Wellbore Fracturing Treatment Based on Real-Time Modeling

The oil and gas industry uses well stimulation techniques to increase the transfer of hydrocarbon resources from a reservoir formation to a wellbore. Such stimulation typically relies on the introduction of a pressurized fracturing fluid into a wellbore. The pressurized fracturing fluid generates fractures downhole in the reservoir formation. As part of the process, a flow network, sometimes referred to as “frac iron,” is constructed between one or more pumps and a wellhead of a borehole. The flow network provides a path to deliver the pressurized fracturing fluid to the borehole so the fracturing fluid may be used to generate and propagate fractures in the reservoir formation.

Like reference numbers and designations in the various drawings indicate like elements.

The description that follows includes example systems, methods, techniques, and program flows that embody embodiments of the disclosure. Unless otherwise specified, use of the terms “connect,” “engage,” “couple,” “attach,” or any other like term describing an interaction between elements is not meant to limit the interaction to a direct interaction between the elements and may also include an indirect interaction between the elements described. Unless otherwise specified, use of the terms “up,” “upper,” “upward,” “uphole,” “upstream,” or other like terms shall be construed as generally away from the bottom, terminal end of a well; likewise, use of the terms “down,” “lower,” “downward,” “downhole,” or other like terms shall be construed as generally toward the bottom, terminal end of the well, regardless of the wellbore orientation. Use of any one or more of the foregoing terms shall not be construed as denoting positions along a perfectly vertical axis. In some instances, a part near the end of the well can be horizontal or even slightly directed upwards. Unless otherwise specified, use of the term “subterranean formation” shall be construed as encompassing both areas below exposed earth and areas below earth covered by water such as ocean or fresh water.

Hydraulic fracturing is a form of energy transfer. In one example approach, the energy transfer initiates with hydraulic horsepower (via positive displacement pumps) that injects a unit volume of an incompressible fluid into the formation via one or more wellheads. In one such example approach, the incompressible fluid includes a certain volume fraction of proppant. The injection process applies energy through compression to convert a low-pressure volume to a high-pressure state. Surface energy consumption is defined as the integration of the horsepower deployed over time. Integrating the horsepower deployed over time provides total surface energy consumption for the entire hydraulic pumping duration.

It can be difficult to balance formation development costs versus formation production. It may not be enough to simply look at the total slurry and fluid volume delivered to a formation without taking into account the dynamic conditions in which fracturing jobs occur. One may instead look at well stimulation as an exchange of energy for production (in any form) within the formation. In one example approach, the energy consumed on the surface may be directly correlated to fuel consumed by the hydraulic horsepower and by the horsepower operating cost. This correlation is direct given the fact that fuel and horsepower maintenance may be valued in units of energy (e.g., million British thermal units (MMBtu) or megawatt-hour (MWh)). The effective energy delivered to formation, however, is different from surface energy consumed since the unit volume of slurry that is pumped from the surface down the wellbore, past the perforations and into formation undergoes a series of energy losses and energy gains before reaching the formation. In practice, completion related variables are often changed with no regard to the impact on total energy consumption and related cost since that relationship is not understood. Similarly, the lack of understanding of effective energy delivered to formation prevents operators from making informed decisions regarding drill space unit (DSU) production optimization.

Furthermore, as a reservoir formation becomes more fractured, the effect of energy transfer into the formation becomes difficult to predict. It may be useful to further understand how energy transferred into the reservoir formation alters the formation; such information may be especially useful in optimizing complex fracture systems that are generated during the hydraulic fracture completion of single wells, multiple wells, multiple wells on a single pad, and multiple wells from multiple pads. Understanding how this energy is being dissipated in the reservoir and the upper effective limits of this energy may enable optimization of asset development all the way from well design to pad design and then on to completion sequencing and well production. In some cases, the optimum completion design may also change depending on when a given well is completed during the sequence of operations.

Traditionally, physic-based models have been used to design hydraulic fracturing treatments. Such models have not proven effective due to the complexity of formations having two or more wellheads. Current fracture models consider predominately Mode 1 (tension) types of failure. Shale applications, however, include multiple fractures along a horizontal wellbore. There may be significant stress interference between different fractures and the levels of stress increase significantly as fractures interfere with each other. High strain levels within regions of high stress interference can cause rock failure in shear that include Mode 2 (in plane shear failure) and Mode 3 (out of plane shear) types of failure. Dilation of Mode 2 and Mode 3 fractures creates a pressure field within the multi-mode fracture system that may be detected far away from the hydraulically initiated fractures themselves.

