of frac tree 120 may vary. Fluid communication between frac tree 120 and well 102 is provided by wellhead 122 disposed at the upper end of well 102 (at the surface). Wellhead 122 provides physical support for frac tree 120 as well as components of fracturing system 100 that extend into well 102, including, for example, a casing string (not shown in FIG. 1). The casing string provides structural support to well 102 and controls fluid communication between well 102 and the surrounding earthen formation 10 through which well 102 extends. Although well 102 is a cased well in this embodiment, in general, system 100 can be used in connection with cased or uncased wells.

In the arrangement described above and shown in FIG. 1, pressurized fracturing fluid is communicated from high pressure fracturing line 118, through frac tree 120, and into well 102 via wellhead 122. As will be discussed further herein, hydraulic fracturing system 100 also routes flowback fluid received from well 102 following the injection of pressurized fracturing fluid into well 102. Particularly, flowback fluid from well 102 may be routed through wellhead 122, frac tree 120, and into flowback tank 126 via flowback line 124 for storage in tank 126. In this embodiment, frac tree 120 includes a frac valve 128 for isolating high pressure fracturing line 118 when flowback fluid is flowed from well 102 into flowback tank 126 via flowback line 124. Additionally, frac tree 120 includes a flowback valve 130 for isolating flowback tank 126 when high pressure fracturing fluid is pumped into well 102 from high pressure fracturing line 118; however, in other embodiments, the configuration of frac tree 120 may vary.

Referring still to FIG. 1, high pressure fracturing line 118 includes a surface sensor package or assembly 132 to measure and indicate the fluid pressure of the fracturing fluid within high pressure fracturing line 118. In this embodiment, fluid injection or flow rate of pressurized fracturing fluid through high pressure fracturing line 118 may be determined by parameters pertaining to surface pumps 116 (i.e., speed, amount of fluid displaced each period, etc.), via surface sensor package 132. In general, the embodiments described herein are not limited to obtaining pressure data from a particular location, and should be interpreted to encompass examples in which net pressure is determined from one or more sensors downhole, one or more sensors at the surface, or combinations thereof.

In this embodiment, fracturing system 100 further includes a monitoring system 134 for monitoring various parameters of well 102 and equipment of fracturing system 100. Monitoring system 134 is in signal communication with surface sensor package 132 (and/or one or more downhole sensors, not shown in FIG. 1 for simplicity) and other equipment of fracturing system 100. Monitoring system 134 is also in communication with an input/output (I/O) unit 136 (e.g., a graphical user interface, a touchscreen interface, or the like) for displaying information to an operator of fracturing system 100 and for receiving user inputs from the operator. The I/O unit 136 may display information related to the operation of fracturing system 100 and may receive user inputs related to operation of fracturing system 100. During operations, I/O unit 136 may communicate received user inputs to monitoring system 134, including the viscosity of the fracturing fluid supplied to well 102, and mesh size and density of the diverting agent provided by blending unit 110. Communication between the I/O unit 136 and monitoring system 134 may be wired, wireless, or a combination thereof.

Monitoring system 134 comprises any suitable device or assembly which is capable of receiving electrical (or other data) signals and transmitting electrical (or other data) signals to other devices. In particular, while not specifically shown, monitoring system 134 may comprise a computer system including a processor and a memory. The processor (e.g., microprocessor, central processing unit, or collection of such processor devices, etc.) may execute machine readable instructions (e.g., non-transitory machine readable medium) provided on the corresponding memory to provide the processor with all of the functionality described herein. The memory of monitoring system 134 may comprise volatile storage (e.g., random access memory), non-volatile storage (e.g., flash storage, read only memory, etc.), or combinations of both volatile and non-volatile storage. Data consumed or produced by the machine readable instructions can also be stored on the memory of monitoring system 134. An RL frac packing module is stored in the memory of monitoring system 134 and is executed by the processor of monitoring system 134. As will be described further herein, the RL frac packing module is generally configured to provide an alarm or another indication of an impending TSO event in an earthen formation (e.g., earthen formation 10), during the performance of a hydraulic fracturing operation. This enables the operator to perform one or more actions to achieve a desired net pressure gain after the TSO event, which may result in an improved frac pack. Although in this embodiment monitoring system 134 comprises a component of the fracturing system 100 shown in FIG. 1, in other embodiments, monitoring system 134 may be utilized in fracturing systems which vary in configuration from fracturing system 100.

