Systems and methods for real-time c factor prediction in hydraulic fracturing operations

By integrating machine learning with physics-informed data, the system generates cumulative compliance values for real-time fracture quality assessment and optimization, addressing the limitations of existing hydraulic fracturing systems.

US20260366409A1Pending Publication Date: 2026-11-19COOK DAVID CAMERON
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
US19/537518
Authority / Receiving Office
US · United States
Patent Type
Applications(United States)
Current Assignee / Owner
Priority Date
2025-02-13
Filing Date
2026-02-12
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Existing hydraulic fracturing systems lack real-time, adaptive, and comprehensive analysis capabilities, failing to accurately predict fracture behavior, integrate complex data streams, and provide interpretable insights, leading to suboptimal operations and environmental impacts.

Method used

Integrate machine learning techniques with physics-informed formation data to generate cumulative compliance values that characterize actual fracture system quality, enabling real-time fracture quality assessment, production forecasting, and operational optimization based on measured formation response.

Benefits of technology

Enables timely and accurate fracture behavior analysis, optimizing operations for efficiency and sustainability by processing complex data streams, adapting to changing conditions, and providing interpretable predictions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US20260366409A1-D00000_ABST
    Figure US20260366409A1-D00000_ABST
Patent Text Reader

Abstract

The present disclosure provides systems and methods for evaluating fracture system compliance associated with a hydraulic stimulation. Stimulation data comprising pressure data and injection rate data associated with the hydraulic stimulation are received from one or more data sources. A plurality of compliance-related values characterizing pressure response behavior of a fracture system are determined from the stimulation data over respective portions of the stimulation. The plurality of compliance-related values are integrated over a defined interval corresponding to a stimulation stage to generate a cumulative compliance metric representing a fracture system state associated with the stimulation stage. The cumulative compliance metric may be used to infer fracture behavior, stimulation effectiveness, or reservoir interaction, and may be generated and used during a stimulation, after completion of a stimulation, or using historical stimulation data.
Need to check novelty before this filing date? Find Prior Art

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Patent Application No. 63 / 758,322, filed Feb. 13, 2025, entitled “Systems and Methods for Real-Time C Factor Prediction in Hydraulic Fracturing Operations” the entirety of which is incorporated herein by referenceFIELD OF INVENTION

[0002] The present disclosure relates to hydraulic fracturing operations, and more particularly to systems and methods for real-time prediction and analysis of compliance factors during hydraulic fracturing using machine learning techniques.BACKGROUND

[0003] Hydraulic fracturing, commonly known as fracking, has become an integral part of the oil and gas industry, enabling the extraction of hydrocarbons from previously inaccessible reservoirs. This technique involves injecting high-pressure fluid into a wellbore to create cracks in deep-rock formations, allowing oil and natural gas to flow more freely. While fracking has revolutionized energy production, it also presents significant challenges in terms of optimization, efficiency, and environmental impact.

[0004] Traditionally, the analysis of fracturing operations has relied heavily on post-job evaluations and limited real-time data. This approach, while providing valuable insights, suffers from significant limitations. The inability to consistently interpret fracture system behavior from available data often results in suboptimal resource utilization and reduced well productivity. Moreover, the complex and dynamic nature of subsurface formations makes it challenging to predict fracture behavior accurately using conventional methods. For example, prior art systems that employ resistance-inductance-capacitance (RLC) models to invert parameters and estimate dominant fracture dimensions, such as fracture length, based on measured signals like pressure or slurry rates (e.g., as described in U.S. Pat. No. 12,037,883), may provide estimates of fracture geometry but fail to distinguish whether fracture growth—particularly in height—is extending into productive rock intervals or unproductive zones, such as depleted areas or offset wells. In contrast, methods that assess formation compliance—the relationship between injection rate changes and resulting pressure responses—have the potential to more directly characterize fracture system quality and responsiveness. However, prior art compliance measurement techniques, such as step-rate tests or diagnostic injection tests, typically provide only discrete, isolated compliance measurements at specific operational points and do not accumulate or integrate compliance behavior over the course of an entire stimulation stage, or take into account geological, reservoir or wellbore differences between one compliance measurement and another. These discrete measurements fail to capture the cumulative fracture system response that develops throughout a fracturing operation, limiting their utility for real-time operational optimization and production forecasting.

[0005] Prior art systems have attempted to address these challenges through various means, including the use of step-rate tests and pressure falloff analysis. However, these methods often provide only discrete, point-in-time measurements fail to capture the cumulative and evolving behavior of fracture systems over a stimulation stage. Furthermore, they typically require interruptions in the fracturing process, leading to operational inefficiencies and increased costs.

[0006] Another significant shortcoming of existing approaches is their limited ability to differentiate between beneficial fracture growth and undesirable fluid diversion. For instance, a decrease in fracture treating pressure could indicate either positive vertical growth into new productive intervals or problematic communication with depleted zones or offset wells or run away perforation erosion. Increase in pressure could indicate proppant hold up or frac containment, or many other system variables. In another example is the Nolte plot or net pressure plot, the classic frac diagnostic plot, but it needs to be calibrated against fracture minimum stress, a value that changes while fracking. The inability to distinguish between these scenarios in real-time can lead to misinterpretation of data and suboptimal decision-making during the fracturing operation.

[0007] Prior work by the present inventor addresses related aspects of hydraulic fracturing analysis and optimization using machine learning techniques. U.S. Pat. No. 12,270,286 discloses systems for training machine learning models on physics-based simulation outputs, including fracture propagation, rock deformation, and reservoir pressure predictions—to enable real-time prediction of optimal production parameters and operational adjustments without requiring execution of computationally expensive simulation models during active fracturing operations. This approach enables rapid optimization based on predicted fracture behavior. U.S. Patent Application Publication No. 2025 / 0252238 discloses systems for generating three-dimensional reservoir geometry visualizations using machine learning-based data flagging and isosurface rendering techniques, providing operators with spatial representations of fracture characteristics and reservoir properties. While these systems offer valuable capabilities for operational optimization and geometric visualization, they share a common limitation: neither simulation-based parameter predictions nor geometric characterizations directly measure actual fracture system effectiveness during stimulation or provide metrics that correlate with production outcomes. Predicted optimal parameters from simulation models represent what should theoretically occur, not what is actually occurring in the formation. Geometric visualizations show where fractures may extend, but do not characterize fracture quality or formation acceptance. There remains a need for real-time measurement and characterization of actual fracture system response during active stimulation operations, where the measured response metric demonstrates predictive correlation with production potential, enabling operators to assess fracture effectiveness and make operational adjustments based on observed formation behavior rather than simulated predictions or geometric estimates alone.

[0008] The present disclosure addresses the aforementioned limitations by providing systems and methods for real-time compliance monitoring during hydraulic fracturing operations. The disclosed systems generate cumulative compliance values that characterize actual fracture system quality throughout stimulation operations, where the cumulative compliance metric demonstrates predictive correlation with production outcomes. By integrating machine learning techniques with physics-informed formation data, the disclosed systems enable real-time fracture quality assessment, production forecasting during active operations, and operational optimization based on measured formation response rather than geometric predictions or simulation-based estimates.

[0009] The advent of advanced sensing technologies and increased data acquisition capabilities has opened new possibilities for more comprehensive fracture analysis. However, the sheer volume and complexity of the data generated during a fracturing operation present their own challenges. Many existing systems struggle to process and interpret this data stream effectively, often resulting in information overload for operators and delayed insights without producing a physically meaningful, stage-level metric that can be compared across wells or stimulations.

[0010] Furthermore, the heterogeneous nature of reservoir formations and the variability in fracturing fluid properties make it difficult to develop universally applicable models for predicting fracture behavior. Each well and formation may exhibit unique characteristics that require tailored analysis and interpretation. Prior art systems often rely on simplified models or generalized assumptions that fail to account for these site-specific nuances, leading to inaccurate predictions and suboptimal fracturing designs.

[0011] The integration of distributed acoustic sensing (DAS) and distributed temperature sensing (DTS) technologies has provided valuable additional data streams for fracture analysis. However, many existing systems lack the capability to effectively correlate this information with traditional pressure and rate data in real-time, missing opportunities for more comprehensive fracture characterization and optimization.

[0012] Another critical limitation of prior art systems is their inability to adapt to changing conditions during the fracturing operation. Reservoir properties, fluid behavior, and fracture propagation patterns may evolve over the course of a job, requiring analytical approaches capable of accommodating changing conditions and variable data characteristics. Static or slowly adapting systems often fail to capture these dynamic changes, leading to increasingly inaccurate assessments as the operation progresses.

[0013] The oil and gas industry has long recognized the potential of machine learning and artificial intelligence in improving various aspects of operations. However, the application of these technologies to historical or real-time fracture analysis has been limited. Many existing systems either lack the sophisticated algorithms necessary for meaningful real-time predictions or suffer from the need for exotic sensors that render their insights uneconomical or less actionable during the critical moments of a fracturing operation.

[0014] Moreover, the interpretability of machine learning models in the context of fracturing operations remains a significant challenge. Many “black box” approaches, while potentially accurate, fail to provide the transparency and explainability required for operators to trust and act upon their predictions confidently. This lack of interpretability can lead to hesitation in implementing AI-driven recommendations, particularly in high-stakes decisions during fracturing operations.

[0015] The industry also faces challenges in effectively integrating real-time fracture analysis with broader well and field management strategies. Many existing systems operate in isolation, failing to consider the broader context of reservoir management, production optimization, and long-term field development plans. This siloed approach can lead to suboptimal decision-making and missed opportunities for holistic optimization.

[0016] In light of these limitations, there is a clear and pressing need for advanced systems and methods capable of providing timely and accurate insights into fracture behavior during hydraulic fracturing operations. Such systems must be able to process and interpret complex, high-velocity data streams, adapt to changing conditions, and provide predictions that are both accurate and interpretable. Furthermore, they should be capable of integrating multiple data sources, including traditional pressure and rate measurements, MWD, offset monitoring or possibly advanced sensing technologies like DAS and DTS, though the use of exotic sensors like DTS and DAS make them cost prohibitive when designing AI models for frac the engineer must consider that the data used to train the model must be available for real-time inference on the day of the frac if a sensor provides valuable insight, it might help you extract learnings from your other sensors, but if it costs hundreds of thousands or millions of dollars to run it will not be run on every well and so it cannot be used as a training parameter in the model.

[0017] The development of such systems would not only enhance the efficiency and effectiveness of individual fracturing operations but also contribute to broader goals of resource conservation, environmental protection, and sustainable energy production. By enabling more precise control over fracture propagation and fluid placement, these advanced systems could help minimize the risk of environmental impacts while maximizing hydrocarbon recovery.

[0018] The limitations of prior art systems in providing real-time, adaptive, and comprehensive analysis of fracturing operations underscore the need for innovative approaches. The integration of advanced data processing techniques, machine learning algorithms, and domain-specific knowledge holds the promise of transforming hydraulic fracturing from a largely reactive process to a proactive, optimized operation. The development of such systems represents a significant opportunity to address the longstanding challenges in fracture analysis and optimization, potentially revolutionizing the field of hydraulic fracturing and contributing to more efficient and sustainable energy production to help meet our growing need for energy to power more advanced artificial intelligence systems. Accordingly, there exists a need for techniques that derive physically meaningful measures of fracture system response from stimulation data and that may be applied during a stimulation, after completion of a stimulation, or using historical data, without reliance on predefined test procedures or fixed analytical models.SUMMARY

[0019] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0020] According to aspects of the present disclosure, systems and methods are provided for evaluating fracture system compliance associated with a hydraulic stimulation. Stimulation data comprising pressure data and injection rate data may be received from one or more data sources associated with the stimulation. A plurality of compliance-related values characterizing pressure response behavior of a fracture system may be determined from the stimulation data over respective portions of the stimulation. In one embodiment, the method includes receiving, by at least one processor, real-time data from at least one sensor during a hydraulic fracturing operation, wherein the real-time data comprises injection rate measurements and pressure measurements. The method includes detecting, by the at least one processor, a rate change event in the hydraulic fracturing operation based on the injection rate measurements. In response to detecting the rate change event, the method includes calculating, by the at least one processor, a change in injection rate (ΔQ) associated with the rate change event, fitting, by the at least one processor, a mathematical function to a subset of the pressure measurements following the rate change event, determining, by the at least one processor, a fitted pressure change (ΔPfit) based on the mathematical function, and calculating, by the at least one processor, a compliance value for the rate change event as a ratio of the change in injection rate to the fitted pressure change (ΔQ / ΔPfit). The method further includes integrating, by the at least one processor, a plurality of calculated compliance values over a time period of the hydraulic fracturing operation to generate a cumulative compliance factor, outputting, by the at least one processor, the cumulative compliance factor to a display device, and adjusting, by the at least one processor, at least one operational parameter of the hydraulic fracturing operation based on the cumulative compliance factor, wherein the at least one operational parameter comprises at least one of: injection rate, proppant concentration, or fluid composition. For example, the cumulative compliance factor may be generated by computing the time integral of dQ / dP (or its discrete approximation ΔQ / ΔP) over the fracturing operation, providing a measure of total fracture system responsiveness.

[0021] The plurality of compliance-related values may be integrated over a defined interval corresponding to a stimulation stage to generate a cumulative compliance metric representing a fracture system state associated with the stimulation stage. The cumulative compliance metric may be used to infer fracture behavior, stimulation effectiveness, reservoir interaction, or combinations thereof. The cumulative compliance metric may be generated and used during a stimulation, after completion of a stimulation, or using historical stimulation data. According to other aspects of the present disclosure, the method may include one or more of the following features. The method may include training a machine learning model using historical hydraulic fracturing data, wherein the historical hydraulic fracturing data comprises injection rates, pressure measurements, and subterranean data, and corresponding compliance values, and generating, by the at least one processor using the trained machine learning model, a real-time predicted compliance factor based on current injection rate and pressure measurements, wherein the real-time predicted compliance factor is generated without requiring discrete rate change events. The method may include outputting, by the at least one processor, the real-time predicted compliance factor to the display device, and adjusting, by the at least one processor, the at least one operational parameter of the hydraulic fracturing operation based on the real-time predicted compliance factor. The mathematical function may comprise an exponential function. The method may include identifying, by the at least one processor, a plurality of events in the real-time data, wherein the plurality of events comprise rate step up events and rate step down events, and generating, by the at least one processor, an event table comprising a plurality of entries, wherein each entry corresponds to an identified event and includes an event type indicator, a start index corresponding to a beginning of the identified event, and an end index corresponding to an end of the identified event, wherein the rate change event is detected based on the generated event table. The method may include iteratively fitting a pressure response model to the subset of pressure measurements, wherein the iterative fitting comprises fitting an initial function to the subset of pressure measurements, calculating an error for the initial function fit, and if the error value is not within a predetermined threshold, executing an iterative fitting process comprising modifying a size of an analysis window containing the subset of pressure measurements, fitting a new function to the modified analysis window, calculating a new error value for the new function fit, and repeating these steps until either the error value meets the predetermined threshold or a maximum number of iterations is reached, selecting a final mathematical function fit based on the iterative fitting process, and determining the fitted pressure change (ΔPfit) based on parameters of the final mathematical function fit. The method may include calculating, by the at least one processor, an event-based alpha value for each rate change event, wherein the event-based alpha value is computed as a ratio of the fitted pressure change (ΔPfit) to the change in injection rate (ΔQ), integrating, by the at least one processor, a plurality of calculated event-based alpha values over the time period of the hydraulic fracturing operation to generate a cumulative alpha factor, wherein the machine learning model is further trained using the event-based alpha values, and wherein the real-time predicted compliance factor and a real-time predicted alpha factor are generated simultaneously by the trained machine learning model, and wherein the method includes calculating an instantaneous shut-in pressure (ISIP) alpha value when an ISIP event is detected in the real-time data, the ISIP alpha value being computed as a ratio of a pressure change during the ISIP event to a corresponding change in injection rate. The method may include filtering, by the at least one processor, the calculated compliance values to exclude compliance values greater than a predetermined threshold, and binning, by the at least one processor, the filtered compliance values and corresponding leak-off data by hydraulic fracturing stage, wherein the predetermined threshold is less than 50, and wherein the binning of compliance values and leak-off data by stage facilitates stage-specific analysis of fracture behavior and fluid efficiency during the hydraulic fracturing operation. The method may include generating, by the at least one processor, a summary statistics report for the hydraulic fracturing operation, wherein the summary statistics report includes a total number of rate step up events processed, a total number of rate step down events processed, and a total number of instantaneous shut-in pressure (ISIP) events processed, wherein the summary statistics report provides an overview of the frequency and distribution of different event types throughout the hydraulic fracturing operation, and wherein the summary statistics report is used to assess the overall behavior of the reservoir and the effectiveness of the fracturing process. The method may include analyzing, by the at least one processor, frequency components of the pressure measurements using a Fourier transform to identify dominant frequencies associated with rock breakage during the hydraulic fracturing operation, correlating, by the at least one processor, the identified dominant frequencies with geological characteristics of the formation being fractured, and adjusting, by the at least one processor, at least one of injection rate or fluid composition to target specific frequency ranges associated with optimal fracture propagation in the formation. The method may include integrating, by the at least one processor, distributed acoustic sensing (DAS) data from a fiber optic cable deployed in the wellbore with the injection rate and pressure measurements, identifying, by the at least one processor using the trained machine learning model, locations of fluid entry into the formation based on patterns in the DAS data correlated with changes in the real-time predicted compliance factor, and generating, by the at least one processor, a dynamic fracture propagation visualization showing estimated fracture growth over time correlated with the identified fluid entry locations and the real-time predicted compliance factor. The historical hydraulic fracturing data may further comprise subterranean formation data selected from at least one of well log data, core sample measurements, seismic or microseismic data, porosity or permeability measurements, or formation pressure data, and wherein the machine learning model is trained, at least in part, using the subterranean formation data in combination with the injection rates, pressure measurements, and corresponding compliance values. The method may include the aforementioned techniques but where rate and pressure are simply inverted and then integrated over time, or integrated over volume or similar metrics used to track the progression of a fracture treatment.