There are similar types of interference between multi-stage horizontal wells. In multi-stage horizontal wells, the amount of stress interference between stages and between different wells can become extreme. It may no longer be a reasonable assumption to rely on Mode 1 models since while Mode 1 failure will occur, shear and compressional failure will also occur resulting in a multi-mode fracture system. The result is a complex fracture system combining Mode 1, Mode 2, and Mode 3 failures. Open fractures may appear well away from the hydraulic fracture and significant pressure fields associated with the hydraulic fracture process may be detected a significant distance away from the hydraulic fracture.

Stress is transmitted through the rock very quickly, but pressure front development or pressure dissipation may be much slower as it is controlled by fracture dilation and leak-off into initially closed fracture systems within the sheared region. Pressure response lags the strain response. Once these fractures are open, however, the pressure signal moves faster. Either way the pressure field from hydraulic fracturing dissipates throughout the complex fracture system and can be used to understand the system itself. For the first wells in a system, the pressure front is mostly behind the newest fractures. After the first wells are fractured, however, the pressures will be much more interactive and connected. Interior wells, therefore, will behave much differently than those on the outside of the reservoir formation or on the outside of a given pad of wells or multiple pads of wells.

Underground sensors such as pressure gauges and Low Frequency Distributed Acoustic Sensing (LFDAS) may be used to characterize fractures in a complex fracture system. For instance, it is possible to use external pressure gauges combined with LFDAS to identify specific distributed acoustic sensing (DAS) behaviors or patterns that are indicative of a pressure communication event in the formation. Data driven models of such fractures have been proposed. Some such models may depend on above-ground measurements of surface energy and underground measurements of stress to determine propagation of fractures in a complex fracturing system. Such models, however, are not predictive in nature and have been proven to be quite limited in enabling effective and relevant decisions in real-time. Most of the current models are look back models that provide basic information on completion designs based on historical data. They do not enable real-time responses to specific events in the completion process.

In the United States unconventional market, some operators are using a trial-and-error method to replace physics-based modelling when designing hydraulic fracturing treatments. The general assumption is that these wells and each fracturing segment will behave the same. In addition, operators rely on people to make decisions in real time when executing a treatment option. Optimally, the person making the decisions has experience in treating wellbores in the given type of formation. In reality, some people have a great deal of relevant experience while others do not, and the decisions made may not be consistent or correct in many situations.

In addition, operators may see significant variations in response between treatment segments. Very little effort has been expended, however, on understanding the cause of these variations or the results of this behavior. The trial-and-error method is time consuming and provides a relatively slow means of making decisions due to a slow feedback loop; with well performance or production the trial-and-error method may take several months to get the desired feedback.

Other operators are using data driven models when designing hydraulic fracturing treatments. In most cases, data driven models only tell the operator what he or she already knows, or what he or she should already know. In shale fracturing, data driven models tend to tell operators that what they are currently doing is best and it is hard to move past this when designing new hydraulic fracturing treatments.

A more predictive data-driven model accelerates the learning curve and reduces the risk of the costly repetition of mistakes. If used properly, this data-driven model may be supported with physics-based solutions to help make treatment changes beneficial both to current pumping operations and to longer-term well production performance.

In one example approach, rather than attempting to create a model for an entire hydraulic fracturing treatment, each hydraulic fracturing treatment is broken down into specific segments and models for each segment are trained using pattern recognition and machine learning. The behaviors observed for each segment are the used to create higher level models that are more robust in nature. Advanced pattern recognition techniques may be used to make multiple predictions during a single treatment and machine learning techniques may be used to make predictions based on those observations.

In one such example approach, formation breakdown behavior (the hydraulic efficiency or the time required to reach design injection rate) may be modeled based on automated breakdown techniques used to provide a consistent procedure to establish flow rate based on pressure response. This consistent procedure means that different breakdown behaviors may be identified, and effective treatment outcomes predicted. In one such example approach, actions recommended based on the observed response may be used to improve the probability of achieving the desired result or the desired outcome.