Hydraulic fracturing system 100 may be employed via methods discussed further herein to form a fracture system within earthen formation 10, as shown schematically in FIG. 2. Particularly, FIG. 2 schematically illustrates a horizontally extending (relative to the surface), subsurface portion of well 102 and the surrounding earthen formation 10 through which well 102 extends. Although in this embodiment well 102 of hydraulic fracturing system 100 includes a horizontal or deviated section, in other embodiments, well 102 may comprise a substantially vertical well.

As shown particularly in FIG. 2, by injecting fracturing fluid into well 102 at a sufficient pressure, fracturing system 100 may form a pair of primary fractures 152 within earthen formation 10. In this embodiment, fractures 152 extend from a generally cylindrical surface 153 of the well 102 in a bi-wing configuration. Fractures 152 generally comprise initial hydraulic (i.e., not natural) fractures formed in earthen formation 10 during a hydraulic fracturing operation (e.g., a frac packing operation) performed by the hydraulic fracturing system 100 shown in FIG. 1.

Referring now to FIG. 3, a schematic example of training a RL frac packing module 305 is shown. As described above, certain embodiments improve and/or the well-completion process using reinforcement learning. Certain embodiments of the present invention integrate a reinforcement learning system (e.g., RL agent) with a third-party simulator for frac packs.

Reinforcement learning seeks to train an agent, which “learns” how to provide optimization guidance or make independent decisions based on interactions with a simulator. The agent “learning” is based on receiving either a reward or a penalty based on its previous recommended action (e.g., adjusting an input parameter for the simulator) and the resulting simulator output that is provided to the agent.

In the example of FIG. 3, the RL frac packing module 305 is represented by a simulator 310 (i.e., the simulator of an RL methodology) and an RL agent 320 (i.e., the agent of the RL methodology). The simulator 310 and the RL agent 320 may be connected by an application programming interface (API), such as the Python API. In some examples, final results of calibration of the simulator 310 (e.g., when simulator 310 input parameter(s) are provided by the RL agent 320 such that simulated net pressure values are within a threshold of actual net pressure values) may be provided to a display coupled to the RL frac packing module 305.

Also, in FIG. 3, actions 330 represents adjustments to simulator 310 input parameter(s) that are implemented by the RL agent 320, such as to iteratively configure the simulator 310 based on actual net pressure values. In FIG. 3, state block 340 represents a comparison of simulated net pressure values (P_{net_pred}) from the simulator 310 to actual net pressure values (P_{net_field}), represented by or received as field data 350. As described above, the field data 350 may be from a downhole gauge, or determined from a surface pressure gauge based on the hydrostatic pressure of the fluid column in the wellbore 102. In this example, the state block 340 determines the difference between actual net pressure value(s) (from the field data 350) and simulated net pressure value(s) (from the simulator 310) as a mean squared error (MSE) calculation, and provides the result of the MSE calculation (e.g., MSE(P_{net_pred}−P_{net_field})) back to the RL agent 320.

Generally, FIG. 3 demonstrates a methodology that includes training a reinforcement learning system (e.g., the RL agent 320) to optimize or improve the calibration of the simulator 310 (e.g., while the simulator 310 is providing simulated net pressure values for a frac pack operation), in real time (or near-real time) during the frac pack operation.