[0022] In various embodiments, the cumulative compliance metric may be used to support operational decision-making, automated or semi-automated control, post-stimulation analysis, performance attribution, well ranking, or training of predictive models. The compliance-related values and cumulative compliance metric may be determined using analytical techniques, data-driven models, machine learning models, or combinations thereof, and may be generated without requiring discrete rate change events or predefined stimulation patterns. According to another aspect of the present disclosure, an apparatus is provided. The apparatus includes at least one processor and a memory storing instructions that, when executed by the at least one processor, cause the apparatus to receive real-time data from at least one sensor during a hydraulic fracturing operation, wherein the real-time data comprises injection rate measurements and pressure measurements, detect a rate change event in the hydraulic fracturing operation based on the injection rate measurements, in response to detecting the rate change event, calculate a change in injection rate (ΔQ) associated with the rate change event, fit a mathematical function to a subset of the pressure measurements following the rate change event, determine a fitted pressure change (ΔPfit) based on the mathematical function, calculate a compliance value for the rate change event as a ratio of the change in injection rate to the fitted pressure change (ΔQ / ΔPfit), integrate a plurality of calculated compliance values over a time period of the hydraulic fracturing operation to generate a cumulative compliance factor, output the cumulative compliance factor to a display device, and adjust at least one operational parameter of the hydraulic fracturing operation based on the cumulative compliance factor, wherein the at least one operational parameter comprises at least one of: injection rate, proppant concentration, or fluid composition.

[0023] The disclosed systems and methods may be implemented using one or more processors executing instructions stored in memory, may operate in real time or in a deferred manner, and may be deployed in local, distributed, or remote computing environments. Data used by the systems and methods may include surface data, downhole data, historical data, inferred data, or simulated data, and data sources need not be synchronous or continuously available. According to other aspects of the present disclosure, the apparatus may include one or more of the following features. The instructions, when executed by the at least one processor, may further cause the apparatus to train a machine learning model using historical hydraulic fracturing data, wherein the historical hydraulic fracturing data comprises injection rates, pressure measurements, and corresponding compliance values, and generate, using the trained machine learning model, a real-time predicted compliance factor based on current injection rate and pressure measurements, wherein the real-time predicted compliance factor is generated without requiring discrete rate change events. The instructions, when executed by the at least one processor, may further cause the apparatus to output the real-time predicted compliance factor to the display device, and adjust the at least one operational parameter of the hydraulic fracturing operation based on the real-time predicted compliance factor. The mathematical function may comprise an exponential function. The instructions, when executed by the at least one processor, may further cause the apparatus to identify a plurality of events in the real-time data, wherein the plurality of events comprise rate step up events and rate step down events, generate an event table comprising a plurality of entries, wherein each entry corresponds to an identified event and includes an event type indicator, a start index corresponding to a beginning of the identified event, and an end index corresponding to an end of the identified event, wherein the rate change event is detected based on the generated event table. The apparatus may be configured to iteratively fit a pressure response model to the subset of pressure measurements, wherein the iterative fitting comprises fitting an initial function to the subset of pressure measurements, calculating an error for the initial function fit, and if the error value is not within a predetermined threshold, executing an iterative fitting process comprising modifying a size of an analysis window containing the subset of pressure measurements, fitting a new function to the modified analysis window, calculating a new error value for the new function fit, and repeating these steps until either the error value meets the predetermined threshold or a maximum number of iterations is reached, selecting a final mathematical function fit based on the iterative fitting process, and determining the fitted pressure change (ΔPfit) based on parameters of the final mathematical function fit. The instructions, when executed by the at least one processor, may further cause the apparatus to calculate an event-based alpha value for each rate change event, wherein the event-based alpha value is computed as a ratio of the fitted pressure change (ΔPfit) to the change in injection rate (ΔQ), integrate a plurality of calculated event-based alpha values over the time period of the hydraulic fracturing operation to generate a cumulative alpha factor, wherein the machine learning model is further trained using the event-based alpha values, and wherein the real-time predicted compliance factor and a real-time predicted alpha factor are generated simultaneously by the trained machine learning model, and wherein the apparatus is configured to calculate an instantaneous shut-in pressure (ISIP) alpha value when an ISIP event is detected in the real-time data, the ISIP alpha value being computed as a ratio of a pressure change during the ISIP event to a corresponding change in injection rate. The instructions, when executed by the at least one processor, may further cause the apparatus to filter the calculated compliance values to exclude compliance values greater than a predetermined threshold, and bin the filtered compliance values and corresponding leak-off data by hydraulic fracturing stage, wherein the predetermined threshold is less than 50, and wherein the binning of compliance values and leak-off data by stage facilitates stage-specific analysis of fracture behavior and fluid efficiency during the hydraulic fracturing operation. The instructions, when executed by the at least one processor, may further cause the apparatus to generate a summary statistics report for the hydraulic fracturing operation, wherein the summary statistics report includes a total number of rate step up events processed, a total number of rate step down events processed, and a total number of instantaneous shut-in pressure (ISIP) events processed, wherein the summary statistics report provides an overview of the frequency and distribution of different event types throughout the hydraulic fracturing operation, and wherein the summary statistics report is used to assess the overall behavior of the reservoir and the effectiveness of the fracturing process. The instructions, when executed by the at least one processor, may further cause the apparatus to analyze frequency components of the pressure measurements using a Fourier transform to identify dominant frequencies associated with rock breakage during the hydraulic fracturing operation, correlate the identified dominant frequencies with geological characteristics of the formation being fractured, and adjust at least one of injection rate or fluid composition to target specific frequency ranges associated with optimal fracture propagation in the formation. The instructions, when executed by the at least one processor, may further cause the apparatus to integrate distributed acoustic sensing (DAS) data from a fiber optic cable deployed in the wellbore with the injection rate and pressure measurements, identify, using the trained machine learning model, locations of fluid entry into the formation based on patterns in the DAS data correlated with changes in the real-time predicted compliance factor, and generate a dynamic fracture propagation visualization showing estimated fracture growth over time correlated with the identified fluid entry locations and the real-time predicted compliance factor. The historical hydraulic fracturing data may further comprise subterranean formation data selected from at least one of well log data, core sample measurements, seismic or microseismic data, porosity or permeability measurements, or formation pressure data, and wherein the machine learning model is trained, at least in part, using the subterranean formation data in combination with the injection rates, pressure measurements, and corresponding compliance values.

[0024] According to another aspect of the present disclosure, a method for monitoring hydraulic fracturing operations manually is provided. The method includes continuously measuring injection rate and pressure during fracturing at a selected sampling interval, using a processor to analytically compute a compliance value C from changes in injection rate relative to pressure change, integrating said compliance values over time to form a C-Factor, displaying the C-Factor on a user interface, and adjusting fracturing parameters based on said C-Factor to optimize reservoir stimulation.

[0025] According to another aspect of the present disclosure, a method for monitoring hydraulic fracturing operations with an AI system is provided. The method includes collecting injection rate, pressure, and offset well data in real time, feeding said data into a machine learning model trained to compute C-Factor in real time, generating, using the machine learning model, a real-time C-Factor based on the collected data, displaying the real-time C-Factor on a user interface, and modifying fracturing parameters to adjust fracture growth based on the real-time C-Factor.

[0026] According to another aspect of the present disclosure, an apparatus for real-time compliance measurement using manual techniques is provided. The apparatus includes a data acquisition subsystem that receives fracturing pressure and rate measurements, a user interface configured to allow a user to flag out event sequences and then manually calculate a compliance value C=ΔQ / ΔP from measured pressure changes and injection rate changes corresponding to the flagged event sequences, a summation module to integrate said compliance values over time into a C-Factor, a display device for presenting the C-Factor to an operator, and a control panel allowing the operator to adjust fracturing parameters based on C-Factor output.

[0027] According to another aspect of the present disclosure, an apparatus for real-time compliance measurement with AI is provided. The apparatus includes a data acquisition module collecting pressure, rate, and optionally other sensors like offset well data at a specified sample interval, a machine learning processor configured to compute C-Factor in real time, a display device for presenting the real-time C-Factor, and a control interface that can adjust fracturing parameters upon detecting unwanted C-Factor outcome.

[0028] According to another aspect of the present disclosure, a non-transitory computer-readable medium storing instructions is provided. When executed by at least one processor, the instructions cause the at least one processor to perform operations comprising receiving real-time data from at least one sensor during a hydraulic fracturing operation, wherein the real-time data comprises injection rate measurements and pressure measurements, detecting a rate change event in the hydraulic fracturing operation based on the injection rate measurements, in response to detecting the rate change event, calculating a change in injection rate (ΔQ) associated with the rate change event, fitting a mathematical function to a subset of the pressure measurements following the rate change event, determining a fitted pressure change (ΔPfit) based on the mathematical function, calculating a compliance value for the rate change event as a ratio of the change in injection rate to the fitted pressure change (ΔQ / ΔPfit), integrating a plurality of calculated compliance values over a time period of the hydraulic fracturing operation to generate a cumulative compliance factor, outputting the cumulative compliance factor to a display device, and adjusting at least one operational parameter of the hydraulic fracturing operation based on the cumulative compliance factor, wherein the at least one operational parameter comprises at least one of: injection rate, proppant concentration, or fluid composition.

[0029] According to another aspect of the present disclosure, a method for generating cumulative compliance values from hydraulic fracturing data is provided. The method includes receiving, by at least one processor, pressure data and injection rate data for one or more stimulation stages of one or more wells, identifying, by the at least one processor, a plurality of rate change events in the pressure and injection rate data, wherein the rate change events are identified using manually placed flags, automatically detected events, or combinations thereof, and wherein the plurality of rate change events comprise at least one of rate step up events or rate step down events, generating, by the at least one processor, an event table comprising a plurality of entries, wherein each entry corresponds to an identified rate change event and includes at least an event type indicator, a start index corresponding to a beginning of the identified event, and an end index corresponding to an end of the identified event, fitting, by the at least one processor, a mathematical function to a subset of the pressure measurements associated with each identified event to determine a fitted pressure change (ΔPfit), wherein in some implementations the mathematical function comprises an exponential function, calculating, by the at least one processor, a compliance value for each identified event as a ratio of injection rate change to fitted pressure change (ΔQ / ΔPfit), filtering the calculated compliance values to exclude compliance values greater than a predetermined threshold, calculating, by the at least one processor, a cumulative compliance value for each stimulation stage by aggregating the filtered compliance values corresponding to that stage, and storing the cumulative compliance values along with associated event table data and stage metadata for subsequent use in modeling, analysis, production forecasting, or economic evaluation.

[0030] According to another aspect of the present disclosure, a method for training a machine learning model to predict cumulative compliance is provided. The cumulative compliance values generated by the method of paragraph may serve as training labels for a predictive model. The method includes assembling, by at least one processor, a training dataset comprising, for each of a plurality of historical stimulation stages, input features comprising at least pressure data, injection rate data, and formation data, wherein the formation data comprises at least one of well log data, core sample measurements, seismic or microseismic data, porosity or permeability measurements, formation pressure data, distributed acoustic sensing (DAS) data, geological or drilling metadata such as well trajectory, formation tops, or stress logs, measurement-while-drilling (MWD) data, distance to geological markers, lithological indicators, or formation stress indicators, and output labels comprising the cumulative compliance values generated for those stages according to the method of paragraph

[0029] , and training a machine learning model on the assembled training dataset, wherein in some implementations the machine learning model comprises a neural network, such that the trained model learns to predict cumulative compliance values from the input features without requiring detection of discrete rate change events during inference. Because geological or drilling metadata may be loaded into the model and used to anticipate compliance behavior and optimize stage design as described herein, the trained model may account for geological, reservoir, and wellbore variations that affect compliance response behavior across different wells and formations.

[0031] According to another aspect of the present disclosure, a method for generating a production forecast using predicted cumulative compliance is provided. Because higher compliance may correlate with higher production as described herein, and because comparing compliance factors across multiple wells or stages may allow operators to identify best practices and optimize their fracturing operations for specific reservoir conditions, cumulative compliance values may be used to evaluate well performance and generate production forecasts. The method includes generating, by the trained machine learning model of paragraph during a live fracturing operation or applied retrospectively to historical data, predicted cumulative compliance values representing accumulated fracture system response, providing the predicted cumulative compliance values to a user interface during the operation by at least one of rendering on a display, transmitting through an application programming interface, or generating a data file for operator access, generating a visual alert when at least one predicted cumulative compliance value falls outside a predetermined range, aggregating the predicted cumulative compliance values across a plurality of stages within a well to generate a well-level cumulative compliance value, correlating well-level cumulative compliance values from a plurality of historically completed wells with measured production outcomes for those wells to construct a production profile relating cumulative compliance to production at one or more time intervals, and applying the production profile to predicted cumulative compliance values for a current or prospective well to generate a production forecast. Because the trained model can be deployed on wells that were not manually flagged for events, the method enables generation of cumulative compliance values and corresponding production forecasts across an entire field of wells, including wells for which manual event flagging was not performed.

[0032] According to another aspect of the present disclosure, a method for identifying optimization parameters using predicted cumulative compliance is provided. Because the production profile described in paragraph

[0031] establishes a quantitative relationship between cumulative compliance and production outcomes, and because controllable operational parameters exhibit measurable influence on the magnitude of the cumulative compliance metric as described herein, the method includes evaluating, by at least one processor, predicted cumulative compliance values across a plurality of wells in relation to controllable operational parameters associated with those wells, wherein the controllable operational parameters comprise at least one of maximum injection rate, time to maximum injection rate, total fluid volume, total proppant mass, proppant intensity, completion system design, fluid viscosity, landing depth relative to geological markers, injection rate, proppant concentration, fluid composition, or stage design, identifying, from the evaluation, which controllable operational parameters are associated with higher predicted cumulative compliance values for a given formation or field, and adjusting one or more of the controllable operational parameters for a current or future hydraulic fracturing operation based on the identified associations to improve predicted cumulative compliance and thereby improve predicted production outcomes.

[0033] According to another aspect of the present disclosure, a method for calculating economic metrics during hydraulic fracturing using predicted cumulative compliance is provided. Because the controllable operational parameters identified in paragraph each have associated operational costs, and because cumulative compliance correlates with production outcomes as described herein, the method includes receiving, by at least one processor, operational cost data associated with one or more controllable operational parameters during a hydraulic fracturing operation, calculating an economic metric based at least in part on the predicted cumulative compliance values and the operational cost data, wherein the economic metric comprises at least one of cost per unit compliance, projected return on investment, or profit-loss projection, providing the economic metric to a user interface during the hydraulic fracturing operation, and adjusting one or more controllable operational parameters during the operation based on the economic metric to optimize economic return. Because the trained machine learning model generates predicted cumulative compliance values in real time as described herein, the economic metric may be updated continuously during the operation, enabling operators to make cost-benefit decisions while fracturing is actively underway rather than after completion.

[0034] According to another aspect of the present disclosure, a system for generating cumulative compliance values from hydraulic fracturing data is provided. The system includes at least one data interface configured to receive pressure data and injection rate data from at least one surface sensor during a hydraulic fracturing operation, at least one processor, and a memory storing instructions that, when executed by the at least one processor, cause the system to identify a plurality of rate change events in the received data using manually placed flags, automatically detected events, or combinations thereof, generate an event table comprising entries for each identified rate change event including at least an event type indicator, a start index, and an end index, fit a mathematical function to pressure measurements associated with each event to determine a fitted pressure change (ΔPfit), wherein in some implementations the mathematical function comprises an exponential function, calculate a compliance value for each event as a ratio of injection rate change to fitted pressure change (ΔQ / ΔPfit), filter the calculated compliance values to exclude values greater than a predetermined threshold, calculate cumulative compliance values for each stimulation stage by aggregating filtered compliance values, and store the cumulative compliance values along with associated event table data and stage metadata.