To date, no one has taken a data driven approach to characterizing specific behaviors and responses at each segment of the treatment process. The ability to characterize such behaviors and responses may, however, not only enable actions to be taken at each segment of an operation, but also may predict what to expect on subsequent treatment segments. For example, observations of pressure response after breakdown may be used to provide suggestions for later in the treatment. Observations of pressure response when proppant hits perforations for proppant sweeps or main proppant schedule, observations of pressure response as proppant concentrations are increased during proppant segments, observations of proppant pulse behavior during operations (e.g., during quick step-down tests or during diverter cycles) and observation of pressure response when the diverter reaches the perforations may be used to characterize performance at each segment of treatment. Similarly, observations of water hammer behavior, observations of offset well interference such as, for example, frac hits observed with fiber optic or offset well pressure monitoring, and observations of DAS interpretation on perforation cluster efficiency in real time may be used to characterize performance at particular segments of treatment. Finally, observations of hydraulic efficiency including time to reach target injection rate and effective energy placed into the formation may be used to characterize performance at given segments of treatment. As treatment continues through each segment, the model refines the predictions and recommendations to reflect further observations and behaviors. At the same time, the model may suggest actions to be taken by the control system at each segment based on specific treatment responses in previous segments. Actions may include performing a quick step-down test to capture pressure pulse data, perforation friction, near wellbore tortuosity and the number of open perforations taking fluid, running a procedure for alleviating wellbore friction and reducing treating pressures, running a near wellbore or far field diversion stage, running a proppant sweep, changing the high viscosity friction reducer (HVFR) concentration, changing the proppant schedule or ending the segment early.

In one example approach, this data driven approach is used to provide initial recommendations for treatment options. The selected options at each treatment segment may change, however, as the system is refined at each treatment segment. based on the observations at each segment. In one such example approach, the system captures all relevant data within a basin to train a model to recognize early indicators and to make logical decisions based on previous results observed over hundreds or thousands of previous treatments. In one example approach, the data is supplemented with information from subject matter experts to formalize a decision-making process. The result may be used by the treatment engineer to accept or decline the recommendations or, in some cases, may be forwarded directly to the system controller to implement the recommended action. The data driven technique provides a means to achieve consistent and repeatable performance for crews across the entire organization and, as experience is gained, may provide real-time decisions guiding automated treatment processes in the future.

In some implementations, a downhole operation or attribute in the wellbore may be modified or updated based on the determining whether the fracturing event might occur in the wellbore. For example, an operation (at the surface or downhole) may be performed and/directed to be performed to change a downhole operation or attribute based on whether the fracturing event might occur in the wellbore. For example, attributes of an actual fracturing operation in the wellbore may be set based on whether the fracturing event might occur. Examples of such attributes of the actual fracturing operation may include depth, composition of the proppant used for fracturing, composition of the fracturing fluid used for fracturing, the pump rate for fracturing, etc. For instance, if the fracturing event is determined to not likely to occur, any one of these attributes may be updated to increase the likelihood that the fracturing event is to occur.

Illustrative examples are given to introduce the reader to the general subject matter discussed herein and are not intended to limit the scope of the disclosed concepts. The following sections describe various additional features and examples with reference to the drawings in which like numerals indicate like elements, and directional descriptions are used to describe the illustrative aspects, but, like the illustrative aspects, should not be used to limit the present disclosure.

illustrates an example well stimulation system, according to aspects of the present disclosure. A typical well stimulation systemincludes one or more pumpsand a flow network connecting the pumpsthrough a boreholeto a reservoir formation. In operation, the flow network conveys pressurized fracturing fluid from the pumps to the reservoir formation. In general, fracturing fluids in well stimulation systems are injected into the wellbore at a high pressure in order to convey sufficient energy to a subterranean formation to cause fracturing in the formation. In some example approaches, well stimulation systems induce fluid pressures in a range of 3,000 to 20,000 pounds/square inch (psi) in the fluid injected into the wellbore with total slurry rate in the range of 10 to 200 barrels per min (bpm).

In the example shown in, well stimulation systemincludes a frac iron configurationconnected through boreholeto reservoir formation. Frac iron configurationincludes frac ironand one or more pumps. In the example approach shown in, frac ironreceives pressurized fracturing fluid pumped by the one or more pumpsfrom fluid sourceand conveys the received pressurized fluid through boreholeto reservoir formation.

In some example approaches, boreholeincludes a casingand an openingat surface. In some such example approaches, the pressurized fracturing fluid passes through an isolation valvebefore entering boreholeat opening. In some example approaches, boreholefurther includes perforationsat certain locations in reservoir formation, the pressurized fracturing fluid passing through the perforationsto cause fractures in the reservoir location.