The RL agent 320 may be trained to configure the simulator 310 such that the simulated net pressure values generated by the simulator 310 match, or are within a threshold amount of, the measured net pressure values from the well 102, such as from field data 350. As described above, measured net pressure values may be determined from field data 350, such as that received from a sensor package, which may include a downhole pressure gauge, a surface pressure sensor, or combinations thereof. The field data 350 may also include bottom hole proppant concentration data, surface tubing and annulus pressure data, data related to a volume of proppant remaining at the surface, washpipe pressure data, or combinations thereof.

In some embodiments, the RL agent 320 iteratively configures the simulator 310 (e.g., in real time or near-real time), such as by modifying or otherwise adjusting simulator 310 input parameter(s) related to the well 102 (represented by actions 330). For example, the simulator 310 input parameter(s) may include a modulus configuration parameter, a toughness configuration parameter, a stress configuration parameter, and/or a leakoff coefficient configuration parameter for the well 102. As described above, the simulator 310 may be considered to be properly configured when the simulated net pressure values match, or are within a threshold amount of, the actual (e.g., measured) net pressure values. Accordingly, the RL agent 320 configures the simulator 310 in real time, or in near-real time.

The RL agent 320 may be configured to be trained using a reward function. For example, the reward function may be such that the RL agent 320 is rewarded for configurations (of the simulator 310 input parameter(s)) that result in simulated net pressure values that are closer to the actual, measured net pressure values 350 from the well 102.

In an embodiment, the simulator 310 is configured to perform an initial simulation, which may be based on geological data of the well 102 (e.g., from one or more logging operations performed for the well 102) and one or more input parameters. As described above, the input parameter(s) may include a modulus, a toughness, a stress, and/or a leakoff coefficient of the well 102. The initial simulation performed by the simulator 310 provides simulated net pressure values (P_{net_pred}), which may be provided as a function of time for the well 102.

The state block 340 receives the simulated net pressure values from the simulator 310, and also receives the actual net pressure values of the well 102 (P_{net_field}) from the field data 350. The state block 340 compares the simulated net pressure values with the actual net pressure values, and provides an indication of a difference between the simulated net pressure values and the actual net pressure values. For example, the indication of the difference is a result of an MSE calculation. The MSE calculation is then provided as an input to the RL agent 320.

The RL agent 320 is configured to adjust the input parameter(s) to the simulator 310, as actions 330 to the simulator 310, based on the indication of difference received from the state block 340. For example, the RL agent 320 adjusts the input parameter(s) to the simulator 310 such that the simulator 310 provides simulated net pressure values that are closer to the actual net pressure values received as field data 350.

The simulator 310 is configured to perform an updated simulation based on the already-determined geological data of the well 102 as well as the adjusted input parameter(s) provided as actions 330 by the RL agent 320. The updated simulation provides updated simulated net pressure values (P_{net_pred}) as a function of time for the well 102.

The above-described process continues iteratively (e.g., in a loop fashion as shown in FIG. 3) until the simulated net pressure values from the simulator 310 are within a threshold amount of the actual net pressure values indicated by the field data 350. For example, with each iteration, the RL agent 320 configures the simulator 310 by adjusting input parameter(s) using certain actions 330. Subsequently, the simulator 310 generates an updated simulated net pressure (P_net,pred), and the state block 340 performs an MSE calculation based on the updated simulated net pressure values and the actual net pressure values indicated by the field data 350, the result of which MSE calculation is provided back into the RL agent 320, for further training of the RL agent 320 to more accurately configure the simulator 310.

In some embodiments, the state block 340 particularly provides an indication of the difference (e.g., the result of an MSE calculation) between an actual net pressure value, indicated by the field data 350, and a corresponding simulated net pressure value from the simulator 310. A corresponding simulated net pressure value may refer to a simulated net pressure value that is simulated for (e.g., predicted) a particular time that is approximately the same as the time associated with the actual net pressure value. For example, field data 350 provides actual net pressure values beginning at time 0, and every 5 seconds thereafter. In this example, corresponding simulated net pressure values may be the net pressure values provided by the simulator 310 for the same time 0, and every 5 seconds thereafter. In this way, the state block 340 compares actual net pressure values with net pressure values that are simulated for approximately the same time, which enables the RL agent 320 to be further trained based on differences between the actual net pressure values from the field data 350 and corresponding simulated net pressure values from the simulator 310.