[0035] According to another aspect of the present disclosure, a system for training and deploying a machine learning model to predict cumulative compliance is provided. The cumulative compliance values generated by the system of paragraph may serve as training labels for a predictive model. The system includes at least one processor and a memory storing instructions that, when executed by the at least one processor, cause the system to assemble a training dataset comprising, for each of a plurality of historical stimulation stages, input features comprising at least pressure data, injection rate data, and formation data, wherein the formation data comprises at least one of well log data, core sample measurements, seismic or microseismic data, porosity or permeability measurements, formation pressure data, distributed acoustic sensing (DAS) data, geological or drilling metadata such as well trajectory, formation tops, or stress logs, measurement-while-drilling (MWD) data, distance to geological markers, lithological indicators, or formation stress indicators, and output labels comprising the cumulative compliance values generated by the system of paragraph

[0034] , train a machine learning model on the assembled training dataset, wherein in some implementations the machine learning model comprises a neural network, and deploy the trained machine learning model to generate predicted cumulative compliance values from pressure, rate, and formation data without requiring detection of discrete rate change events during inference. Because geological or drilling metadata may be loaded into the model as described herein, the trained model may account for geological and wellbore variations affecting compliance response.

[0036] According to another aspect of the present disclosure, a system for generating production forecasts using predicted cumulative compliance is provided. Because higher compliance may correlate with higher production as described herein, and because comparing compliance factors across multiple wells or stages may allow operators to identify best practices, the system includes at least one processor and a memory storing instructions that, when executed by the at least one processor, cause the system to generate predicted cumulative compliance values using the trained model of paragraph during a live fracturing operation or applied retrospectively to historical well data, provide the predicted cumulative compliance values to a user interface by one or more of display rendering, application programming interface transmission, or data-file generation, generate a visual alert when at least one predicted cumulative compliance value falls outside a predetermined range, aggregate predicted cumulative compliance values across stages within a well to generate well-level cumulative compliance values, correlate well-level cumulative compliance values from historically completed wells with measured production outcomes to construct a production profile, and apply the production profile to predicted cumulative compliance values for a current or prospective well to generate a production forecast. Because the trained model can be applied to wells that were not manually flagged for events, the system enables production forecasting across an entire field including wells for which manual event flagging was not performed.

[0037] According to another aspect of the present disclosure, a system for identifying optimization parameters using predicted cumulative compliance is provided. Because controllable operational parameters exhibit measurable influence on the magnitude of the cumulative compliance metric as described herein, and because the production profile of paragraph

[0036] quantifies the relationship between cumulative compliance and production, the system includes at least one processor and a memory storing instructions that, when executed by the at least one processor, cause the system to evaluate predicted cumulative compliance values across a plurality of wells in relation to controllable operational parameters, wherein the controllable operational parameters comprise at least one of maximum injection rate, time to maximum injection rate, total fluid volume, total proppant mass, proppant intensity, completion system design, fluid viscosity, landing depth relative to geological markers, injection rate, proppant concentration, fluid composition, or stage design, identify which controllable operational parameters are associated with higher predicted cumulative compliance values for a given formation or field, and transmit control signals or recommendations to adjust one or more controllable operational parameters for a current or future operation to improve predicted cumulative compliance and production outcomes.

[0038] According to another aspect of the present disclosure, a system for calculating economic metrics during hydraulic fracturing using predicted cumulative compliance is provided. Because the controllable operational parameters of paragraph

[0037] each have associated operational costs, and because cumulative compliance correlates with production outcomes as described herein, the system includes at least one processor and a memory storing instructions that, when executed by the at least one processor, cause the system to receive operational cost data associated with controllable operational parameters during a hydraulic fracturing operation, calculate an economic metric based at least in part on the predicted cumulative compliance values and the operational cost data, wherein the economic metric comprises at least one of cost per unit compliance, projected return on investment, or profit-loss projection, provide the economic metric to a user interface during the operation, and transmit control signals to hydraulic fracturing equipment to adjust one or more controllable operational parameters based on the economic metric. Because predicted cumulative compliance values are generated in real time as described herein, the economic metric may be updated continuously during the operation, enabling cost-benefit optimization while fracturing is actively underway.

[0039] The foregoing general description of the illustrative embodiments and the following detailed description thereof are merely exemplary aspects of the teachings of this disclosure and are not restrictive.BRIEF DESCRIPTION OF FIGURES

[0040] Non-limiting and non-exhaustive examples are described with reference to the following figures.

[0041] FIG. 1 illustrates a flowchart for a method of real-time compliance factor prediction in hydraulic fracturing operations, according to aspects of the present disclosure.

[0042] FIG. 2 illustrates a flowchart for a machine learning process to predict real-time compliance factors, according to an embodiment.

[0043] FIG. 3 illustrates a flowchart for monitoring hydraulic fracturing operations, in accordance with example embodiments.

[0044] FIG. 4 illustrates a flowchart for an AI-based method for monitoring hydraulic fracturing operations, according to aspects of the present disclosure.

[0045] FIG. 5 illustrates a block diagram of a real-time compliance factor prediction system, according to an embodiment.

[0046] FIG. 6 illustrates a sequence diagram of data flow in a real-time hydraulic fracturing compliance monitoring system, according to aspects of the present disclosure.

[0047] FIG. 7 illustrates a flowchart of a compliance factor calculation process, in accordance with example embodiments.

[0048] FIG. 8 illustrates a block diagram of a hydraulic fracturing system, according to an embodiment.

[0049] FIG. 9 illustrates a system diagram of a C-Factor Calculation System, according to aspects of the present disclosure.

[0050] FIG. 10 illustrates a block diagram of a software architecture for a C-Factor calculation system, in accordance with example embodiments.

[0051] FIG. 11 illustrates an exemplary pressure response event during a hydraulic fracturing operation in which the operation continued before the pressure response reached a stabilized value, requiring the fitted pressure change (ΔPfit) to be determined by fitting a mathematical function to the available subset of pressure measurements and extrapolating to estimate the asymptotic pressure (Poo).

[0052] FIG. 12 illustrates an exemplary manually calculated cumulative compliance value for a stimulation stage, plotted alongside corresponding pressure and injection rate traces, showing the cumulative compliance increasing in discrete steps corresponding to individual rate change events.

[0053] FIG. 13 illustrates an exemplary AI-predicted cumulative compliance value for a stimulation stage, plotted alongside corresponding pressure and injection rate traces, showing the cumulative compliance generated continuously by a trained machine learning model without requiring detection of discrete rate change events.

[0054] FIG. 14 illustrates an exemplary pressure response event during a hydraulic fracturing operation in which the pressure response reached a stabilized asymptotic value (Poo) prior to the next operational change.DETAILED DESCRIPTION

[0055] The following description sets forth exemplary aspects of the present disclosure. It should be recognized, however, that such description is not intended as a limitation on the scope of the present disclosure. Rather, the description also encompasses combinations and modifications to those exemplary aspects described herein.

[0056] A detailed description of systems, devices, and methods consistent with embodiments of the present disclosure is provided below. While several embodiments are described, it should be understood that disclosure is not limited to any one embodiment, but instead encompasses numerous alternatives, modifications, and equivalents. In addition, while numerous specific details are set forth in the following description in order to provide a thorough understanding of the embodiments disclosed herein, some embodiments can be practiced without some or all of these details. Moreover, for the purpose of clarity, certain technical material that is known in the related art has not been described in detail in order to avoid unnecessarily obscuring the disclosure.

[0057] As used herein, a compliance-related value refers to any value derived from pressure data and injection behavior that characterizes fracture system response, including analytically determined values, inferred values, predicted values, or learned representations, and is not limited to any specific mathematical form, event detection technique, or timing of computation. In some cases, the compliance factor, denoted as C, may be utilized as a key parameter in hydraulic fracturing operations to provide information about the behavior of the fracture network during the stimulation process. In its traditional formulation, compliance may be defined as the ratio of volume change to pressure change (ΔV / ΔP). In operational implementations, particularly for real-time monitoring during active fracturing operations, individual compliance values may be calculated as the ratio of injection rate change to pressure change (ΔQ / ΔP), which serves as a practical proxy for volumetric compliance and enables calculation without requiring mathematical integration of flow rates to determine cumulative volumes. However, prior art compliance measurement approaches, such as step-rate tests, typically measure compliance as discrete, isolated values at specific operational points and do not integrate or accumulate these compliance measurements over time, nor do they incorporate geological, reservoir, or wellbore contextual information that may influence compliance behavior between measurements. Similarly, prior art fracture characterization systems—such as fracture geometry modeling tools or real-time monitoring systems that estimate fracture dimensions (e.g., fracture half-length, height, or width)—focus primarily on predicting geometric parameters rather than characterizing fracture system quality or responsiveness in ways that directly correlate with production outcomes. In contrast, the disclosed systems and methods integrate a plurality of compliance values—calculated over numerous rate change events throughout a stimulation stage and optionally informed by formation properties such as measurement-while-drilling (MWD) data, gamma ray logs, or distance to geological tops or markers—to generate a cumulative compliance metric (referred to herein as “C-Factor” or “cumulative compliance factor”). This cumulative metric, formed by integrating or summing compliance values over time (e.g., Σ[ΔQ / AP] over multiple events), represents the accumulated fracture system responsiveness throughout the stimulation and provides a fundamentally different characterization of fracture behavior than discrete compliance snapshots or geometric dimension estimates. FIG. 12 illustrates an exemplary manually calculated cumulative compliance value for a stimulation stage. The figure shows three traces plotted over time: an upper trace representing injection rate measurements, a middle trace representing pressure measurements, and a lower stepped trace representing the cumulative compliance value, which increases in discrete steps as each rate change event is detected and its corresponding compliance value is added to the cumulative total. Unlike geometry-focused predictions that estimate physical fracture size, the cumulative C-Factor characterizes fracture system quality and effectiveness in ways that enable predictive correlations with production outcomes. The cumulative compliance factor (C-Factor) may be used as an indicator of overall fracture system quality, stimulation effectiveness, and production potential. A higher cumulative C-Factor may generally suggest better overall fracture propagation and reservoir stimulation, while a lower cumulative value may indicate less effective fracturing or potential problems such as fluid diversion into undesired zones.

[0058] As used herein, “C-Factor” or “cumulative compliance factor” refers to a machine learning model output that represents predicted cumulative fracture system compliance during hydraulic fracturing operations. Unlike discrete compliance measurements used in conventional step-rate tests or diagnostic injection tests—which provide isolated compliance snapshots at specific operational points (typically 3-5 measurements per test)—the C-Factor represents a continuously-generated prediction of accumulated fracture system response throughout a stimulation stage. The C-Factor is generated by a trained machine learning model (such as a neural network) that receives as inputs real-time pressure measurements, injection rate measurements, and physics-informed formation features (such as measurement-while-drilling data, gamma ray logs, distance to geological markers, or other subterranean formation characteristics), and outputs predicted cumulative compliance values. The machine learning model may be trained using historical hydraulic fracturing data wherein cumulative compliance ground truth values are generated by detecting numerous rate change events (e.g., 20 to over 80 events per stage), computing individual compliance values (such as ΔQ / ΔP) for each detected event through curve fitting techniques, and integrating or summing these individual compliance values to generate cumulative training targets (e.g., Σ[ΔQ / AP] over events in a stage). FIG. 11 illustrates an exemplary pressure response event in which the operation continued before the pressure response reached a stabilized value, requiring mathematical estimation of the asymptotic pressure. FIG. 11 shows two traces: a longer trace representing the measured hydraulic fracturing pressure during the event, and a shorter trace representing the fitted exponential function applied to a subset of the measured pressure data. Because the fracturing operation continued before the pressure stabilized, the fitted function extrapolates to estimate the asymptotic pressure (P∞) that would have been reached if sufficient time had elapsed, thereby enabling determination of the fitted pressure change (ΔPfit) for compliance calculation despite the truncated measurement window. FIG. 14 illustrates an exemplary pressure response event during a hydraulic fracturing operation in which the pressure response reached a stabilized asymptotic value (P∞) prior to the next operational change. FIG. 14 similarly shows two traces: a longer trace representing the measured hydraulic fracturing pressure during the event, and a shorter trace representing the fitted exponential function. In contrast to FIG. 11, where the operation continued before pressure stabilization requiring mathematical extrapolation, FIG. 14 demonstrates the case where sufficient time elapsed between rate change events for the pressure to approach its equilibrium value, providing a direct measurement against which the accuracy of the curve fitting technique may be validated. The physics-informed formation features associated with each training example enable the machine learning model to learn how geological, reservoir, and wellbore context influences the relationship between pressure / rate behavior and cumulative compliance, such that the model can distinguish compliance responses across varying formation conditions. Once trained, the machine learning model generates C-Factor predictions continuously during active fracturing operations without requiring manual detection of discrete rate change events, enabling real-time characterization of fracture system quality and responsiveness. FIG. 13 illustrates an exemplary AI-predicted cumulative compliance value for the same type of stimulation stage. The figure shows three traces plotted over time: an upper trace representing injection rate, a middle trace representing pressure and a lower trace showing the cumulative compliance, generated continuously by the trained machine learning model. The C-Factor may correlate with production outcomes in ways not achievable with non-integrated discrete compliance measurements or geometric fracture characterization approaches, and may be used for operational adjustments, post-stimulation analysis, well ranking, or economic optimization during hydraulic fracturing operations. The distinguishing characteristic of the C-Factor is that it represents a machine learning-predicted cumulative compliance metric informed by physics-based formation features, rather than a manually-calculated sum of discrete measurements or a geometric fracture dimension estimate.

[0059] The disclosed systems and methods may be applied to various hydraulic fracturing operation configurations. In some implementations, the operation may involve discrete stimulation stages, where each stage corresponds to a defined set of perforation clusters or intervals that are isolated (e.g., via plug-and-perf techniques) and fractured individually or concurrently. Cumulative compliance values may be generated for each discrete stage, whether stages are stimulated sequentially (one after another) or concurrently (multiple stages fractured simultaneously, such as in simul-frac or zipper frac operations). The trained machine learning model may process sensor data from multiple discrete stages and generate corresponding cumulative compliance values for each stage, enabling real-time monitoring and optimization across multiple concurrent stimulation activities.

[0060] In other implementations, the operation may involve continuous or substantially continuous fluid injection throughout multiple perforation intervals without mechanical stage isolation, such as in open hole completions, slotted liner completions, or heel-to-toe fracturing operations where the wellbore is initially perforated along its length and fracturing fluid is injected continuously while operational parameters are dynamically adjusted during the operation. In such continuous operations, “stages” may be defined operationally or dynamically based on changes in operational parameters, introduction of diverting agents, or observed changes in formation response. For example, a first operational stage may correspond to an initial injection interval, and a subsequent stage may be initiated when diverting agents are introduced to redirect fluid to different formation zones, when proppant concentration or type is changed (such as during sand ramps or proppant switches), or when cumulative compliance values indicate a plateau or change in formation responsiveness suggesting a transition to different stimulation behavior. The trained machine learning model may generate cumulative compliance values continuously throughout such operations, and the cumulative values at different time intervals or operational phases may be used to assess formation response, determine optimal timing for parameter adjustments (such as introduction of diverting agents or changes to fluid composition or proppant concentration), and enable real-time operational decisions based on observed fracture system behavior.

[0061] In continuous fracturing implementations, the cumulative compliance metric may be monitored for indicators of formation response saturation or diminishing returns, such as when the rate of cumulative compliance increase plateaus or decreases, suggesting that the current treatment zone has reached a state where continued fluid injection is less effective. Upon detecting such indicators, operational adjustments may be triggered-such as introducing diverting agents to redirect fluid to untreated or less-stimulated formation zones, transitioning to different proppant concentrations or types, or adjusting injection rates or fluid composition. This approach enables dynamic optimization throughout continuous operations, where “stages” are not predefined by mechanical isolation but are instead defined by operational transitions driven by real-time formation response characterization via cumulative compliance monitoring. The cumulative compliance values generated by the trained machine learning model represent the accumulated fracture system response throughout the stimulation stage. This accumulated response metric may characterize fracture system behavior, formation responsiveness, fracture network quality, stimulation effectiveness, or other aspects of the fracture system's interaction with the stimulation process. In some implementations, the cumulative compliance values may correlate with production potential, economic value, or other operational outcomes.

[0062] The formation data utilized by the disclosed systems and methods maybe obtained from various sources and measurement technologies. Measurement-while-drilling (MWD) data may include measurements acquired during the drilling process, such as gamma ray measurements, resistivity measurements, inclination and azimuth data, and other downhole parameters collected by sensor packages integrated into the drill string. Well log data may include measurements obtained through wireline logging operations or logging-while-drilling, providing information about formation properties such as porosity, density, acoustic velocity, or lithology. Distributed acoustic sensing (DAS) data may be obtained from fiber optic sensing systems deployed along the wellbore, as described in greater detail elsewhere in this disclosure. Distance to geological markers may be calculated based on wellbore trajectory data and identified formation boundaries, tops, or other stratigraphic features, representing the spatial relationship between a measurement point in the wellbore and known geological reference features.

[0063] By incorporating these diverse formation data sources as input features to the machine learning models, the disclosed systems account for geological, reservoir, and wellbore variations that affect compliance response behavior. This physics-informed approach enables the models to distinguish between compliance variations caused by operational parameters (such as injection rate changes) versus those caused by formation property variations (such as changes in rock mechanical properties or proximity to formation boundaries). Prior art compliance measurement techniques that rely solely on pressure and rate data, without accounting for formation context, fail to capture these geological influences on compliance behavior.