In one example approach, well stimulation systemincludes a computer system. In one such example approach, computer systemincludes a processorand a memory. In one example approach, instructions are stored in memorythat, when executed by processor, allow the processorto control the fracturing process and to capture data representing pressure fields generated via fracturing.

In one example approach the computer systemreceives an effective energy model for energy loss in pressurized fracturing fluid passing into and through boreholeto a reservoir formationduring hydraulic fracturing. The model may be used at each segment identified for the hydraulic fracturing treatment to determine effective energy delivered to the reservoir formation as a function of surface energy added to the fracturing fluid when raising the fracturing fluid to a high-pressure state; gravitational potential energy gains in the pressurized fracturing fluid as the fluid travels down to the reservoir formation; and energy losses in the pressurized fracturing fluid as the fluid travels down to the reservoir formation. In one example approach, an operator may use computer systemto apply the effective energy model to a selected reservoir formation and may select, based on the effective energy model, an operational cost for hydraulic fracturing of the selected reservoir formation. The operator may then use the computer system to control the one or more pumpsin frac iron configurationto achieve the selected operational cost.

As noted above, surface energy consumption may be expressed as units of energy. For instance, surface energy consumption may be expressed as horsepower times hours, or horsepower hours, which may be considered a unit of energy. The surface energy used in treatment of one or more wells may be calculated by multiplying pressure, rate, and operational time for the duration of the well treatment to arrive at surface horsepower hours consumed. Surface horsepower hours consumed may also be calculated by integrating pressure with respect to volume pumped for the duration of the treatment. That is, surface energy may be determined as the energy input to elevate a unit volume of slurry (fluid containing a certain volume fraction of proppant) from a low-pressure state to a high-pressure state.

As the unit volume of slurry traverses down boreholeit is met with assistance in the form of gravitational potential energy (hydrostatic pressure) and with resistance in the form of pipe friction. Additionally, when the unit of slurry flows past perforationsand near wellbore tortuous regions, the unit undergoes additional energy losses before reaching the formation. This relationship may be expressed via a derivation of Bernoulli's equation. Bernoulli's equation also adheres to conservation of energy. As a result, the pressure losses or gains may be represented in terms of energy if the individual terms of Bernoulli's equation are integrated with respect to volume.

This then becomes the backbone of hydraulic fracturing energy analyses, where the effective energy delivered to the formation may be calculated by adding gravitational potential energy to the surface energy component and subtracting all the energy loss contributions (e.g., pipe friction and perforation friction). As noted above, the energy received at the perforationsis not all converted to Mode 1 fractures. Some is lost at the perforationsand via tortuosity/near wellbore (NWB). The remainder is dissipated as a pressure field generated via Mode 2 and Mode 3 fractures in the vicinity of the Mode 1 fractures.

By understanding this energy system and applying completion strategies and technologies to reduce system losses such as pipe friction, perforation friction, NWB losses and pressure field propagation, one may lower surface energy consumption while maintaining or improving the effective energy delivered bottomhole. As the ratio of effective energy downhole to surface energy is maximized, so is the ratio of effective energy to operational cost maximized.

In the example approach of, a predictive data driven process supplements the energy model above by identifying behavioral patterns at each segment identified for the the hydraulic fracturing treatment, by attributing probable causes for the different behavioral patterns and then by creating recommended actions when encountering each behavioral pattern. In some example approaches, the action is recommended based on one or more of previous observed results in the data, on results from physical modeling and on information from subject matter experts. By breaking up the hydraulic fracturing treatments into logical segments, it is possible to create higher level models with reduced uncertainty based on the behavioral patterns observed over multiple segments. By relating the observed behaviors of multiple observations, the model is able to make predictions with higher levels of accuracy. In addition, for real-time operations, the ability to quickly identify relevant behaviors enables fast and effective decisions to be made very quickly, even within the same pumping stage.