After the RL agent 320 iteratively configures the simulator 310 to provide simulated net pressure values that match (or are within a threshold of) the actual readings of net pressure, the RL frac packing module 305 may be configured to provide an indication of a TSO event. For example, a TSO event may be indicated when the simulated net pressure values are sufficiently close to the actual net pressure values (e.g., the difference between the simulated and actual net pressure values is less than a first threshold amount), and a slope (e.g., a rate of change) of the actual net pressure values deviates from a slope of the simulated net pressure values by more than a second threshold amount. As described above, the TSO event refers to the point at which the proppant inhibits additional fracture growth, and such TSO event is characterized by a distinct change (e.g., an increase) in the net pressure slope as a function of time. In the example of FIG. 3, such distinct change in slope of the actual net pressure values is determined by its deviation from (e.g., by more than the second threshold amount) the simulated net pressure values from the simulator 310.

Following the TSO event, additional proppant that enters the fracture may inflate the fracture and results in a net pressure gain. In some examples, a frac packing plan is designed and implemented to manage an amount of net pressure gain after the TSO event, through decisions such as adding or subtracting proppant at the wellbore surface, reducing a pump rate for packing inside the well bore, and the like, which improves the likelihood of a relatively high-quality frac packing job, or frac pack.

In some examples, the RL frac packing module 305 is thus further configured to provide a recommendation to an operator to adjust the pump rate, surface pressure, and/or proppant volume to achieve a particular or desired net pressure gain following the TSO event. For example, the RL frac packing module 305 may display a simulation of net pressure (e.g., from the simulator 310), which enables the operator to determine what steps may be taken in order to achieve the desired amount of net pressure gain after the TSO event.

The recommendation from the RL frac packing module 305 may pertain to parameters useful to control a well-completion technique, such as a frac packing operation. By adjusting proppant volume, changing the pump rate, and/or adjusting a choke openness, an operator may achieve a desired net pressure gain after the TSO event identified by the RL frac packing module 305.

In other examples, the RL frac packing module 305 is further configured to automatically adjust the pump rate, surface pressure, and/or proppant volume to achieve the particular or desired net pressure gain following the TSO event. For example, the RL frac packing module 305 may provide command(s) to various control interface(s) in order to control various systems in order to adjust the desired parameters to achieve the particular or desired net pressure gain following the TSO event.

The machine/reinforcement learning approach of the embodiments described herein can thus be used to improve real-time decision making by an operator, and/or to automate certain control actions in order to achieve a particular or desired net pressure gain following the TSO event. In other embodiments, a human operator may refer to generated simulations (e.g., from the simulator 310) to make an operator decision to achieve the desired net pressure gain. The RL agent 320 may be coupled to a custom user interface to enable or assist operational decisions by a frac pack engineer.

In some embodiments, the RL agent 320 may be trained prior to being deployed at, or used in conjunction with, a first well. For example, the RL agent 320 may be trained for a particular geographic region, and then subsequently deployed on other wells in that geographic region. In on example, a second well is in the same geographic region as the first well, and the RL agent 320 is trained using data collected from a previously performed well completion operation for the second well. Subsequently, the RL agent 320 may be more quickly trained for the first well, due to its having been previously trained on the second well in the same geographic region. The RL agent 320 may also be deployed on other wells in the same region.

In some embodiments, the RL agent 320 is configured to achieve an improved, more accurate net pressure gain (e.g., in terms of pounds/ft²of proppant formation) and a void-free annular pack in order to ensure good connectivity to the reservoir. The RL agent 320 may also thus reduce issues or problems related to pumping operations during the frac pack.