[0064] In some aspects, positive compliance (C>0) may occur when pressure increases and the fracture absorbs a meaningful amount of volume. This may indicate that the rock is responding elastically, and the fracture is expanding in a way that sustains new volume. Positive compliance may suggest a large or effective fracture network, one that remains open long enough to contribute to production.

[0065] In some cases, if the summed “C-factor” is high, it may indicate that:

[0066] 1. The fractures were successfully expanded into new productive rock.

[0067] 2. The fractures were sustained and connected to an effective drainage area.

[0068] 3. There was minimal restriction to fluid expansion.

[0069] 4. Proppant placement may have been more effective because fractures were still open.

[0070] In some implementations, negative or near-zero compliance (C≈0 or C<0) may occur if a rate increase does not result in a meaningful volume increase, suggesting that the fracture is not growing efficiently. This may indicate that:

[0071] 1. The fractures are hitting barriers (e.g., breaking into a depleted parent well).

[0072] 2. The fractures are not propagating effectively due to stress constraints.

[0073] 3. The fracture system is losing connectivity (e.g., frac stages not communicating well with the reservoir).

[0074] 4. Proppant is being prematurely restricted or settled.

[0075] In some aspects, higher compliance may correlate with higher production. This may suggest that more compliant fractures (higher compliance values) are leading to better reservoir drainage and overall well performance. A more compliant system (higher C-factor) may indicate that the fractures remain open longer, connect well, and sustain productive pathways.

[0076] In some cases, if compliance is low, it may indicate fracture stiffness, limited connectivity, or rapid fracture closure—potentially reducing drainage area effectiveness. Conversely, if compliance is high, the fracture may keep accepting fluid without major restriction, suggesting a better long-term permeability network.

[0077] The temporal evolution of cumulative compliance during a fracturing operation may provide additional operational insights. In some implementations, cumulative compliance values may increase continuously throughout a stimulation stage as fractures propagate and formation response develops. However, in some cases, the rate of cumulative compliance increase may decrease or stabilize over time, indicating that the formation's capacity to accept additional fluid injection at current operational parameters may be approaching saturation or that fracture propagation into the target zone may be diminishing. Such temporal patterns—for example, when cumulative compliance values plateau, exhibit reduced rate of increase, or show diminishing responsiveness—may indicate operational transition points where adjustments to stimulation parameters could enhance treatment effectiveness. Operators may use real-time monitoring of cumulative compliance evolution to identify these transition points and determine optimal timing for operational modifications, such as changes to injection rate, introduction of diverting agents to redirect fluid to different formation zones, adjustments to proppant concentration or type, or transitions to different fluid compositions. This capability to detect and respond to temporal changes in formation responsiveness, as characterized by cumulative compliance evolution, enables dynamic operational optimization throughout fracturing operations.

[0078] As used herein, integrating may include summing, accumulating, applying one or more weighted averaging techniques, or using other suitable algorithms that combine multiple compliance values into a single or composite measure. In some implementations, the integration process may involve calculating a cumulative C-factor over time or across multiple fracturing stages.

[0079] As used herein, “real-time” when applied to data acquisition, processing, prediction, or system operation refers to performance of operations during the course of a hydraulic fracturing operation with latency sufficiently low to enable operational adjustments, decision-making, or system responses during active stimulation activities, as distinguished from post-completion analysis, deferred processing, or retrospective evaluation. Real-time operations encompass processing and output occurring concurrently with, or within a practically usable timeframe during, the ongoing hydraulic fracturing operation, such that the generated information (e.g., compliance predictions, alerts, recommendations) can inform or influence operational parameters while the stimulation is actively proceeding. The term “real-time” does not require instantaneous processing with zero latency or responses within specific millisecond thresholds; rather, it encompasses any processing speed that enables the system's outputs to be actionable during the fracturing operation. This includes continuous or pseudo-continuous output that updates as new sensor data is received, for example, real-time may include generating compliance predictions or alerts within timeframes ranging from sub-second intervals to several minutes, provided such timeframes enable operators or automated control systems to make adjustments to injection rates, fluid composition, proppant concentration, pumping pressure, or other operational parameters before completion of a stimulation stage or the overall fracturing operation. Furthermore, ‘real-time’ implies a temporal link between the current physical state of the wellbore and the generated prediction, enabling a closed-loop or human-in-the-loop response while the formation is under pressure. Real-time performance is contrasted with offline, post-job, or delayed analysis that occurs hours, days, or weeks after stimulation completion and therefore cannot influence the active fracturing process. In various implementations, real-time processing may involve streaming data analysis, online learning, continuous prediction updates, or feedback control loops operating during active hydraulic fracturing, and may accommodate practical operational latencies associated with data transmission, computational processing, and control system response times typical of industrial hydraulic fracturing environments.

[0080] In some aspects, the compliance factor and its integration may be used in conjunction with other fracturing parameters and measurements to provide a comprehensive analysis of the fracturing operation. For example, the compliance factor may be correlated with pressure data, injection rates, and seismic measurements to gain deeper insights into fracture propagation and reservoir behavior.

[0081] In some cases, real-time monitoring and analysis of the compliance factor may enable operators to make informed decisions during the fracturing process. For instance, if a sudden decrease in compliance is observed, it may prompt adjustments to the injection rate, fluid composition, or proppant concentration to optimize fracture growth and prevent potential issues such as screenouts or undesired fluid diversion.

[0082] The compliance factor may also be utilized in post-job analysis to evaluate the effectiveness of different fracturing strategies and to inform future well designs and completion plans. By comparing the compliance factors across multiple wells or stages, operators may identify best practices and optimize their fracturing operations for specific reservoir conditions.

[0083] In some implementations, machine learning algorithms may be employed to analyze historical compliance factor data along with other relevant parameters to develop predictive models for fracture behavior and well performance. These models may assist in real-time decision-making during fracturing operations and in the planning of future wells.

[0084] Field validation across multiple hydraulic fracturing operations in diverse geological settings has demonstrated consistent relationships between operational parameters and the cumulative compliance metric. Empirical observations from stimulation operations spanning different shale formations and jurisdictions indicate that controllable parameters—including injection rate profiles and total fluid volumes—exhibit measurable influence on the magnitude of the cumulative compliance metric. Additionally, formation-specific characteristics, such as those captured in well log data or formation tops as described in geological metadata loaded into the model, and well placement parameters relative to these tops have been observed to correlate with compliance factor variability, enabling identification of optimal distances to measurable geological layers for enhanced rock quality assessment. Importantly, wells exhibiting higher cumulative compliance metrics across their stimulation stages have demonstrated correlation with improved production outcomes in subsequent production periods, suggesting the metric's utility as a leading indicator for well performance. These observations support the metric's application not only for real-time operational adjustments but also for predictive modeling of production potential and comparative evaluation of completion strategies across heterogeneous reservoir conditions.

[0085] The integration of compliance factor analysis with advanced sensing technologies, such as distributed acoustic sensing (DAS) and distributed temperature sensing (DTS), may provide a more comprehensive understanding of fracture propagation and fluid distribution within the reservoir. This integrated approach may enable more precise control over fracture growth and optimize resource utilization during hydraulic fracturing operations.

[0086] As used herein, integrating may include summing, accumulating, applying one or more weighted averaging techniques, or using other suitable algorithms that combine multiple compliance values into a single or composite measure.

[0087] In some cases, a method for real-time compliance factor prediction in hydraulic fracturing operations may be implemented as illustrated in FIG. 1. FIG. 1 depicts a flowchart illustrating a sequence of steps for predicting and utilizing compliance factors during hydraulic fracturing operations.

[0088] The method may begin with a step 100 where real-time data is received from at least one sensor during a hydraulic fracturing operation. In some cases, the real-time data may include injection rate measurements and pressure measurements.

[0089] Following the data reception, a step 102 may involve detecting a rate change event in the hydraulic fracturing operation. The rate change detection may be based on the injection rate measurements received in step 100.

[0090] A decision point 104 may then determine if a rate change event is detected. If no event is detected, the process may return to step 100 to continue receiving real-time data. If an event is detected, the process may proceed to subsequent steps for analysis.

[0091] In some implementations, a step 106 may calculate a change in injection rate (ΔQ) associated with the detected rate change event. This calculation may provide a quantitative measure of how the injection rate has changed during the event.

[0092] A step 108 may involve fitting a mathematical function to a subset of the pressure measurements following the rate change event. In some cases, the mathematical function may comprise an exponential function, which may accurately model the pressure response behavior.

[0093] Following the function fitting, a step 110 may determine a fitted pressure change (ΔPfit) based on the mathematical function. This fitted pressure change may represent the pressure response to the rate change event.

[0094] A step 112 may then calculate a compliance value for the rate change event. In some implementations, the compliance value may be calculated as a ratio of the change in injection rate to the fitted pressure change (ΔQ / ΔPfit).

[0095] The method may include a step 114 that integrates a plurality of calculated compliance values over a time period of the hydraulic fracturing operation. This integration may generate a cumulative compliance factor, which may provide a broader perspective on fracture behavior compared to individual compliance value measurements.

[0096] A step 116 may output the cumulative compliance factor to a display device, allowing operators to visualize and interpret the compliance data in real-time.

[0097] Finally, a step 118 may involve adjusting at least one operational parameter of the hydraulic fracturing operation based on the cumulative compliance factor. In some cases, the adjusted operational parameter may include at least one of: injection rate, proppant concentration, or fluid composition.

[0098] In some implementations, the method may include identifying a plurality of events in the real-time data, including rate step up events and rate step down events. An event table may be generated with entries for each identified event, including an event type indicator, a start index corresponding to the beginning of the identified event, and an end index corresponding to the end of the identified event. The rate change event detected in step 102 may be based on this generated event table.

[0099] By following this sequence of steps, the method may enable real-time prediction and utilization of compliance factors during hydraulic fracturing operations, potentially allowing for more informed decision-making and optimization of the fracturing process.

[0100] In field implementations spanning operations across multiple jurisdictions and diverse geological formations, the disclosed systems and methods have been validated through rigorous blind testing protocols. In one exemplary blind study conducted across multiple well sites, the system's machine learning model—trained on integrated cumulative compliance metrics from tens to hundreds of rate change events per fracturing stage, sufficient to capture fracture physics across multiple growth regimes, demonstrated the capability to forecast post-fracturing production volumes at 1 month to 12 month intervals during active stimulation operations with prediction accuracy within 3% of subsequently measured production data. This forecasting capability, enabled by the granular event-dense analysis (detecting and integrating compliance responses from 20 to over 80 micro-events per stage, compared to conventional step-rate tests which typically employ 3-5 discrete rate steps), represents a substantial improvement over conventional real-time monitoring systems, which typically lack sufficient event density or physics-informed ML integration to achieve such predictive accuracy during operations. The ability to generate accurate production forecasts during fracturing—rather than weeks or months post-completion-enables real-time economic optimization decisions that were previously infeasible with diagnostic injection tests or lower-resolution monitoring approaches. These field results demonstrate that the combination of event-dense compliance calculation (via automatic detection of naturally occurring rate variations), continuous integration over stimulation intervals, and ML training on the resulting cumulative metrics yields a technically and economically meaningful advancement over prior art approaches limited to pre-frac diagnostics or sparse event analysis.

[0101] The capability to process dozens to over one hundred compliance-related events per stimulation stage distinguishes the disclosed approach from conventional diagnostic injection tests. Traditional step-rate tests (SRT) or mini-frac procedures typically involve 3 to 5 predetermined rate changes with substantial durations (e.g., 10-30 minutes per rate step), requiring operational interruptions and yielding only a sparse sampling of formation response. In contrast, the disclosed systems continuously monitor naturally occurring rate variations during main fracturing stages—such as minor fluctuations from equipment cycling, proppant additions, or intentional micro-adjustments—detecting these as compliance-relevant events at sampling rates of 0.1 Hz or higher. By automatically flagging hundreds of such micro-events per stage, iteratively fitting pressure response models (e.g., exponential decay functions with error-thresholded window optimization) to transient pressure data following each detected rate change, and integrating the resulting compliance values continuously over the stimulation interval (e.g., via numerical integration ∫(dQ / dP) dt or discrete summation Σ(ΔQ / ΔPfit)), the system generates a cumulative compliance metric with resolution and fidelity unattainable through sparse diagnostic testing. This event-dense, integration-based approach enables the machine learning model to be trained on datasets comprising thousands of stage-level cumulative metrics correlated with production outcomes, thereby learning physics-informed relationships (e.g., formation compliance evolution, fracture connectivity patterns, interference signatures) that generalize across wells and formations. The ML model, once trained, can then predict compliance and production in real-time without requiring any discrete rate change events, as the model has internalized the underlying fracture response physics from the high-resolution training data. This combination—event density during training, continuous integration for cumulative metric generation, and ML-enabled event-free prediction during deployment—represents a novel integration of data acquisition scale, mathematical aggregation, and predictive modeling not disclosed or suggested by prior art focused on sparse diagnostic tests or qualitative real-time monitoring.

[0102] A critical distinction of the disclosed approach lies in the recognition that integrating compliance-related values into a cumulative metric—rather than treating individual compliance measurements as discrete, independent snapshots—yields information about fracture system behavior and production potential that is not apparent from individual measurements alone. Prior art compliance analysis (e.g., step-rate tests, mini-frac analysis) generates discrete compliance values (C1, C2, C3, etc.) at specific rate steps, which are interpreted individually to infer fracture geometry, leak-off coefficients, or closure stress at that moment in time. These discrete measurements are not accumulated or integrated; each is treated as providing independent information about instantaneous formation response. The prior art provides no teaching or suggestion that summing, integrating, or otherwise accumulating compliance values over a stimulation stage would generate a cumulative metric with predictive value for production outcomes. Furthermore, even if a person of ordinary skill in the art (POSA) considered integrating compliance measurements, the practical barriers were substantial: no software infrastructure existed to systematically detect tens to hundreds of naturally occurring events per stage, fit pressure response models to each with error-based convergence, calculate event-specific compliance values, and integrate these continuously over the stimulation interval in real-time during main-stage fracturing operations. The disclosed systems overcome these barriers through automated event detection, iterative model fitting, and continuous numerical integration (e.g., > (ΔQ / ΔPfit× At)), enabling generation of cumulative compliance metrics that, when used to train machine learning models on historical data from hundreds to thousands of stages, demonstrate correlation with production outcomes and achieve real-time forecasting accuracy (e.g., within 3% for 6-month production predictions) previously unattainable. The transformation from discrete, stepwise compliance measurements treated as independent information to continuous, integrated cumulative metrics with demonstrated production predictivity represents a non-obvious advance that required both the conceptual recognition of integration's value and the systematic implementation infrastructure to realize it at scale during active fracturing operations.

[0103] The event-dense cumulative compliance generation methodology described herein itself represents a substantial advancement over prior art diagnostic approaches, independent of the machine learning prediction capabilities built upon it. Prior to the disclosed systems, no practical infrastructure existed to systematically detect tens to over one hundred naturally-occurring rate change micro-events per fracturing stage, iteratively fit pressure response models to each event with error-based convergence, calculate event-specific compliance values, and integrate these continuously into a cumulative metric during uninterrupted main-stage fracturing operations. Conventional diagnostic injection tests (e.g., step-rate tests, mini-fracs) required operational interruptions, employed predetermined rate sequences yielding only 3 to 5 discrete measurements, and provided no software or methodological framework for accumulating compliance measurements over stimulation intervals. The disclosed manual or automated event detection algorithms, iterative curve-fitting techniques, and continuous numerical integration infrastructure overcome practical barriers that prevented prior art from generating event-dense cumulative compliance metrics. This cumulative metric generation capability—yielding stage-level compliance values representing accumulated formation response over entire stimulation intervals rather than isolated snapshots at predefined test points—constitutes a technical advancement that enables subsequent applications including production correlation analysis and machine learning model training. The recognition that such event-dense cumulative metrics could be generated through systematic software infrastructure, and that these cumulative metrics would provide information about fracture system behavior and production potential not apparent from discrete measurements alone, required both conceptual insight and substantial implementation effort to realize at scale during active fracturing operations. Thus, the disclosed systems and methods advance the art in multiple dimensions: first, by enabling automated generation of event-dense cumulative compliance metrics through systematic detection, fitting, and integration infrastructure not present in prior art; and second, by training machine learning models on these cumulative metrics with physics-informed formation features to enable continuous real-time prediction capabilities that further enhance operational utility beyond the cumulative metric generation itself.

[0104] The disclosed systems and methods implement a multi-stage technical workflow that transforms discrete, event-based compliance measurements into continuous, real-time compliance predictions with demonstrated production forecasting capability. The workflow comprises several inventive steps addressing practical operational constraints and technical challenges not resolved by prior art:

[0105] First, raw time-series pressure and injection rate data acquired during hydraulic fracturing operations are analyzed to detect rate change events, typically yielding tens of events per stimulation stage (e.g., 20 to 80 or more events capturing breakdown, pad injection, proppant ramp-up phases, mid-stage rate adjustments, and equipment cycling). For each detected event, a pressure response model (e.g., exponential decay function) is iteratively fitted to transient pressure data following the rate change to determine a fitted pressure parameter representing the asymptotic pressure (P∞) that would be reached if sufficient time elapsed. This curve-fitting approach addresses a critical operational constraint: hydraulic fracturing operations proceed continuously without waiting for pressure stabilization, necessitating estimation of Poo via mathematical modeling rather than direct measurement. The fitted Poo for each event, combined with the corresponding rate change (ΔQ), enables calculation of an event-specific compliance measurement (ΔQ / ΔP∞). The discrete compliance measurements for all detected events within a stage are summed to generate a cumulative compliance value (C-Factor) representing the total accumulated compliance for that stimulation stage.