In one example approach, Low Frequency Distributed Acoustic Sensing (LFDAS) may be used in the presence of external pressure gauges to detect DAS behaviors or patterns that are indicative of a pressure communication event. Data captured by the LFDAS sensors and by the external pressure gauges may then be correlated using a physics-based method and/or used to train a machine learning system such as a neural network to operate solely with LFDAS sensors to detect such DAS behaviors in the absence of external pressure gauges. The machine learning process may include supervised and unsupervised learning approaches. In some example approaches, the pressure data may be processed and tagged with events of interest. In some such approaches, the events of interest include deviations from a static base line, deviations from a slowly declining pressure rate as pressure bleeds off into the formation, rate of pressure changes where different rates of change may indicate different events (e.g. an approaching fracture with a slow rate of pressure increase or a fracture intersecting a pressure gauge where the rate of pressure change can be associated with the location of the fracture, sudden drops in pressure during fracturing operations where the pressure drop may be associated with fault activation, or gradual changes to pressure rates as fractures may intersect existing fracture networks). These tags may then be used for supervised machine learning approaches. The tagged events may include events flagged and/or removed/ignored during unsupervised learning where the machine learning model development process identifies similar events and groups them in order to identify features that may then be classified by a subject matter expert and used for supervised learning approaches. These pressure events and features may then be used to train a machine learning model where, for instance, the strain data is used as the measured/applied as input data in cases where external pressure gauges may not be available or coarsely spaced. Data from fracture operations may also be included in the data set and one or more of treatment well pressure, rate, fluid chemical composition, completion parameters, target formation, microseismic data captured with Distributed Acoustic Sensing (DAS) systems/geophone/accelerometers and associated processed data, treatment well flow allocation and uniformity index using DAS, instantaneous and cumulative flow rate per cluster, formation parameters like depth/permeability/porosity/reservoir rock/temperature/pressure and geographical data like GPS position for formation specific models. This approach of building machine learning models where strain may be used as a proxy or a substitute for pressure enables the use of distributed sensing where the distributed fiber may be deployed on demand during fracturing operations.

In the example shown in, monitoring wells.and.(collectively, monitoring wells) include a borehole. Boreholeincludes LFDAS sensorsand external pressure gauges. During fracturing operations, sensorsmonitor for DAS behaviors or patterns that are indicative of a pressure communication event while pressure gaugesmonitor for the pressure communication event itself. In one example approach, the data is stored and correlated using a physics-based method and/or used to train a machine learning system such as a neural network to detect pressure communication events in adjacent wells. The use of LFDAS sensors is described in further detail in U.S. patent application Ser. No. 18/441,976, filed Feb. 14, 2024, the description of which is incorporated herein by reference.

illustrates another example approach to a well stimulation system, according to aspects of the present disclosure. As in the well stimulation systemof, well stimulation systemincludes one or more pumpsand a flow network connecting the pumpsthrough a boreholeto a reservoir formation. In operation, the flow network conveys pressurized fracturing fluid from the pumps to the reservoir formation. In general, fracturing fluids in well stimulation systems are injected into the wellbore at a high pressure in order to convey sufficient energy to a subterranean formation to cause fracturing in the formation. In some example approaches, well stimulation systems induce fluid pressures in a range of 3,000 to 20,000 pounds/square inch (psi) in the fluid injected into the wellbore with total slurry rate in the range of 10 to 200 barrels per min (bpm).

In the example shown in, well stimulation systemincludes a frac iron configurationconnected through boreholeto reservoir formation. Frac iron configurationincludes frac ironand one or more pumps. In the example approach shown in, frac ironreceives pressurized fracturing fluid pumped by the one or more pumpsfrom fluid sourceand conveys the received pressurized fluid through boreholeto reservoir formation.

In some example approaches, boreholeincludes a casingand an openingat surface. In some such example approaches, the pressurized fracturing fluid passes through an isolation valvebefore entering boreholeat opening. In some example approaches, boreholefurther includes perforationsat certain locations in reservoir formation, the pressurized fracturing fluid passing through the perforationsto cause fractures in the reservoir location.

In one example approach, well stimulation systemincludes a computer system. In one such example approach, computer systemincludes a processorand a memory. In one example approach, instructions are stored in memorythat, when executed by processor, allow the processorto control the fracturing process and to capture data representing pressure fields generated via fracturing. In one such example approach, memoryfurther includes instructions stored in memorythat, when executed by processor, allow the processorto execute a machine learning system that receives training datastored in memoryand generates a model for interpreting LFDAS sensor data to detect and measure pressure communication events as discussed above. Processorthen uses the model to process LFDAS signals received during stimulation of the reservoir formation.