In some embodiments, the RL agent 320 is trained in a simulation environment, such as prior to being deployed at, or used in conjunction with, a first well. For example, the simulation environment may act as the learning environment for the RL agent 320. Using model-free RL, the RL agent 320 may be configured to learn (e.g., be trained) in a simulated environment, in which the RL agent 320 is observes (e.g., receives data related to) the environment after every action (e.g., adjusting input parameter(s) as in block 330 in FIG. 3) and further continues to adjust the various input parameter(s) based on receiving a “reward” (e.g., which reward may be based on feedback such as that from state block 340). In some embodiments, the simulator 310 may implement a StimPlan™ simulation model, such the RL agent 320 may be configured to optimizes or improve calibration coefficients of the simulation model implemented by the simulator 310. For example, one or more embodiments of the RL agent 320 may be configured to optimize or improve calibration coefficients in a StimPlan™ simulation.

In some embodiments, the trained model implemented by the RL agent 320 may be tested and/or otherwise validated in a simulation environment, and may also be benchmarked against historical frac packing operations data (e.g., data from previously-completed frac packs).

Turning now to FIG. 4, a method 400 for calibrating a frac pack simulator during a completion operation is shown in further detail. In particular, the method 400 may be useful to calibrate the RL frac packing module 305 described above, including training the RL agent 320 thereof. Accordingly, the method 400 is useful for well 102 extending through a subterranean formation.

The method 400 begins in block 402 with performing an initial simulation based on geological data of the well 102 and an input parameter. The initial simulation provides simulated net pressure values as a function of time for the well 102. As described above, the simulator 310 is configured to perform an initial simulation, which may be based on geological data of the well 102 (e.g., from one or more logging operations performed for the well 102) and one or more input parameters. The input parameter(s) may include a modulus, a toughness, a stress, and/or a leakoff coefficient of the well 102. The initial simulation performed by the simulator 310 provides simulated net pressure values (P_{net_pred}), which may be provided as a function of time for the well 102.

The method 400 continues in block 404 with receiving an indication of an actual net pressure value in the well 102. As described above, field data 350 may be indicative of the actual net pressure values, and the RL frac packing module 305 (e.g., the state block 340) is configured to receive the field data 350 and to determine an actual net pressure value associated therewith.

The method 400 then continues in block 406 with adjusting the input parameter to the simulator 310 based on a difference between the actual net pressure value and a corresponding simulated net pressure value (e.g., as determined by the state block 340). The difference may include the result of an MSE calculation between the actual net pressure value and the corresponding simulated net pressure value. As described above, actions 330 represent adjustments to simulator 310 input parameter(s) that are implemented by the RL agent 320, such as to iteratively configure the simulator 310 based on actual net pressure values. Thus, based on a comparison of simulated net pressure values (P_{net_pred}) from the simulator 310 to actual net pressure values (P_{net_field}), represented by or received as field data 350, the RL agent 320 is configured to adjust the input parameter(s) to the simulator 310, as actions 330 to the simulator 310. For example, the RL agent 320 adjusts the input parameter(s) to the simulator 310 such that the simulator 310 provides simulated net pressure values that are closer to the actual net pressure values received as field data 350.

The method 400 continues further in block 408 with performing an updated simulation based on the geological data of the well 102 and the adjusted input parameter. The updated simulation provides updated simulated net pressure values as a function of time for the well 102. As described above, the simulator 310 is configured to perform an updated simulation based on the already-determined geological data of the well 102 as well as the adjusted input parameter(s) provided as actions 330 by the RL agent 320. The updated simulation provides updated simulated net pressure values (P_{net_pred}) as a function of time for the well 102.

In block 410, the method 400 iteratively adjusts the input parameter to the simulator 310 by repeating step 406 and step 408, with the corresponding simulated net pressure value being from the updated simulated net pressure values. As described above with respect to FIG. 3, the process continues in a loop fashion until the simulated net pressure values from the simulator 310 are within a threshold amount of the actual net pressure values indicated by the field data 350. For example, with each iteration, the RL agent 320 configures the simulator 310 by adjusting input parameter(s) using certain actions 330. Subsequently, the simulator 310 generates an updated simulated net pressure (P_net,pred), and the state block 340 performs an MSE calculation based on the updated simulated net pressure values and the actual net pressure values indicated by the field data 350, the result of which MSE calculation is provided back into the RL agent 320, for further training of the RL agent 320 to more accurately configure the simulator 310.