[0106] Second, a machine learning model—such as a neural network—is trained using historical data comprising the cumulative compliance values (C-Factors) calculated for completed fracturing stages, along with corresponding time-series pressure and rate data and physics-informed features that provide contextual information explaining variation in compliance behavior. These features may include formation characteristics derived from measurement-while-drilling (MWD) data (e.g., gamma ray logs, distance to geological formation tops or boundaries, lithological indicators), well trajectory information, and operational parameters. Critically, the neural network is trained to predict the cumulative compliance value (C-Factor) at any given time during a fracturing stage based on the pressure, rate, and formation data available up to that time. By incorporating these physics-informed features, the machine learning model learns relationships between formation properties, operational conditions, and cumulative compliance evolution that are not apparent from pressure and rate data alone, enabling the model to account for well-to-well and stage-to-stage variability attributable to geological heterogeneity.

[0107] Third, the trained machine learning model is deployed during subsequent hydraulic fracturing operations to generate cumulative compliance predictions in real-time at high temporal resolution (e.g., predictions at 1-second intervals or higher frequency). Remarkably, the model, having been trained on cumulative compliance values that were originally derived from discrete compliance measurements (calculated from tens of events per stage, each spanning durations of approximately 6 to 300 seconds), learns the underlying physics governing formation response and directly predicts the cumulative compliance value—the running total of accumulated compliance—at each time point throughout the stimulation operation. The model does not generate individual instantaneous compliance measurements that must then be integrated; rather, it generates the cumulative value directly, as it was trained to predict the integrated sum. This transformation from training data based on discrete, manually flagged events (30-80 per stage) to deployment predictions that output cumulative compliance continuously (second-by-second throughout the stage) represents a technical achievement that prior art discrete compliance analysis (e.g., step-rate tests yielding 3 to 5 discrete compliance snapshots) cannot accomplish.

[0108] Fourth, cumulative compliance metrics may be aggregated across multiple stages within a well (e.g., by summing maximum or final compliance values from each stage) to generate a well-level cumulative compliance value, as contemplated by the well and field management strategies disclosed herein (e.g.,

[0012] ). These well-level values, when correlated with measured production data from a plurality of historically completed wells across a field or geological formation, enable construction of a production profile or predictive equation relating cumulative compliance to production outcomes at specified time intervals (e.g., 30-day, 180-day, or 365-day cumulative production). When a new well is fractured, the real-time cumulative compliance metric calculated during active stimulation operations can be input into the production profile equation to generate production forecasts—enabling operators to predict production outcomes weeks or months in advance, during the fracturing operation itself, rather than waiting for post-completion production data. Field implementations have demonstrated the capability to forecast 6-month production volumes with accuracy within approximately 3% of subsequently measured production, enabling real-time economic optimization decisions (e.g., whether to continue pumping, adjust stage parameters, or terminate operations based on diminishing compliance returns and cost-benefit projections) during active fracturing.

[0109] This multi-stage workflow—comprising event detection and pressure curve fitting for discrete compliance calculation during training data generation, machine learning training with physics-informed features, real-time continuous prediction of cumulative compliance values via trained model deployment, hierarchical aggregation from stage to well level, field-level production profile construction, and real-time production forecasting for ongoing operations—represents a systematic technical solution to challenges in hydraulic fracturing analysis that prior art approaches treating compliance as discrete, independent measurements for geometry estimation do not address. The disclosed integration of these steps enables capabilities (continuous cumulative compliance prediction, real-time production forecasting during fracturing, economic optimization during operations) that are not achievable through prior art discrete compliance analysis or generic real-time monitoring systems lacking the physics-informed machine learning framework. In some cases, the C-Factor Calculation System 900 may be implemented to provide real-time compliance analysis during hydraulic fracturing operations. The C-Factor Calculation System 900 may integrate over the course of a frac into a new parameter called C-Factor. In some implementations, the C-Factor Calculation System 900 may measure the second-by-second or minute-by-minute changes in injection rate relative to pressure changes, providing a real-time measure of fracture system compliance.

[0110] The Data Acquisition Layer 902 of the C-Factor Calculation System 900 may collect real-time data from various sensors. In some cases, the Pressure Sensors 904 and Flow Rate Sensors 906 may provide continuous measurements of injection rate and pressure during fracturing at a selected sampling interval. The Distributed Acoustic Sensing (DAS) System 908 may capture high-resolution acoustic data, which may be integrated with the injection rate and pressure measurements to provide additional insights into fracture behavior.

[0111] As used herein, a “machine learning model” refers to a mathematical and / or algorithmic representation of a relationship between inputs and outputs, as generated using a machine learning process and stored in memory. A “trained machine learning model” refers to a machine learning model that has been trained using training data to learn relationships between input parameters and output values. Training a machine learning model may involve applying elements from a training dataset to input nodes or input parameters of the model, and using a training algorithm to adjust connections, weights, coefficients, or other parameters within the model to produce desired output values. Once trained, an input may be submitted to the trained machine learning model, which generates an output based on the learned relationships.

[0112] In some implementations, the trained machine learning model may comprise a neural network. A neural network may include an input layer of nodes, one or more intermediate layers, and an output layer of nodes, where connections between nodes are created and adjusted during training. Training may involve applying training data elements to input nodes and using a training algorithm to adjust connections and weights between nodes in adjacent layers to produce desired values at output nodes. In some implementations, the machine learning model may be trained using supervised learning, wherein training data includes labeled input-output pairs, such as historical sensor measurements paired with corresponding cumulative compliance values derived from previously completed fracturing stages. The trained model may then predict cumulative compliance values for new input data based on the learned relationships.

[0113] Supervised machine-learning process include algorithms that receive a training set relating a number of inputs to a number of outputs, and seek to find one or more mathematical relations relating inputs to outputs.

[0114] The Data Processing Layer 918 may utilize the High-Performance Computing Cluster 920 to process the acquired sensor data. In some implementations, the CPU Nodes 922 and GPU Nodes 924 may work in parallel to perform real-time calculations of compliance values. The compliance value C for each rate change event may be computed analytically from changes in injection rate relative to pressure change using the equation:C=Δ⁢QΔ⁢Pfitwhere C represents the compliance value for the rate change event, ΔQ represents the change in injection rate associated with the rate change event, and ΔP_fit represents the fitted pressure change determined from the mathematical function fitted to the subset of pressure measurements following the rate change event.In various implementations, the subset of pressure measurements may be selected to capture the pressure response behavior during a fracture system's stabilization period following a rate change event. The subset may be determined by identifying a pressure minimum within the pressure measurements recorded after detection of the rate change event, and selecting pressure measurements extending from the identified pressure minimum through a subsequent pressure maximum, thereby isolating the pressure stabilization curve. This selective windowing may focus the mathematical function fitting on the portion of the pressure response that exhibits characteristic exponential stabilization behavior, while excluding initial transient effects or subsequent operational changes that may not reflect the fracture system's compliance response to the specific rate change event. In some cases, the subset may be dynamically adjusted during an iterative fitting process to improve fit quality or satisfy predetermined convergence criteria. For example, if initial fitting parameters indicate physically unrealistic behavior (such as excessively long time constants or negative decay rates), the analysis window may be progressively narrowed by removing later pressure measurements until the fitted parameters fall within physically plausible ranges. The subset selection may involve excluding measurements recorded before the pressure minimum (which may represent pre-event baseline conditions or declining pressure from prior activities) and measurements recorded substantially after the pressure maximum (which may represent subsequent rate changes, pump changes, or other operational or reservoir events unrelated to the compliance response being analyzed). In various embodiments, the subset may comprise measurements spanning timeframes ranging from several seconds to several minutes, with the specific duration determined by the observed stabilization behavior of the fracture system for the particular rate change event. This targeted subset selection may enable more accurate compliance calculations by focusing the curve fitting algorithm on pressure data that directly reflects the fracture system's response to a specific injection rate change, thereby improving the signal-to-noise ratio and reducing interference from extraneous operational factors. Where ΔQ represents the change in injection rate and ΔPfit represents the fitted pressure change.

[0116] In some implementations, historical training data for the machine learning model may be generated through manual analysis of previously completed fracturing operations. This process may involve identifying rate change events in historical stimulation data, fitting mathematical models (such as exponential decay functions) to pressure responses following each event to determine fitted pressure parameters, calculating compliance values for each event (e.g., ΔQ / ΔPfit), and integrating or summing these discrete compliance values over a stimulation stage to generate cumulative compliance values that serve as training labels. These manually generated cumulative compliance values, paired with corresponding sensor data and formation characteristics, may be used to train the machine learning model to recognize patterns and relationships between operational parameters, formation properties, and fracture system response.

[0117] The Real-Time Stream Processing 926 component may continuously integrate these compliance values over time to form the C-Factor. This integration process may involve summing or applying weighted averaging techniques to combine multiple compliance values into a composite measure.

[0118] In some aspects, the Machine Learning Layer 930 may employ machine learning algorithms to analyze the real-time data and generate predictions. The Model Training Module 932 may be trained using historical hydraulic fracturing data, including injection rates, pressure measurements, and corresponding compliance values. The Real-Time Inference Engine 934 may then use this trained model to compute C-Factor in real-time, even without the need for discrete step tests.

[0119] The Feature Extraction Pipeline 936 may incorporate signal processing techniques to analyze frequency components of the pressure measurements. In some implementations, this analysis may be used to identify dominant frequencies associated with rock breakage during the hydraulic fracturing operation.

[0120] The Visualization and Control Layer 938 may present the calculated C-Factor and other relevant data through the Real-Time Dashboard 940. This may allow operators to visualize and interpret the compliance data in real-time. The Automated Control System 942 may use the real-time C-Factor predictions to adjust fracturing parameters such as injection rate, proppant concentration, or fluid composition.

[0121] In some cases, the Alert Management System 944 may be configured to detect and flag negative compliance events quickly. For instance, if a sudden decrease in compliance is observed, it may trigger an alert prompting adjustments to the fracturing process to optimize fracture growth and prevent potential issues such as screenouts or undesired fluid diversion.

[0122] The C-Factor Calculation System 900 may characterize fracture system behavior to differentiate between beneficial fracture growth and undesirable fluid diversion. The system may monitor cumulative compliance evolution during stimulation operations. Increasing cumulative compliance values may indicate active fracture development and effective access to fresh formation intervals. Observed changes in compliance behavior may inform operational decisions, enabling adjustments to maximize productive fracture growth. In some aspects, the machine learning framework 1020 may utilize the tensor processing module 1022 and neural network module 1024 to implement advanced predictive models. These models may analyze patterns in the compliance data to forecast fracture behavior and optimize fracturing operations in real-time.

[0123] The visualization layer 1028 may leverage the graphics library 1030 and plotting module 1032 to create dynamic, real-time visualizations of the C-Factor and related parameters. This may enable operators to monitor the fracturing process and make data-driven decisions.

[0124] In some implementations, the control system layer 1034 may use the protocol interface 1036 to communicate with the hydraulic fracturing equipment, allowing for automated or semi-automated adjustments to fracturing parameters based on the real-time C-Factor analysis.

[0125] The C-Factor Calculation System 900 may provide several advantages in hydraulic fracturing operations. By enabling real-time monitoring and analysis, operators may no longer need to wait until the stage's conclusion or final production data to realize if a fracture is drifting into offset wells or not achieving vertical growth. This real-time insight may lead to improved resource utilization, as the frac can be better targeted to the main zone, potentially improving initial production (IP) and estimated ultimate recovery (EUR).

[0126] In some cases, the system's ability to analyze negative compliance “dips” may allow for the identification of partial infiltration events, differentiating them from fully beneficial large compliance expansions into fresh zones. This detailed analysis may provide operators with a more nuanced understanding of fracture behavior and help identify potential issues such as fluid diversion into undesired zones.

[0127] The integration of machine learning techniques in the C-Factor Calculation System 900 may enable the development of predictive models for fracture behavior and well performance. These models may assist in real-time decision-making during fracturing operations and in the planning of future wells, potentially leading to optimized fracturing strategies and improved overall field development.

[0128] In some cases, the method for real-time compliance factor prediction in hydraulic fracturing operations may be implemented as illustrated in FIG. 6. FIG. 6 depicts a sequence diagram showing the interactions between components of the system, including at least one sensor, at least one processor, a display device, a machine learning model, and hydraulic fracturing equipment.

[0129] The process may begin with a real time data step S1, where real-time data, including injection rate and pressure measurements, may be transmitted from the at least one sensor to the at least one processor. The at least one processor may then execute a series of sequential calculations and analyses.

[0130] A rate change detection step S2 may detect rate change events in the data. Following detection of a rate change event, an injection rate calculation step S3 may calculate the change in injection rate (ΔQ).

[0131] In parallel with the injection rate calculation, a mathematical function fitting step S4 may fit mathematical functions to pressure measurements. In some cases, the mathematical function fitting step S4 may include iteratively fitting a pressure response model to the subset of pressure measurements. This iterative process may involve adjusting the analysis window and recalculating error values until a predetermined threshold is met or a maximum number of iterations is reached.

[0132] A pressure change determination step S5 may determine the fitted pressure change (ΔPfit) based on the mathematical function. A compliance value calculation step S6 may then calculate the compliance value (ΔQ / ΔPfit).

[0133] A compliance integration step S7 may integrate compliance values to generate a cumulative compliance factor. The at least one processor may then execute a compliance factor output step S8, outputting the cumulative compliance factor to the display device.

[0134] In an operational parameter adjustment step S9, operational parameters may be adjusted and sent to the hydraulic fracturing equipment based on the calculated compliance factor.

[0135] Concurrently with the traditional compliance calculation path, the machine learning model may execute a parallel processing path. A model training step S10 may train the machine learning model using historical hydraulic fracturing data. A current data input step S11 may receive current injection rate and pressure data.

[0136] The machine learning model may then execute a compliance prediction step S12, generating real-time predicted compliance factors. These predictions may be output to the display device in a prediction output step S13.

[0137] In some cases, the system may calculate an event-based alpha value for each rate change event. The event-based alpha value may be computed as a ratio of the fitted pressure change (ΔPfit) to the change in injection rate (ΔQ). The system may integrate a plurality of calculated event-based alpha values to generate a cumulative alpha factor.

[0138] The system may also calculate an instantaneous shut-in pressure (ISIP) alpha value when an ISIP event is detected in the real-time data. The ISIP alpha value may be computed as a ratio of a pressure change during the ISIP event to a corresponding change in injection rate.

[0139] In some implementations, the system may integrate distributed acoustic sensing (DAS) data from a fiber optic cable deployed in the wellbore with the injection rate and pressure measurements. The system may use the trained machine learning model to identify locations of fluid entry into the formation based on patterns in the DAS data correlated with changes in the real-time predicted compliance factor.

[0140] The system may generate a dynamic fracture propagation visualization showing estimated fracture growth over time correlated with the identified fluid entry locations and the real-time predicted compliance factor.

[0141] Finally, a parameter adjustment step S14 may send additional operational parameter adjustments to the hydraulic fracturing equipment based on the machine learning predictions.

[0142] This sequence diagram illustrates the parallel processing paths between traditional compliance calculation (steps S2-S7) and machine learning prediction (steps S10-S12), both contributing to operational decisions through steps S9 and S14 respectively. The integration of these two approaches may provide a comprehensive analysis of fracture behavior and enable real-time optimization of hydraulic fracturing operations.

[0143] In some cases, a compliance calculation process 700 may be implemented for real-time analysis of hydraulic fracturing operations. FIG. 7 illustrates a flowchart of the compliance calculation process 700, which may include multiple interconnected modules that process real-time data to calculate and utilize compliance factors during hydraulic fracturing operations.

[0144] The compliance calculation process 700 may begin with a rate change detection module 702 that receives input from a data stream module 716. The data stream module 716 may provide a continuous flow of real-time data from sensors at the hydraulic fracturing site. The rate change detection module 702 may analyze this data to identify significant changes in injection rate that warrant further analysis.

[0145] Upon detecting a rate change event, the process 700 may branch into two parallel processing paths. In the first path, an injection rate calculation module 704 may determine the magnitude of changes in injection rate (ΔQ). The output from the injection rate calculation module 704 may feed into a compliance value computation module 712.

[0146] In the second path, a pressure response analysis module 706 may process the pressure data associated with the detected rate change event. Within the pressure response analysis module 706, a mathematical function fitting module 708 may fit a mathematical function, such as an exponential curve, to the pressure data. The output from the mathematical function fitting module 708 may flow to a pressure change determination module 710, which may calculate the fitted pressure change (ΔPfit) based on the mathematical function. The pressure change determination module 710 may then provide input to the compliance value computation module 712.