In one example approach the computer systemreceives an effective energy model for energy loss in pressurized fracturing fluid passing into and through boreholeto a reservoir formationduring hydraulic fracturing. The model may be used to determine effective energy delivered to the reservoir formation as a function of surface energy added to the fracturing fluid to raise the fracturing fluid to a high-pressure state; gravitational potential energy gains in the pressurized fracturing fluid as the fluid travels down to the reservoir formation; and energy losses in the pressurized fracturing fluid as the fluid travels down to the reservoir formation. In one example approach, an operator may use computer systemto apply the effective energy model to a selected reservoir formation and may select, based on the effective energy model, an operational cost for hydraulic fracturing of the selected reservoir formation. The operator may then use the computer system to control the one or more pumpsin frac iron configurationto achieve the selected operational cost.

In the example shown in, monitoring wells.and.(collectively, monitoring wells) include a borehole. Boreholeincludes LFDAS sensors. During fracturing operations, computer systemuses sensorsto monitor DAS behaviors or patterns that are indicative of a pressure communication event. In one example approach, the LFDAS signal data is stored and analyzed by processorusing the model for interpreting LFDAS sensor data.

In one example approach, monitoring wellsequipped with LFDAS sensorsare established throughout the reservoir formationto capture information representing pressure interference between pads and to provide vital information to help optimize treatments for a new pad. Such an approach is less expensive than the combination of LFDAS sensorsand pressure gaugesdiscussed in, while presenting accurate detection of the scope of the pressure field around the fracturing event.

is a flow diagram illustrating an example method for anticipating and reacting to events that may occur during a segment of a hydraulic fracturing treatment of one or more wells, according to aspects of the present disclosure. The early detection and mitigation of adverse events during hydraulic fracturing treatment would go a long way in reducing the risks and costs of such treatments. At the same time, the early detection and encouragement of positive events at segments of hydraulic fracturing treatment further reduce the risk and cost of such treatments. The challenge in developing a useful data driven model for real time applications is to identify early, key indicators that correlate to specific outcomes (patterns). When these indicators (patterns) are identified, they may be used in real time to predict outcomes and therefore enable decisions to take preventative action. For hydraulic fracturing, real time decisions may involve 1) making specific changes to the current pumping stage to prevent an undesirable outcome; 2) making specific changes to the stages following the current pumping stage, and 3) making changes in the design of hydraulic fracturing treatments for other wells to prevent undesirable outcomes over a broader sample of stages and wells. Since these indicators may be predictive in nature, analysis of recent and historical data may be used during fracture design and planning to take preventative steps to avoid future problems based upon regional observations (patterns) from existing data.

Pattern recognition may be used to identify the patterns that lead to specific events or outcomes. Pattern recognition is the ability of machines to identify patterns in data, and then to use those patterns to make decisions or predictions using computer algorithms. The benefits of pattern recognition include identification (behavioral, structural, audio, visual patterns may be used for identification), discovery (enable thinking out of the box and seeing things not normally seen), prediction (forecasting data and making predictions), decision-making (making decisions based on reliable, data-based insights) and big-data analytics (patterns may be found in large data sets and may be used to predict future outcomes).

In one example approach, Explainable Artificial Intelligence (XAI) is used to support pattern recognition when designing hydraulic fracturing treatments. XAI is a set of processes and methods that allow human users to comprehend and trust the results and output created by machine learning algorithms. One goal of XAI is to create a suite of new or modified machine learning techniques that produce explainable models that, when combined with effective explanation techniques, enable end users to understand, appropriately trust, and effectively manage the information. Explainable AI eliminates the “Black Box” often associated with Data Analytics. It therefore allows engineers to understand why something is happening and then how to utilize this information to create the most value.

As data sets grow, it becomes more difficult to extract nuanced information. The combination of AI and pattern recognition in tools such as XAI provides a mechanism that may be used to identify patterns in data that represent early indicators, precursors or warning signs of a probable result, outcome or event. Such early indicators, precursors and warning signs may then be used to increase the probability of positive results and to decrease the probability of negative results. When applied to hydraulic fracturing treatments, a combination of AI and pattern recognition provide a mechanism for modeling each segment of the hydraulic fracturing treatment, gaining insight in how modifications to the treatment at each segment effect that segment, how those in-segment modifications effect subsequent segments in the hydraulic fracturing treatment for that well and how the in-segment modifications of the well effect hydraulic fracturing treatment of other wells in the formation. Identification of early indicators, precursors or warning signs also makes it possible to identify actions that will either increase the probability of a certain event occurring during a segment identified for the hydraulic fracturing treatment or that will decrease the probability of a certain event occurring, depending on the desired outcome.