Finally, the method 400 continues in block 412 with providing an indication of an event at the well based on the actual net pressure value and the corresponding simulated net pressure value. One example of such a well event is a TSO event, as described above. In this example, the indication of the TSO event is provided responsive to the difference between the actual net pressure value and the corresponding simulated net pressure value being less than a first threshold amount, and a slope of actual net pressure values deviating from a slope of simulated net pressure values by more than a second threshold amount. For example, a TSO event may be indicated when the simulated net pressure values are sufficiently close to the actual net pressure values (e.g., the difference between the simulated and actual net pressure values is less than a first threshold amount), and a slope (e.g., a rate of change) of the actual net pressure values deviates from a slope of the simulated net pressure values by more than a second threshold amount.

FIG. 5 is a graph 500 of exemplary simulated net pressure values and actual net pressure values provided by one or more embodiments described herein. In the graph 500, simulated net pressure values as a function of time (e.g., from the simulator 310) are represented by 510, while actual net pressure values as a function of time (e.g., from a pressure sensor) are represented by 520. As described above, the RL agent 320 implements a reinforcement learning approach to iteratively configure the simulator 310, such that, over time, the simulated net pressure values 510 more closely match the actual net pressure values 520.

As the RL agent 320 iteratively configures the simulator 310 based on the actual net pressure values 520, the simulated net pressure 510 begins to converge with the actual net pressure values 520. Presenting the graph 500 to an operator (e.g., which graph 500 may be updated over time according to the iterative reinforcement learning process described above) may allow the operator to ascertain whether a TSO event has occurred. In the example of FIG. 5, a TSO event occurs at time 530 (e.g., at approximately 20 minutes elapsed in the graph 500). As described above, the TSO event is characterized by a distinct change (e.g., increase) in slope of the actual net pressure 520, particularly relative to the simulated net pressure 510.

In this example, the indication of the TSO event is provided responsive to the difference between the actual net pressure value 520 and the corresponding simulated net pressure value 510 being less than a first threshold amount, and a slope of actual net pressure values 520 deviating from a slope of simulated net pressure values 510 by more than a second threshold amount. At time 530, such deviation occurs because the simulated net pressure values 510 were sufficiently close to the actual net pressure values 520 (e.g., indicating an approximately accurate calibration of input parameter(s) of the simulator 310), and the slope of the actual net pressure values 520 increased relative to the simulated net pressure values 510. As described above, the “closeness” or accuracy of the simulated net pressure values 510 relative to the actual net pressure values 520 may be determined relative to a first threshold value. Also, the deviation of the slope of the actual net pressure values 520 relative to the simulated net pressure values 510 may be relative to a second threshold value.

At the time 530 of the TSO event, an operator may introduce or reduce an amount of proppant to affect the net pressure gain following the time 530 of the TSO event. In the example of FIG. 5, the desired gain in net pressure (e.g., following the TSO event at 530) is approximately 500 psi (i.e., the difference between about 1200 psi and about 700 psi). Other examples may have different amounts of desired gain in net pressure. By achieving the desired net pressure gain after the TSO event at 530, the operators may achieve a more high-quality frac pack.

Following the TSO event at 530, an increasing amount of proppant is illustrated by 540. As described above, during the frac packing process, fluid is pumped first followed by proppant. The increasing amount of proppant 540 can be pumped after the TSO event at 530. As the amount of proppant 540 increases, the actual net pressure 520 builds as a function of time.

In some scenarios, as the actual net pressure 520 increases, the increased pressure can also lead to a “well bore screen out,” which may indicate that the frac packing is complete. Keeping various volume and pressure states into consideration, and viewing the graph 500 (and/or responding to recommendations or automated actions as described above) the operator is able to manage ceasing to provide (e.g., fill with) proppant at an appropriate time prior to a time at which the well bore screen out would otherwise occur.