[0147] The compliance value computation module 712 may process inputs from both paths to calculate compliance values. In some cases, the compliance value may be computed as the ratio of the change in injection rate to the fitted pressure change (ΔQ / ΔPfit). These calculated compliance values may then flow to a cumulative compliance integration module 714, which may generate integrated compliance factors over time.

[0148] In some implementations, the compliance calculation process 700 may include a step of filtering the calculated compliance values to exclude values greater than a predetermined threshold. This filtering step may help remove outliers or anomalous data points that could skew the analysis. The predetermined threshold may be set based on historical data or expert knowledge of the expected range of compliance values for the specific formation being fractured.

[0149] Following the filtering step, the process 700 may include binning the filtered compliance values and corresponding leak-off data by hydraulic fracturing stage. This binning process may allow for stage-specific analysis of fracture behavior and fluid efficiency during the hydraulic fracturing operation.

[0150] The output from the cumulative compliance integration module 714 may connect to both a display interface module 718 and a parameter adjustment module 720. The display interface module 718 may present the compliance data and other relevant information to operators through visual representations such as graphs, charts, or numerical displays.

[0151] The parameter adjustment module 720 may use the compliance data to modify fracturing parameters in real-time. In some cases, the system may not only provide human-readable alarms but also auto-adjust fracturing parameters such as flow rates, fluid type, or offset well back pressure based on the calculated compliance signals. This automated adjustment capability may allow for rapid optimization of the fracturing process in response to changing subsurface conditions.

[0152] In some implementations, the compliance calculation process 700 may generate a summary statistics report for the hydraulic fracturing operation. This report may include information such as the total number of rate change events processed, the distribution of compliance values across different stages, and any significant trends or anomalies detected during the operation.

[0153] The process 700 may also be capable of measuring compliance signals when the injection rate is constant but the pressure changes or oscillates. This capability may provide additional insights into formation behavior and fracture propagation even during periods of steady-state injection.

[0154] In some cases, geological or drilling metadata, such as well trajectory, formation tops, or stress logs, may be loaded into the model before fracturing begins. This pre-loaded data may be used to anticipate likely negative compliance intervals and optimize stage design, potentially improving the overall efficiency of the fracturing operation.

[0155] By analyzing negative compliance ‘dips,’ the process 700 may flag partial infiltration events, differentiating them from fully beneficial large compliance expansions into fresh zones. This detailed analysis may provide operators with a more nuanced understanding of fracture behavior and help identify potential issues such as fluid diversion into undesired zones.

[0156] The compliance calculation process 700 may incorporate continuous data flow from the data stream module 716, allowing for ongoing monitoring and adjustment of the hydraulic fracturing operation. The modules may be arranged to enable parallel processing of rate and pressure data, with the results converging at the compliance value computation module 712 for integrated analysis. This architecture may allow for efficient, real-time calculation and utilization of compliance factors throughout the hydraulic fracturing operation.

[0157] In some cases, a hydraulic fracturing system 800 may be implemented for performing hydraulic fracturing operations. FIG. 8 illustrates a block diagram of the hydraulic fracturing system 800, which may include several interconnected components.

[0158] The hydraulic fracturing system 800 may include a target well 802 into which fracturing fluid may be injected. A pump 804 may be connected to the target well 802 to provide the necessary pressure for fluid injection. The pump 804 may receive fluid from a fluid reservoir 806 and proppant from a proppant supply 808. In some implementations, a mixing unit 810 may be used to combine the fluid and proppant before delivery to the pump 804.

[0159] The hydraulic fracturing system 800 may incorporate monitoring components to measure various parameters during the fracturing operation. These monitoring components may include a pressure sensor 812 and a flow rate sensor 814. The pressure sensor 812 may measure the injection pressure at various points in the system, while the flow rate sensor 814 may measure the rate of fluid injection into the target well 802.

[0160] A control unit 816 may be connected to the monitoring components and other system elements. The control unit 816 may process the sensor data and manage the operation of the system components. In some cases, the control unit 816 may implement the real time data step, rate change detection step, injection rate calculation step, mathematical function fitting step, pressure change determination step, compliance value calculation step, and compliance integration step as described in previous sections.

[0161] A display 818 may be connected to the control unit 816 for presenting operational data and system status information. The display 818 may be used to implement the compliance factor output step, allowing operators to visualize compliance factors and other relevant data in real-time.

[0162] The hydraulic fracturing system 800 may also include a machine learning module 820 connected to the control unit 816. The machine learning module 820 may receive data from the sensors and provide predictive analytics to assist in system operation. In some implementations, the machine learning module 820 may perform the model training step, current data input step, compliance prediction step, and prediction output step.

[0163] In some cases, the historical hydraulic fracturing data used by the machine learning module 820 may further comprise subterranean formation data. This subterranean formation data may include well log data, core sample measurements, seismic or microseismic data, porosity or permeability measurements, or formation pressure data. By incorporating this additional data, the machine learning module 820 may develop more comprehensive models for predicting compliance factors and optimizing fracturing operations.

[0164] The control unit 816 may analyze frequency components of the pressure measurements using a Fourier transform. This analysis may be performed on the data collected by the pressure sensor 812. By examining the frequency domain representation of the pressure data, the control unit 816 may identify dominant frequencies associated with rock breakage during the hydraulic fracturing operation.

[0165] In some implementations, the control unit 816 may correlate the identified dominant frequencies with geological characteristics of the formation being fractured. This correlation may provide insights into the fracturing process and the response of different geological layers to the applied pressure.

[0166] Based on the frequency analysis and correlation with geological characteristics, the control unit 816 may adjust operational parameters of the hydraulic fracturing system 800. For example, the control unit 816 may adjust the injection rate controlled by the pump 804 or modify the fluid composition in the mixing unit 810. These adjustments may be made to target specific frequency ranges associated with optimal fracture propagation in the formation.

[0167] The system components may be arranged to enable fluid flow from the fluid reservoir 806 and proppant supply 808 through the mixing unit 810, which then feeds into the pump 804 for injection into the target well 802. The pressure sensor 812 and flow rate sensor 814 may monitor the injection process, providing real-time data to the control unit 816 for analysis and display.

[0168] By integrating these various components and leveraging the capabilities of the machine learning module 820, the hydraulic fracturing system 800 may enable real-time monitoring, analysis, and optimization of fracturing operations based on calculated compliance factors and other relevant parameters.

[0169] FIG. 2 illustrates a flowchart for a machine learning process to predict real-time compliance factors during hydraulic fracturing operations. The process begins with a model training step 200, where a machine learning model may be trained using historical hydraulic fracturing data. In some cases, the historical data may include injection rates, pressure measurements, and corresponding compliance values. The machine learning model may also be trained using subterranean formation data, which may include well log data, core sample measurements, seismic or microseismic data, porosity or permeability measurements, or formation pressure data.

[0170] Following the model training step 200, the process moves to a current data input step 202. In this step, current injection rate and pressure measurements may be received as input to the trained machine learning model. These real-time measurements may provide the model with up-to-date information about the ongoing hydraulic fracturing operation.

[0171] The process then continues to a compliance prediction step 204. In this step, the system may generate a real-time predicted compliance factor using the trained machine learning model. In some cases, the machine learning model may generate the real-time predicted compliance factor without requiring discrete rate change events. This capability may allow for continuous prediction of compliance factors throughout the fracturing operation.

[0172] In some implementations, the machine learning model may simultaneously generate a real-time predicted alpha factor along with the real-time predicted compliance factor. The alpha factor may provide additional information about the formation's response to the fracturing process.

[0173] After generating the predictions, the process moves to a prediction output step 206. In this step, the predicted compliance factor may be output to a display device. This real-time display of compliance predictions may allow operators to monitor the fracturing process and make informed decisions.

[0174] The final step in the process is a parameter adjustment step 208. In this step, operational parameters of the hydraulic fracturing operation may be adjusted based on the real-time predicted compliance factor. These adjustments may include changes to injection rate, proppant concentration, or fluid composition. By making these adjustments in response to real-time predictions, the system may optimize the fracturing process and improve overall efficiency.

[0175] The machine learning process illustrated in FIG. 2 may provide a continuous feedback loop for hydraulic fracturing operations. By combining historical data analysis with real-time predictions and parameter adjustments, the system may enable more precise control and optimization of the fracturing process.

[0176] In some implementations, the machine learning models used for predicting compliance factors and optimizing hydraulic fracturing operations may be trained using a variety of techniques and data sources. The training process may involve several steps and considerations to ensure the models can accurately predict compliance factors and provide valuable insights for real-time decision-making.

[0177] The model training step may begin with data collection and preprocessing. Historical hydraulic fracturing data from multiple wells and operations may be gathered, including injection rates, pressure measurements, and corresponding compliance values. This data may be cleaned, normalized, and formatted to ensure consistency and reliability for model training.

[0178] In some cases, the training data may also include subterranean formation data such as well log data, core sample measurements, seismic or microseismic data, porosity or permeability measurements, and formation pressure data. By incorporating this geological information, the models may develop a more comprehensive understanding of the relationship between formation characteristics and compliance factors.

[0179] Feature engineering may be performed to extract relevant information from the raw data. This process may involve creating new features or transforming existing ones to capture important aspects of the hydraulic fracturing process. For example, the system may calculate rate change events, pressure derivatives, or frequency domain features from the time-series data.

[0180] The machine learning models may employ various algorithms depending on the specific requirements of the prediction task. In some implementations, supervised learning techniques such as random forests, gradient boosting machines, or deep neural networks may be used. These models may be trained to map input features (e.g., injection rates, pressure measurements, geological data) to output predictions (e.g., compliance factors, alpha factors).

[0181] Cross-validation techniques may be utilized to assess the model's performance and prevent overfitting. The training data may be split into training, validation, and test sets to evaluate the model's generalization capabilities. Hyperparameter tuning may be performed using techniques such as grid search, random search, or Bayesian optimization to find the optimal model configuration.

[0182] In some cases, ensemble methods may be employed to combine predictions from multiple models. This approach may improve the overall accuracy and robustness of the predictions by leveraging the strengths of different algorithms.

[0183] The training process may also incorporate techniques to handle the temporal nature of the hydraulic fracturing data. Recurrent neural networks (RNNs) or long short-term memory (LSTM) networks may be used to capture time-dependent patterns and relationships in the data.

[0184] Transfer learning techniques may be applied to leverage knowledge gained from training on one dataset to improve performance on related tasks or datasets. This approach may be particularly useful when dealing with limited data for specific well types or geological formations.

[0185] The machine learning models may be designed to handle real-time data streams and provide continuous predictions during hydraulic fracturing operations. Online learning algorithms may be implemented to allow the models to adapt and improve their predictions as new data becomes available during the fracturing process.

[0186] In some implementations, the training process may include techniques to quantify uncertainty in the model predictions. Bayesian neural networks or ensemble methods may be used to provide confidence intervals or probability distributions for the predicted compliance factors.

[0187] The training process may also focus on developing interpretable models that can provide insights into the factors influencing compliance predictions. Techniques such as SHΔP (SHapley Additive explanations) values or LIME (Local Interpretable Model-agnostic Explanations) may be used to explain individual predictions and identify the most important features driving the model's decisions.

[0188] Regularization techniques, such as L1 or L2 regularization, may be applied during training to prevent overfitting and improve the model's generalization capabilities. Dropout layers may be incorporated in neural network architectures to further enhance robustness and reduce overfitting.

[0189] In some cases, the training process may involve curriculum learning, where the model is initially trained on simpler examples before progressing to more complex scenarios. This approach may help the model develop a better understanding of the underlying patterns in the data.

[0190] The machine learning models may be periodically retrained or fine-tuned as new data becomes available. This ongoing learning process may help the models adapt to changes in formation characteristics, equipment configurations, or operational practices over time.

[0191] In some cases, a method for manually monitoring hydraulic fracturing operations may be implemented as illustrated in FIG. 3. FIG. 3 depicts a flowchart illustrating a sequence of steps for monitoring and adjusting hydraulic fracturing operations based on calculated compliance values.

[0192] The method may begin with a step 300 where injection rate and pressure are continuously measured during the fracturing operation at a selected sampling interval. In some implementations, a data acquisition subsystem may receive these fracturing pressure and rate measurements from sensors deployed in the wellbore and surface equipment.

[0193] Following the continuous measurement, a step 302 may involve using a processor to analytically compute a compliance value C from changes in injection rate or fluid volume relative to pressure change. The compliance value C may be calculated as the ratio of injection rate change to pressure change (ΔQ / ΔP) or as the ratio of volume change to pressure change (ΔV / ΔP) for specific rate change events detected in the measured data. In real-time implementations, the rate-based formulation (ΔQ / ΔP) may enable continuous compliance monitoring without requiring integration of flow rates to determine cumulative volumes.

[0194] A step 304 may then integrate the calculated compliance values over time to form a C-Factor. In some cases, a summation module may perform this integration, accumulating individual compliance values to provide a cumulative measure of formation response over the course of the fracturing operation.

[0195] The method may include a step 306 where the C-Factor is displayed on a user interface. In some implementations, a display device may present the C-Factor to operators, allowing for real-time visualization of the cumulative compliance behavior during the fracturing process.

[0196] Finally, a step 308 may involve adjusting fracturing parameters based on the displayed C-Factor to optimize reservoir stimulation. In some cases, a control panel may allow operators to manually adjust parameters such as injection rate, proppant concentration, or fluid composition based on their interpretation of the C-Factor trends.

[0197] By following this sequence of steps, the method may enable manual monitoring and optimization of hydraulic fracturing operations using real-time compliance calculations. The integration of compliance values into a C-Factor may provide a broader perspective on fracture behavior compared to individual compliance measurements, potentially allowing for more informed decision-making during the stimulation process.

[0198] FIG. 4 illustrates a flowchart for an AI-based method for monitoring hydraulic fracturing operations. The method begins with a step 400 where real-time injection, pressure, and offset well data are collected. In some cases, the data acquisition may involve sensors deployed at the wellsite to measure various parameters relevant to the hydraulic fracturing process.

[0199] The process then moves to a step 402 where the collected data is fed into a trained machine learning model. In some implementations, the machine learning model may have been previously trained on historical hydraulic fracturing data, including injection rates, pressure measurements, and corresponding compliance values.

[0200] In a step 404, the machine learning model generates a real-time C-Factor. The C-Factor, as previously defined, is a cumulative or integrated measure of compliance values over time or across multiple fracturing stages. By utilizing a machine learning approach, the method may compute the C-Factor in real time without requiring manual calculations or discrete rate change events.

[0201] The process continues to a step 406 where the real-time C-Factor is displayed on a user interface. In some cases, the display device may present the C-Factor along with other relevant operational parameters, allowing operators to visualize and interpret the data in real-time.

[0202] Finally, in a step 408, fracturing parameters are modified based on the C-Factor. This step may involve adjusting operational parameters such as injection rate, proppant concentration, or fluid composition in response to the real-time C-Factor predictions.

[0203] In some implementations, the method may utilize a data acquisition module to collect pressure, rate, and offset well data at specified sampling intervals. This module may interface with various sensors and measurement devices deployed at the wellsite to ensure accurate and timely data collection.

[0204] The machine learning processor used in the method may be configured to compute the C-Factor in real time based on the incoming data streams. This real-time computation may allow for rapid assessment of fracture behavior and formation response during the hydraulic fracturing operation.

[0205] By employing a machine learning model to compute the C-Factor in real time, the method may offer advantages over manual calculation methods. The AI-based approach may be able to identify complex patterns and relationships in the data that may not be immediately apparent through traditional analysis techniques.

[0206] The display device used in step 406 may be part of a larger visualization system that presents the real-time C-Factor along with other relevant operational data. This comprehensive display may enable operators to make informed decisions quickly based on the most up-to-date information available.

[0207] In some cases, the method may incorporate a control interface that allows for the adjustment of fracturing parameters based on the real-time C-Factor predictions. This interface may provide operators with the ability to fine-tune the fracturing process in response to changing subsurface conditions, potentially improving the overall efficiency and effectiveness of the hydraulic fracturing operation.

[0208] In some cases, a real-time compliance factor prediction system 500 may be implemented for hydraulic fracturing operations. FIG. 5 illustrates a block diagram of the compliance prediction system 500, which may include several interconnected modules and components.

[0209] The compliance prediction system 500 may include a data acquisition module 502 that receives input from multiple sensor types. The data acquisition module 502 may incorporate pressure sensors 504, flow rate sensors 506, a distributed acoustic sensing (DAS) system 508, and offset well sensors 510. These sensors may provide real-time data during hydraulic fracturing operations. In some cases, the pressure sensors 504 and flow rate sensors 506 may be used to measure injection rates and pressures, while the DAS system 508 may provide additional data on wellbore conditions. The offset well sensors 510 may collect pressure or flow data from nearby wells, which may be integrated into the compliance calculation to detect negative compliance events.

[0210] The system 500 may include a data processing module 512 that processes the acquired sensor data. The data processing module 512 may contain a rate change detector 514 for identifying rate change events, a curve fitting algorithm 516 for analyzing pressure responses, and a compliance calculator 518 for computing compliance values. In some implementations, the rate change detector 514 may identify a plurality of events in the real-time data and generate an event table. The curve fitting algorithm 516 may perform iterative fitting of a pressure response model to subsets of pressure measurements.