In the example shown in, the method includes defining contributing factors that lead to a fracturing event (). In some example approaches, the events and the contributing factors are associated with one or more fracturing stages. Data analytics and pattern recognition are then applied to fracturing data to identify precursors for each contributing factor (). This early identification of precursors may be used to anticipate and react to events at each segment identified for the hydraulic fracturing treatment using probabilistic tools. The method then determines, based on the identified precursors, whether the event might occur () and proposes actions to change (positively or negatively) the likelihood of the event occurring (). In one example approach, the method further includes modifying parameters of the hydraulic fracturing treatment to compensate for the detected fracturing event.

As noted above, in one example approach, the hydraulic fracturing treatment of a well may be broken into two or more segments. In one such example approach, the segments may be the traditional segments for making decisions during hydraulic fracturing jobs. A typical hydraulic fracturing treatment might then be broken down into several sections, with each section providing certain insights based on patterns associated with the segments. While some patterns may be understood or recognized, there may be patterns that are not readily observed. It is here that pattern recognition tools may help identify new patterns that may be used to help predict the outcomes or results within a segment and across one or more hydraulic fracturing treatments.

The method described above may be used to identify patterns from fracturing data and to determine how the patterns interact to drive events and actions to mitigate events. To date, certain aspects of hydraulic fracturing treatments have been identified as potential sources of patterns in fracturing data. The patterns represent early indicators, precursors or warning signs of a probable result, outcome or event in stages of hydraulic fracturing treatment. The following segments of the hydraulic fracturing treatment have been identified as potential sources of patterns: 1) Fracture Breakdown, time to rate or hydraulic efficiency analytics, 2) Pressure Pulse analytics, 3) Quick Step-Down Test analytics. 4) Diverter response, 5) Pressure response to first proppant at perforations and later concentration changes, 6) In well fiber perforation flow distribution, 7) Offset well fiber optic and offset well pressure data for frac hit detection, 8) ISIP and pressure decline analytics, 9) Pressure swings suggesting water quality problems/FR issues, 10) Effective hydraulic fracture energy delivered to the reservoir, and 11) Potential to integrate additional data from other sources, such as early production data, production data, smart drill bit data, and drilling data. Patterns, once identified, may then be analyzed to determine how they are connected or related.

Segments identified across different fracturing treatments will be discussed next.illustrate segments of hydraulic fracturing treatment across different wells within a common basin and events that occurred during the respective hydraulic fracturing treatments, according to aspects of the present disclosure. In the examples shown in, treating pressureis pressure at the borehole. Slurry proppant concentrationand Borehole slurry proppant concentrationare proppant concentrations in pounds/gallon. Slurry rateis in barrels per minute (bpm). Calculated borehole pressureand backside pressureare in psi. In the examples shown in, Borehole slurry proppant concentrationlags slurry proppant concentration.

As can be seen in, fracture breakdown (segment) occurs in an early segment of hydraulic fracturing treatment. Fracture automation provides an opportunity to perform analytics on breakdown behavior because the pumping process is consistent. In one example approach, patterns in breakdown behavior are used to create a predictive model that may be used immediately in the breakdown segmentand in subsequent segments. Parameters identified include perforation cluster efficiency and near wellbore tortuosity and complexity. As seen in, patterns detected in the breakdown segmentare used to create a first order model that adds immediate value and that enables immediate recommendations/actions. In one example approach, potential corrective actions include using fine particulates to hold fractures open in areas that normal proppant cannot reach, using 100 mesh sweeps to reduce tortuosity, adjusting proppant ramp. altering perforation strategy to help achieve improved distribution and performing a pressure pulse or quick step-down test.

Pressure Pulse, Quick Step-Down (,and), Diverter Response () and Rate Modulation provide other opportunities for evaluating wellbore conditions. Unlike the physics driven models used to calculate parameters, this model relates observed pressure behavior to a probable result: treating pressure/injection rate, step down pressure behavior, pressure pulse behavior or diverter response. Patterns recognized here are used to further strengthen the model for fracture breakdown and to reduce uncertainty. Anomalies are also useful in characterizing hydraulic treatment behavior. In the example shown in, inclement weather impacted treatment at segment.

The data driven model described above becomes more robust with a higher degree of certainty with additional information. In some example approaches, for instance, physics models are used to help understand the data driven relationships identified in pressure pulse test and analysis and quick step-down test and analysis. Probable outcomes for these stages include expected pressure response as proppant reaches perforations, early screen out and well productivity performance.