The foregoing embodiments may reduce pumping problems—such as by improving safety, reducing cost, and reducing problems associated with deferred production—due to unmanaged screen outs during fracking operations, including frac packing operations. Accordingly, the foregoing embodiments may help achieve a higher near-wellbore proppant concentration, which enhances fluid connectivity to the reservoir.

For example, certain embodiments disclosed herein may increase a success rate for creating high-conductivity fractures in an ultra-high permeability environment. Certain embodiments may also increase a success rate for proppant placement, where high fluid leak-off rates are evident.

In other low-carbon scenarios, the disclosed embodiments may apply where there is a need to drill wells into reservoirs, and may increase volume of CO2 stored in such reservoirs. In a geothermal energy production scenario, the disclosed embodiments may enable drilling of relatively deep wells, while increasing or maximizing the surface area for heat transfer.

Referring to FIG. 6, an embodiment of a computer system 600 suitable for implementing one or more embodiments disclosed herein is shown. For example, the monitoring system 134 shown in FIG. 1 may be configured in a manner similar to the computer system 600 shown in FIG. 6. The computer system 600 includes a processor 602 (which may be referred to as a central processor unit or CPU) that is in communication with one or more memory devices 604, and input/output (I/O) devices 606. The processor 602 may be implemented as one or more CPU chips. The memory devices 604 of computer system 600 may include secondary storage (e.g., one or more disk drives, etc.), a non-volatile memory device such as read only memory (ROM), and a volatile memory device such as random access memory (RAM). In some contexts, the secondary storage ROM, and/or RAM comprising the memory devices 604 of computer system 600 may be referred to as a non-transitory computer readable medium or a computer readable storage media. I/O devices 326 may include printers, video monitors, liquid crystal displays (LCDs), touch screen displays, keyboards, keypads, switches, dials, mice, and/or other well-known input devices.

It is understood that by programming and/or loading executable instructions onto the computer system 600, at least one of the CPU 602, the memory devices 604 are changed, transforming the computer system 600 in part into a particular machine or apparatus having the novel functionality taught by the present disclosure. Additionally, after the computer system 600 is turned on or booted, the CPU 602 may execute a computer program or application. For example, the CPU 602 may execute software or firmware stored in the memory devices 604. The software stored in the memory devices 604 and executed by CPU 602 may comprise the RL frac packing module 305 shown in FIG. 3. During execution, an application may load instructions into the CPU 602, for example load some of the instructions of the application into a cache of the CPU 602. In some contexts, an application that is executed may be said to configure the CPU 602 to do something, e.g., to configure the CPU 602 to perform the function or functions promoted by the subject application. When the CPU 602 is configured in this way by the application, the CPU 602 becomes a specific purpose computer or a specific purpose machine.

While exemplary embodiments have been shown and described, modifications thereof can be made by one skilled in the art without departing from the scope or teachings herein. The embodiments described herein are exemplary only and are not limiting. Many variations and modifications of the systems, apparatus, and processes described herein are possible and are within the scope of the disclosure. For example, the relative dimensions of various parts, the materials from which the various parts are made, and other parameters can be varied. Accordingly, the scope of protection is not limited to the embodiments described herein, but is only limited by the claims that follow, the scope of which shall include all equivalents of the subject matter of the claims. Unless expressly stated otherwise, the steps in a method claim may be performed in any order. The recitation of identifiers such as (a), (b), (c) or (1), (2), (3) before steps in a method claim are not intended to and do not specify a particular order to the steps, but rather are used to simplify subsequent reference to such steps.

SYSTEMS AND METHODS FOR CALIBRATING WELL-COMPLETION TECHNIQUES

Information

Publication Number

Date Filed

Date Published

Inventors

Original Assignees

CPC

International Classifications

Abstract

Description

Claims

CROSS-REFERENCE TO RELATED APPLICATIONS

Provisional Applications (1)