[0211] A machine learning module 520 may be connected to the data processing module 512. The machine learning module 520 may include historical data 522 storage and incorporate model training 524 capabilities. The machine learning module 520 may also contain a real-time prediction component 526 for generating compliance predictions. In some cases, the machine learning module 520 may be trained using historical hydraulic fracturing data to compute compliance factors in real-time.

[0212] An output module 528 may receive data from the machine learning module 520. The output module 528 may include a display device 530 for presenting compliance data and predictions, and a control interface 532 for system operation. The control interface 532 may connect to hydraulic fracturing equipment 534 through control signals.

[0213] The compliance prediction system 500 may process data flows between modules as indicated by solid arrows in the diagram. The data acquisition module 502 may send data to the data processing module 512, which may exchange information with the machine learning module 520. The machine learning module 520 may provide predictions to the output module 528, which may then control the hydraulic fracturing equipment 534 through the control interface 532.

[0214] In some implementations, the compliance prediction system 500 may comprise at least one processor and a memory storing instructions for executing the various modules and components described. The system may receive real-time data from the sensors during hydraulic fracturing operations and process this data to calculate compliance factors and generate predictions.

[0215] The compliance calculation may be refined using real-time downhole sensor readings, such as bottomhole pressure logs, to improve the accuracy of volume change (AV) calculations. In some cases, the bottomhole pressure used in the compliance computation may itself be predicted by an AI model trained on historical data, enabling continuous pressure estimation even in the absence of direct downhole gauges.

[0216] By leveraging this comprehensive system architecture, the compliance prediction system 500 may enable real-time monitoring, analysis, and optimization of hydraulic fracturing operations based on calculated compliance factors and C-factors.

[0217] In some cases, a real-time hydraulic fracturing compliance monitoring system may implement a data flow process as illustrated in FIG. 6. FIG. 6 depicts a sequence diagram showing the interactions between components of the system, including at least one sensor, at least one processor, a display device, a machine learning model, and hydraulic fracturing equipment.

[0218] The process may begin with a real time data step S1, where real-time data, including injection rate and pressure measurements, may be transmitted from the at least one sensor to the at least one processor. The at least one processor may then execute a series of sequential calculations and analyses.

[0219] A rate change detection step S2 may detect rate change events in the data. Following detection of a rate change event, an injection rate calculation step S3 may calculate the change in injection rate (ΔQ).

[0220] In parallel with the injection rate calculation, a mathematical function fitting step S4 may fit mathematical functions to pressure measurements. In some cases, the mathematical function fitting step S4 may include iteratively fitting a pressure response model to the subset of pressure measurements. This iterative process may involve adjusting the analysis window and recalculating error values until a predetermined threshold is met or a maximum number of iterations is reached.

[0221] A pressure change determination step S5 may determine the fitted pressure change (ΔPfit) based on the mathematical function. A compliance value calculation step S6 may then calculate the compliance value (ΔQ / ΔPfit).

[0222] A compliance integration step S7 may integrate compliance values to generate a cumulative compliance factor. The at least one processor may then execute a compliance factor output step S8, outputting the cumulative compliance factor to the display device.

[0223] In an operational parameter adjustment step S9, operational parameters may be adjusted and sent to the hydraulic fracturing equipment based on the calculated compliance factor.

[0224] Concurrently with the traditional compliance calculation path, the machine learning model may execute a parallel processing path. A model training step S10 may train the machine learning model using historical hydraulic fracturing data. A current data input step S11 may receive current injection rate and pressure data.

[0225] The machine learning model may then execute a compliance prediction step S12, generating real-time predicted compliance factors. These predictions may be output to the display device in a prediction output step S13.

[0226] In some cases, the system may calculate an event-based alpha value for each rate change event. The event-based alpha value may be computed as a ratio of the fitted pressure change (ΔPfit) to the change in injection rate (ΔQ). The system may integrate a plurality of calculated event-based alpha values to generate a cumulative alpha factor.

[0227] The system may also calculate an instantaneous shut-in pressure (ISIP) alpha value when an ISIP event is detected in the real-time data. The ISIP alpha value may be computed as a ratio of a pressure change during the ISIP event to a corresponding change in injection rate.

[0228] In some implementations, the system may integrate distributed acoustic sensing (DAS) data from a fiber optic cable deployed in the wellbore with the injection rate and pressure measurements. The system may use the trained machine learning model to identify locations of fluid entry into the formation based on patterns in the DAS data correlated with changes in the real-time predicted compliance factor.

[0229] The system may generate a dynamic fracture propagation visualization showing estimated fracture growth over time correlated with the identified fluid entry locations and the real-time predicted compliance factor.

[0230] Finally, a parameter adjustment step S14 may send additional operational parameter adjustments to the hydraulic fracturing equipment based on the machine learning predictions.

[0231] This sequence diagram illustrates the parallel processing paths between traditional compliance calculation (steps S2-S7) and machine learning prediction (steps S10-S12), both contributing to operational decisions through steps S9 and S14 respectively. The integration of these two approaches may provide a comprehensive analysis of fracture behavior and enable real-time optimization of hydraulic fracturing operations.

[0232] In some cases, a compliance calculation process 700 may be implemented for real-time analysis of hydraulic fracturing operations. FIG. 7 illustrates a flowchart of the compliance calculation process 700, which may include multiple interconnected modules that process real-time data to calculate and utilize compliance factors during hydraulic fracturing operations.

[0233] The compliance calculation process 700 may begin with a rate change detection module 702 that receives input from a data stream module 716. The data stream module 716 may provide a continuous flow of real-time data from sensors at the hydraulic fracturing site. The rate change detection module 702 may analyze this data to identify significant changes in injection rate that warrant further analysis.

[0234] Upon detecting a rate change event, the process 700 may branch into two parallel processing paths. In the first path, an injection rate calculation module 704 may determine the magnitude of changes in injection rate (ΔQ). The output from the injection rate calculation module 704 may feed into a compliance value computation module 712.

[0235] In the second path, a pressure response analysis module 706 may process the pressure data associated with the detected rate change event. Within the pressure response analysis module 706, a mathematical function fitting module 708 may fit a mathematical function, such as an exponential curve, to the pressure data. The output from the mathematical function fitting module 708 may flow to a pressure change determination module 710, which may calculate the fitted pressure change (ΔPfit) based on the mathematical function. The pressure change determination module 710 may then provide input to the compliance value computation module 712.

[0236] The compliance value computation module 712 may process inputs from both paths to calculate compliance values. In some cases, the compliance value may be computed as the ratio of the change in injection rate to the fitted pressure change (ΔQ / ΔPfit). These calculated compliance values may then flow to a cumulative compliance integration module 714, which may generate integrated compliance factors over time.

[0237] In some implementations, the compliance calculation process 700 may include a step of filtering the calculated compliance values to exclude values greater than a predetermined threshold. This filtering step may help remove outliers or anomalous data points that could skew the analysis. The predetermined threshold may be set based on historical data or expert knowledge of the expected range of compliance values for the specific formation being fractured.

[0238] Following the filtering step, the process 700 may include binning the filtered compliance values and corresponding leak-off data by hydraulic fracturing stage. This binning process may allow for stage-specific analysis of fracture behavior and fluid efficiency during the hydraulic fracturing operation.

[0239] The output from the cumulative compliance integration module 714 may connect to both a display interface module 718 and a parameter adjustment module 720. The display interface module 718 may present the compliance data and other relevant information to operators through visual representations such as graphs, charts, or numerical displays.

[0240] The parameter adjustment module 720 may use the compliance data to modify fracturing parameters in real-time. In some cases, the system may not only provide human-readable alarms but also auto-adjust fracturing parameters such as flow rates, fluid type, or offset well back pressure based on the calculated compliance signals. This automated adjustment capability may allow for rapid optimization of the fracturing process in response to changing subsurface conditions.

[0241] In some implementations, the compliance calculation process 700 may generate a summary statistics report for the hydraulic fracturing operation. This report may include information such as the total number of rate change events processed, the distribution of compliance values across different stages, and any significant trends or anomalies detected during the operation.

[0242] The process 700 may also be capable of measuring compliance signals when the injection rate is constant but the pressure changes or oscillates. This capability may provide additional insights into formation behavior and fracture propagation even during periods of steady-state injection.

[0243] In some cases, geological or drilling metadata, such as well trajectory, formation tops, or stress logs, may be loaded into the model before fracturing begins. This pre-loaded data may be used to anticipate likely negative compliance intervals and optimize stage design, potentially improving the overall efficiency of the fracturing operation.

[0244] By analyzing negative compliance ‘dips,’ the process 700 may flag partial infiltration events, differentiating them from fully beneficial large compliance expansions into fresh zones. This detailed analysis may provide operators with a more nuanced understanding of fracture behavior and help identify potential issues such as fluid diversion into undesired zones.

[0245] The compliance calculation process 700 may incorporate continuous data flow from the data stream module 716, allowing for ongoing monitoring and adjustment of the hydraulic fracturing operation. The modules may be arranged to enable parallel processing of rate and pressure data, with the results converging at the compliance value computation module 712 for integrated analysis. This architecture may allow for efficient, real-time calculation and utilization of compliance factors throughout the hydraulic fracturing operation.

[0246] In some cases, a hydraulic fracturing system 800 may be implemented for performing hydraulic fracturing operations. FIG. 8 illustrates a block diagram of the hydraulic fracturing system 800, which may include several interconnected components.

[0247] The hydraulic fracturing system 800 may include a target well 802 into which fracturing fluid may be injected. A pump 804 may be connected to the target well 802 to provide the necessary pressure for fluid injection. The pump 804 may receive fluid from a fluid reservoir 806 and proppant from a proppant supply 808. In some implementations, a mixing unit 810 may be used to combine the fluid and proppant before delivery to the pump 804.

[0248] The hydraulic fracturing system 800 may incorporate monitoring components to measure various parameters during the fracturing operation. These monitoring components may include a pressure sensor 812 and a flow rate sensor 814. The pressure sensor 812 may measure the injection pressure at various points in the system, while the flow rate sensor 814 may measure the rate of fluid injection into the target well 802.

[0249] A control unit 816 may be connected to the monitoring components and other system elements. The control unit 816 may process the sensor data and manage the operation of the system components. In some cases, the control unit 816 may implement the real time data step, rate change detection step, injection rate calculation step, mathematical function fitting step, pressure change determination step, compliance value calculation step, and compliance integration step as described in previous sections.

[0250] A display 818 may be connected to the control unit 816 for presenting operational data and system status information. The display 818 may be used to implement the compliance factor output step, allowing operators to visualize compliance factors and other relevant data in real-time.

[0251] The hydraulic fracturing system 800 may also include a machine learning module 820 connected to the control unit 816. The machine learning module 820 may receive data from the sensors and provide predictive analytics to assist in system operation. In some implementations, the machine learning module 820 may perform the model training step, current data input step, compliance prediction step, and prediction output step.

[0252] In some cases, the historical hydraulic fracturing data used by the machine learning module 820 may further comprise subterranean formation data. This subterranean formation data may include well log data, core sample measurements, seismic or microseismic data, porosity or permeability measurements, or formation pressure data. By incorporating this additional data, the machine learning module 820 may develop more comprehensive models for predicting compliance factors and optimizing fracturing operations.

[0253] The control unit 816 may analyze frequency components of the pressure measurements using a Fourier transform. This analysis may be performed on the data collected by the pressure sensor 812. By examining the frequency domain representation of the pressure data, the control unit 816 may identify dominant frequencies associated with rock breakage during the hydraulic fracturing operation.

[0254] In some implementations, the control unit 816 may correlate the identified dominant frequencies with geological characteristics of the formation being fractured. This correlation may provide insights into the fracturing process and the response of different geological layers to the applied pressure.

[0255] Based on the frequency analysis and correlation with geological characteristics, the control unit 816 may adjust operational parameters of the hydraulic fracturing system 800. For example, the control unit 816 may adjust the injection rate controlled by the pump 804 or modify the fluid composition in the mixing unit 810. These adjustments may be made to target specific frequency ranges associated with optimal fracture propagation in the formation.

[0256] The system components may be arranged to enable fluid flow from the fluid reservoir 806 and proppant supply 808 through the mixing unit 810, which then feeds into the pump 804 for injection into the target well 802. The pressure sensor 812 and flow rate sensor 814 may monitor the injection process, providing real-time data to the control unit 816 for analysis and display.

[0257] By integrating these various components and leveraging the capabilities of the machine learning module 820, the hydraulic fracturing system 800 may enable real-time monitoring, analysis, and optimization of fracturing operations based on calculated compliance factors and other relevant parameters.

[0258] In some cases, a C-Factor Calculation System 900 may be implemented for real-time analysis and optimization of hydraulic fracturing operations. FIG. 9 illustrates a system diagram of the C-Factor Calculation System 900, which may include multiple interconnected layers for data acquisition, transmission, processing, machine learning, and visualization and control.

[0259] The C-Factor Calculation System 900 may include a Data Acquisition Layer 902 comprising various sensors and systems for collecting data relevant to C-Factor calculations. The Data Acquisition Layer 902 may incorporate Pressure Sensors 904, Flow Rate Sensors 906, a Distributed Acoustic Sensing (DAS) System 908, and Offset Well Sensors 910. These components may gather different types of data used for computing compliance factors and C-Factors.

[0260] In some implementations, the Data Transmission Layer 912 may facilitate the transfer of collected data from the acquisition layer to subsequent processing stages. The Data Transmission Layer 912 may include a Fiber Optic Network 914 and a Satellite Communication System 916, which may enable high-bandwidth, low-latency data transfer and provide backup connectivity in remote locations, respectively.

[0261] The Data Processing Layer 918 may handle the processing of transmitted data. The Data Processing Layer 918 may feature a High-Performance Computing Cluster 920, which may include CPU Nodes 922 and GPU Nodes 924. This layer may also incorporate Real-Time Stream Processing 926 and a Time-Series Database 928 for managing and analyzing the incoming data streams.

[0262] In some cases, the Machine Learning Layer 930 may process the data using advanced algorithms. The Machine Learning Layer 930 may consist of a Model Training Module 932, a Real-Time Inference Engine 934, and a Feature Extraction Pipeline 936. These components may work together to analyze the processed data and generate insights for C-Factor calculations.

[0263] The Visualization and Control Layer 938 may present the analyzed data and enable system management. The Visualization and Control Layer 938 may include a Real-Time Dashboard 940 for data visualization, an Automated Control System 942, and an Alert Management System 944.

[0264] In some implementations, the Feature Extraction Pipeline 936 may incorporate signal processing techniques to analyze frequency components of the pressure measurements using a Fourier transform. This analysis may enable the identification of dominant frequencies associated with rock breakage during the hydraulic fracturing operation.

[0265] The Real-Time Inference Engine 934 may correlate the identified dominant frequencies with geological characteristics of the formation being fractured. This correlation may provide insights into the fracturing process and formation response.

[0266] Based on the insights generated by the Machine Learning Layer 930, the Automated Control System 942 may adjust at least one of injection rate or fluid composition to target specific frequency ranges associated with optimal fracture propagation in the formation. These adjustments may be implemented through the hydraulic fracturing equipment to optimize the fracturing process in real-time.

[0267] The components of the C-Factor Calculation System 900 may interact through various connections. Data from the sensors in the Data Acquisition Layer 902 may flow to the Data Transmission Layer 912, which may then send the data to the Data Processing Layer 918. The Data Processing Layer 918 may exchange data with the Machine Learning Layer 930, which in turn may provide outputs to the Visualization and Control Layer 938. The Visualization and Control Layer 938 may send signals back to other parts of the system, such as the Automated Control System 942.

[0268] The C-Factor Calculation System 900 may operate by continuously collecting data, transmitting the data for processing, applying machine learning algorithms, and presenting the results for visualization and control. The system may employ various technologies including fiber optic networks, satellite communication, high-performance computing, and real-time data processing to perform C-Factor calculations and related analyses, potentially enabling more precise control over fracture growth and optimizing resource utilization during hydraulic fracturing operations.

[0269] In some cases, a software architecture 1000 may be implemented for a C-Factor calculation system. FIG. 10 illustrates a block diagram of the software architecture 1000, which may include multiple interconnected layers that handle different aspects of system operation.

[0270] The software architecture 1000 may include an operating system layer 1002. In some implementations, the operating system layer 1002 may comprise a real time operating system 1004 and a hypervisor 1006. The real time operating system 1004 may communicate with the hypervisor 1006 to manage system resources and timing-critical operations.

[0271] A middleware layer 1008 may be included in the software architecture 1000. The middleware layer 1008 may contain a message queuing system 1010 that interfaces with a distributed file system 1012. In some cases, the message queuing system 1010 may manage data flow between components while the distributed file system 1012 may handle data storage and retrieval.

[0272] The software architecture 1000 may incorporate a data processing layer 1014. The data processing layer 1014 may include a stream processing framework 1016 and a signal processing library 1018. In some implementations, the stream processing framework 1016 may process real-time data streams while the signal processing library 1018 may perform signal analysis operations.

[0273] A machine learning framework 1020 may be included in the software architecture 1000. The machine learning framework 1020 may comprise a tensor processing module 1022, a neural network module 1024, and a machine learning library 1026. In some cases, the tensor processing module 1022 may communicate with the neural network module 1024, which may in turn interface with the machine learning library 1026.

[0274] The software architecture 1000 may include a visualization layer 1028. The visualization layer 1028 may contain a graphics library 1030 and a plotting module 1032. In some implementations, the graphics library 1030 may provide visualization capabilities while the plotting module 1032 may generate data plots and visual representations.

[0275] A control system layer 1034 may be incorporated into the software architecture 1000. The control system layer 1034 may include a protocol interface 1036 and a monitoring system 1038. In some cases, the protocol interface 1036 may manage communication protocols while the monitoring system 1038 may oversee system operation.

[0276] The layers of the software architecture 1000 may be interconnected through various data paths, allowing information to flow between components. In some implementations, the middleware layer 1008 may connect to both the data processing layer 1014 and the machine learning framework 1020. The visualization layer 1028 and control system layer 1034 may interface with other layers to provide user interaction and system control capabilities.

[0277] In some cases, the software architecture 1000 may enable the implementation of various steps and modules described in previous sections. For example, the stream processing framework 1016 in the data processing layer 1014 may facilitate the real time data step, rate change detection step, and injection rate calculation step. The signal processing library 1018 may support the mathematical function fitting step and pressure change determination step.

[0278] The machine learning framework 1020 may enable the model training step, current data input step, and compliance prediction step. The visualization layer 1028 may support the compliance factor output step and prediction output step, while the control system layer 1034 may facilitate the operational parameter adjustment step and parameter adjustment step.

[0279] By integrating these various layers and components, the software architecture 1000 may provide a comprehensive framework for implementing the C-Factor calculation system, enabling real-time analysis and optimization of hydraulic fracturing operations.

[0280] The Real-Time Data Acquisition system, in most cases, collects data at high frequency (typically 1 Hz) to provide a detailed representation of the fracturing operation, different frequencies and their uses will be understood by those skilled in the art. This high-resolution data collection can capture rapid changes in pressure and flow rates that occur during fracturing, allowing for analysis and decision-making. The system's ability to monitor offset-well pressures can be useful for identifying potential communication between wells, which can impact fracture propagation and well performance.

[0281] The Compliance Computation Engine builds upon this data stream by calculating parameters such as pressure change (ΔP(t)) and injection rate change (ΔQ(t)). The integration of these rate-change-to-pressure-change ratios into a running “C-Factor” provides a measure of fracturing effectiveness over time. This approach allows for an understanding of fracture behavior compared to methods that rely on discrete measurements. The engine's ability to handle both step-up and step-down scenarios can be useful, as it allows for analysis of fracture response under various operational conditions.

[0282] When C-Factor growth flattens, such events can indicate unproductive fracturing or fluid diversion into offset wells, which are issues in hydraulic fracturing operations. By tracking long-term compliance trends, the system can provide insights into the effectiveness of the fracturing process and help identify potential problems that may impact well performance.

[0283] The optional Machine-Learning Module represents an advancement in fracturing analysis. By employing neural networks or other AI models trained on historical data, this module can interpret complex compliance patterns that may be challenging for human operators to discern. The ability to differentiate between beneficial high compliance and unproductive compliance can be useful, as it allows operators to optimize fracture growth while minimizing unproductive fluid diversion. The detection and flagging of negative compliance events can enable responses to suboptimal conditions, potentially saving time and resources.

[0284] The Advisory or Automated Control system represents the culmination of the architecture's capabilities. By leveraging the insights generated by the previous modules, this system can provide real-time optimization of the fracturing process. The ability to recommend or automatically implement changes to pump parameters, fluid type, or offset well pressurization represents an advancement in fracturing control. These adjustments aim to maintain effective fracturing conditions, promoting efficient fracture growth while minimizing unwanted fluid diversion.

[0285] This system architecture has the potential to improve the efficiency and effectiveness of hydraulic fracturing operations. By providing real-time, data-driven insights and control, it can help operators maximize production while minimizing costs and environmental impact. The integration of machine learning and automated control systems represents a step towards more intelligent and adaptive fracturing operations. In some implementations, the system may incorporate real-time economic modeling that evaluates operational costs—including equipment hourly rates, fluid volume costs (per barrel), proppant costs (per pound), service provider rates, and stage-specific expenditures—in conjunction with predicted production outcomes derived from the cumulative compliance metric calculated during active stimulation operations. This enables real-time calculation and display of economic efficiency metrics, such as cost-per-unit-compliance, projected return on investment (ROI), or net present value (NPV) estimates, during fracturing activities rather than weeks or months post-completion. For example, the system may compute instantaneous profit-and-loss projections by combining the ML-predicted 6-month or 12-month production forecast (generated from the novel C-Factor method) with accumulated stage costs and forecast commodity prices, thereby enabling operators to make economically informed decisions—such as determining optimal stage durations, rate schedules, fluid compositions, or whether to continue or terminate pumping based on diminishing compliance returns—concurrently with fracturing operations. This real-time economic optimization capability, facilitated by the integration of event-dense compliance analysis with production forecasting and cost tracking, represents a practical application that extends beyond mere data collection or mathematical calculation, providing actionable economic intelligence for optimizing capital allocation and operational efficiency during hydraulic fracturing. This capability may enhance both the technical and economic performance of hydraulic fracturing operations and contribute to more efficient energy production.

[0286] The systems and methods described herein may be implemented using various computing hardware and software components known in the art. The computing system may include one or more processors (e.g., CPUs, GPUs, or other processing units) configured to execute machine-readable instructions stored in memory. Memory may include volatile memory (e.g., RAM) for temporary data storage and non-volatile storage (e.g., solid-state drives, hard drives) for persistent data storage.

[0287] Data acquisition from hydraulic fracturing operations may be accomplished using pressure sensors, flow rate sensors, and formation measurement systems known in the art. These sensors may communicate with the computing system via wired or wireless communication networks using standard industrial communication protocols.

[0288] As used herein, “surface pumping equipment” refers to equipment deployed at the surface for injecting fracturing fluid into a wellbore during hydraulic fracturing operations, including but not limited to pumps, blenders, surface treating iron, wellheads, coiled tubing units, flow control equipment, and associated piping and manifolds. Surface pumping equipment is distinguished from downhole equipment (e.g., perforating guns, packers) that is deployed within the wellbore.

[0289] As used herein, “surface sensor” refers to a sensor or measurement device deployed at or near the surface location of a wellbore during hydraulic fracturing operations, as distinguished from downhole sensors deployed within the wellbore. Surface sensors may include, but are not limited to, pressure sensors, flow rate sensors, or flow meters positioned at the wellhead, on surface treating equipment, in surface piping, or at other surface locations where they can measure operational parameters during fluid injection. Surface sensors may provide real-time measurements of injection pressure, injection rate, proppant concentration, or other operational parameters relevant to compliance factor determination. The term “at least one surface sensor” encompasses implementations using a single sensor capable of measuring multiple parameters, multiple discrete sensors each measuring different parameters, or any combination thereof.

[0290] Machine learning models may be implemented using any suitable machine learning framework or library (e.g., neural network frameworks, ensemble learning methods, regression techniques). The models may leverage hardware acceleration capabilities available on the computing platform to enable real-time inference during hydraulic fracturing operations.

[0291] Visualization of compliance factors and operational parameters may be presented to operators via graphical user interfaces implemented using web-based or desktop visualization frameworks. The system may provide real-time updates to displayed data as new sensor measurements are received and processed.

[0292] This hardware ecosystem can be configured to work in concert to enable the real-time analysis and optimization described in the claim. The high-precision sensors can be configured to provide accurate input data, which are transmitted to the processing center via low-latency networks. The high-performance computing cluster, augmented by specialized machine learning hardware, can be configured to process this data in real-time, running the trained models to predict compliance factors. The results can then be immediately displayed on the visualization systems, allowing operators to make informed decisions. Simultaneously, the automated control systems can be configured to implement optimizations based on the model outputs, adjusting fracturing parameters to improve efficiency and productivity.

[0293] Machine readable storage including machine-readable instructions, when executed, to implement a method or realize an apparatus in any of the examples of the present application. Various techniques, or certain aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in tangible media, such as floppy diskettes, CD-ROMs, hard drives, a non-transitory computer readable storage medium, or any other machine-readable storage medium wherein, when the program code is loaded into and executed by a machine, such as a computer, the machine becomes an apparatus for practicing the various techniques. In the case of program code execution on programmable computers, the computing device may include a processor, a storage medium readable by the processor (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. The volatile and non-volatile memory and / or storage elements may be a RAM, an EPROM, a flash drive, an optical drive, a magnetic hard drive, or another medium for storing electronic data. The eNB (or other base station) and UE (or other mobile station) may also include a transceiver component, a counter component, a processing component, and / or a clock component or timer component. One or more programs that may implement or utilize the various techniques described herein may use an application programming interface (API), reusable controls, and the like. Such programs may be implemented in a high-level procedural or an object-oriented programming language to communicate with a computer system. However, the program(s) may be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or an interpreted language, and combined with hardware implementations.

[0294] It should be understood that many of the functional units described in this specification may be implemented as one or more components, which is a term used to more particularly emphasize their implementation independence. For example, a component may be implemented as a hardware circuit comprising custom very large scale integration (VLSI) circuits or gate arrays, off-the-shelf semiconductors such as logic chips, transistors, or other discrete components. A component may also be implemented in programmable hardware devices such as field programmable gate arrays, programmable array logic, programmable logic devices, or the like.

[0295] Components may also be implemented in software for execution by various types of processors. An identified component of executable code may, for instance, comprise one or more physical or logical blocks of computer instructions, which may, for instance, be organized as an object, a procedure, or a function. Nevertheless, the executables of an identified component need not be physically located together, but may comprise disparate instructions stored in different locations that, when joined logically together, comprise the component and achieve the stated purpose for the component.

[0296] As used herein, “user interface” refers to any means by which cumulative compliance values or other operational data are made accessible to an operator or user, including but not limited to visual displays (such as monitors, screens, or dashboards), application programming interfaces (APIs) through which data may be transmitted to external systems or applications for display or further processing, data export functions, printed reports, audible alerts, or any combination thereof. The term encompasses both direct presentation of data to an operator (e.g., on a display screen at the wellsite) and indirect presentation through intermediate systems (e.g., transmitting data via API to a remote system where it is displayed to an operator). Displaying data “on a user interface” includes making the data available through any such interface, regardless of whether the system performs the final visual rendering or transmits the data to another system that performs the rendering.

[0297] Indeed, a component of executable code may be a single instruction, or many instructions, and may even be distributed over several different code segments, among different programs, and across several memory devices. Similarly, operational data may be identified and illustrated herein within components, and may be embodied in any suitable form and organized within any suitable type of data structure. The operational data may be collected as a single data set, or may be distributed over different locations including over different storage devices, and may exist, at least partially, merely as electronic signals on a system or network. The components may be passive or active, including agents operable to perform desired functions.

[0298] Reference throughout this specification to “an example” means that a particular feature, structure, or characteristic described in connection with the example is included in at least one embodiment of the present invention. Thus, appearances of the phrase “in an example” in various places throughout this specification are not necessarily all referring to the same embodiment.

[0299] As used herein, a plurality of items, structural elements, compositional elements, and / or materials may be presented in a common list for convenience. However, these lists should be construed as though each member of the list is individually identified as a separate and unique member. Thus, no individual member of such list should be construed as a de facto equivalent of any other member of the same list solely based on its presentation in a common group without indications to the contrary. In addition, various embodiments and examples of the present invention may be referred to herein along with alternatives for the various components thereof. It is understood that such embodiments, examples, and alternatives are not to be construed as de facto equivalents of one another, but are to be considered as separate and autonomous representations of the present invention.

[0300] Although the foregoing has been described in some detail for purposes of clarity, it will be apparent that certain changes and modifications may be made without departing from the principles thereof. It should be noted that there are many alternative ways of implementing both the processes and apparatuses described herein. Accordingly, the present embodiments are to be considered illustrative and not restrictive, and the invention is not to be limited to the details given herein, but may be modified within the scope and equivalents of the appended claims.

[0301] Those having skill in the art will appreciate that many changes may be made to the details of the above-described embodiments without departing from the underlying principles of the invention. The scope of the present invention should, therefore, be determined only by the following claims.

Claims

1. A method for compliance monitoring during a hydraulic fracturing operation, comprising:a. receiving, during an active hydraulic fracturing operation, pressure data and injection rate data from at least one sensor, and formation data comprising at least one of: well log data, core sample measurements, seismic data, microseismic data, porosity measurements, permeability measurements, formation pressure data, distributed acoustic sensing (DAS) data, or geological or drilling metadata including formation tops or stress logs;b. inputting, by a computing system, the received pressure data, injection rate data, and formation data into a trained machine learning model;c. generating, by the trained machine learning model, cumulative compliance values representing accumulated fracture system response based on the received pressure data, injection rate data, and formation data; andd. providing the cumulative compliance values to a user display device during the active hydraulic fracturing operation.

2. The method of claim 1, wherein the trained machine learning model was trained on historical cumulative compliance values derived from summing discrete compliance measurements from previously completed fracturing stages.

3. The method of claim 1, wherein generating cumulative compliance values comprises generating values at a temporal resolution sufficient to enable operational adjustments during a stimulation stage.

4. The method of claim 1, further comprising generating a forecast of well performance by comparing the cumulative compliance values with historical cumulative compliance factors from multiple previously completed wells or stages to identify relationships between cumulative compliance and production outcomes.

5. The method of claim 1, further comprising evaluating cumulative compliance values in conjunction with operational cost data during the hydraulic fracturing operation to support decisions that maximize production while minimizing costs.

6. The method of claim 5, wherein the operational cost data comprises costs associated with at least one of injection rate, proppant concentration, or fluid composition.

7. The method of claim 1, wherein providing comprises generating an alert when the cumulative compliance values indicate a declining cumulative compliance rate or a forecast of the final cumulative compliance value that falls outside a predetermined range.

8. The method of claim 1, further comprising transmitting control signals to hydraulic fracturing equipment to adjust at least one operational parameter based on the cumulative compliance values, wherein the at least one operational parameter comprises at least one of injection rate, proppant concentration, or fluid composition.

9. The method of claim 1, wherein the trained machine learning model comprises a neural network trained to generate the cumulative compliance values from pressure patterns, rate patterns, and formation characteristics without requiring detection of discrete rate change events during operation.

10. The method of claim 1, wherein the hydraulic fracturing operation comprises fracturing multiple stimulation stages within a wellbore, the method further comprising generating a well-level cumulative compliance value by aggregating cumulative compliance values across the multiple stimulation stages.

11. A system for compliance monitoring during a hydraulic fracturing operation, comprising:a. at least one data interface configured to receive, during an active hydraulic fracturing operation, pressure data and injection rate data from at least one sensor, and formation data comprising at least one of: well log data, core sample measurements, seismic data, microseismic data, porosity measurements, permeability measurements, formation pressure data, distributed acoustic sensing (DAS) data, or geological or drilling metadata including formation tops or stress logs;b. at least one processor; andc. a non-transitory memory storing instructions that, when executed by the at least one processor, cause the system to:i. input the received pressure data, injection rate data, and formation data into a trained machine learning model;ii. generate, via the trained machine learning model, cumulative compliance values representing accumulated fracture system response based on the received pressure data, injection rate data, and formation data; andiii. provide the cumulative compliance values to a user display device during the active hydraulic fracturing operation.

12. The system of claim 11, wherein the trained machine learning model was trained on historical cumulative compliance values derived from summing discrete compliance measurements from previously completed fracturing stages.

13. The system of claim 11, wherein the instructions cause the system to generate cumulative compliance values at a temporal resolution sufficient to enable operational adjustments during a stimulation stage.

14. The system of claim 11, wherein the instructions further cause the system to generate a forecast of well performance by comparing the cumulative compliance values with historical cumulative compliance factors from multiple previously completed wells or stages to identify relationships between cumulative compliance and production outcomes.

15. The system of claim 11, wherein the instructions further cause the system to evaluate cumulative compliance values in conjunction with operational cost data during the hydraulic fracturing operation to support decisions that maximize production while minimizing costs.

16. The system of claim 15, wherein the operational cost data comprises costs associated with at least one of injection rate, proppant concentration, or fluid composition.

17. The system of claim 11, wherein the instructions cause the system to generate an alert when the cumulative compliance values indicate a declining cumulative compliance rate or forecast a final cumulative compliance value that falls outside a predetermined range.

18. The system of claim 11, wherein the instructions further cause the system to transmit control signals to hydraulic fracturing equipment to adjust at least one operational parameter based on the cumulative compliance values, wherein the at least one operational parameter comprises at least one of injection rate, proppant concentration, or fluid composition.

19. The system of claim 11, wherein the trained machine learning model comprises a neural network trained to generate the cumulative compliance values from pressure patterns, rate patterns, and formation characteristics without requiring detection of discrete rate change events during operation.

20. The system of claim 11, wherein the hydraulic fracturing operation comprises fracturing multiple stimulation stages within a wellbore, and wherein the instructions further cause the system to generate a well-level cumulative compliance value by aggregating cumulative compliance values across the multiple stimulation stages.