Predicting Frackable Intervals

This disclosure relates generally to predicting frackable intervals.

Well stimulation refers to interventions performed on wells to increase production by improving the flow of hydrocarbons from the reservoir into the wellbore. Fracking is a well stimulation technique where formations are fractured by a pressurized liquid. During fracking, a fluid is injected under high-pressure into a wellbore to create cracks in the subsurface through which hydrocarbons, water, and brine flow more freely.

Well stimulation is performed to increase the flow of oil and gas from drilled wells. In examples, well stimulation efforts, such as fracking, are applied in unconventional wells, tight wells, and conventional wells to increase hydrocarbon recovery. For example, fracturing is used in unconventional wells along with horizontal drilling for increased production. Fracturing is used in conventional wells, for example, to stimulate production.

In examples, a tight well is a well with low permeability. Tight wells are stimulated to enhance productivity near the wellbore region to obtain commercial and sustainable production from the formation. Stimulations at tight wells are achieved using a stimulation treatment by which a reactive or non-reactive fluid is injected into a reservoir at a pressure above a minimum in-situ stress to create a highly permeable path (e.g., fracture) in the formation to bypass near wellbore damage induced by drilling operations. The fracture extends deep into the formation to enable the flow of hydrocarbons. Creating fractures in formations is known as hydraulic fracturing or fracking, in which the rock formation is fractured by a hydraulically pressurized fluid made of water, sand, proppants, chemicals, or any combinations thereof.

Embodiments described herein predict frackable intervals. Frackable refers to being suitable for hydraulic fracturing, or fracking. Rock fabric data from well logs is integrated with well drilling data and corresponding historical performance data to create labeled rock fabric data. A machine learning model is trained to predict a probability of a successful fracture for at least one interval using the labeled rock fabric data. In some embodiments, an integrated multi-disciplinary engineering design process is implemented to achieve an effective fracture geometry. For example, a stimulation design generated as described herein incorporates the well history, location, rock, and fluid properties. Traditionally, a human selects intervals where fracture will be initiated and subsequently performs the stimulation job using manual processes. Even though manual processes are in place to select intervals to initiate fracture, historical performance shows on average 40% of the perforated intervals using traditional manual processes are unable to frack the rock causing the cancellation of stimulation stages. Consequently, stimulation operations are not efficient and incur unnecessary costs in unknowingly attending to unfrackable intervals determined using manual processes.

Some advantages of the present techniques include an improvement to the identification of fracking intervals by leveraging rock fabric properties and historical data. Traditional techniques are limited to mapping frackability using seismic data, predicting geomechanical properties, predicting rock properties measured in the lab, or predicting fracture geometry based on real-time data. By contrast, the present techniques decouple geomechanics from the prediction and apply rock fabric data to discover frackable intervals along with leveraging existing well data. In this manner, the present techniques operate without geomechanical input for prediction, such as geomechanical input derived from petrophysical data. The present techniques use rock fabric data to avoid uncertainty on geomechanics parameter predictions. Traditional techniques suffer from the inability to robustly predict frackable intervals in tight wells or unconventional wells where parameters of the well data fails to indicate that an interval is not frackable. Traditional techniques use detailed core analysis and calibration test to generate reliable predictions.

The present techniques use rock fabric data from well logs that are integrated with the historical frack jobs. The rock fabric data is then processed using advanced data analytics and machine learning to predict the frackable intervals. The present techniques save time and resources by avoiding attempts to stimulate unfrackable intervals. Accordingly, in some embodiments the present techniques robustly detect unfrackable intervals.

shows a workflowfor predicting frackable intervals. As shown in the example of, historical fracking jobs, rock fabric data, and well drilling data. In examples, historical fracking jobs includes 102 historical hydraulic fracturing performance. In examples, rock fabric data is obtained from well logs, including but not limited to, formation bulk density, sonic, gamma ray, resistivity, wellbore caliper, neutron porosity, total porosity, effective porosity, and volume of quartz. One or any combination of well logs is integrated with the historical frack jobs. In examples, the well drilling datais captured from a predefined time period, such as the past 10 years. Well drilling data describes how the well is completed, such as a vertical well or horizontal well. Well drilling data also describes the mud weight used across reservoir interval, completion tubing strength, and other well completion data including but not limited to well head data, borehole diameters, casing size, casing lengths, and casing materials. In examples, the well drilling dataused for predictions are the well drilling data that statistically influences the resulting prediction. In examples, the historical fracking jobs, rock fabric data, and well drilling dataare compiled into a single database. In examples, the historical fracking jobs, rock fabric data, and well drilling dataare standardized and stored in the database. Data standardization converts the historical fracking jobs, rock fabric data, and well drilling datainto a standard format for input into a machine learning model. In examples, the historical fracking jobs, rock fabric data, and well drilling dataare obtained from various organizations and service providers. Data standardization ensures that data from various organizations and service providers are in a predetermined format used for input to a machine learning model. For example, data standardization includes converting measurements into the metric system or converting dates into a single format.

Data preparationidentifies patterns or clusters of wellbore intervals where frackability issues were encountered using advanced data analytics applied to the historical fracking jobs, rock fabric data, and well drilling data. In some embodiments, input data is quality checked by subject matter experts. In examples, quality checking the data includes, but is not limited to: 1) ensuring that perforation intervals are reported at a correct depth; 2) reconciling formation tops using palynology, coring, and well based and structural field correlations; 3) applying environmental correction and noise reduction to petrophysical data; 4) validating data from stimulation and perforation using post job reports and daily operations; and any combinations thereof.

The prepared data is merged into a single data-frame that is a function of well number, perforation depth, and operational time. The data-frame has more than 119 input parameters for each perforation interval from different disciplines (e.g., drilling, services organizations, stimulation data, subject matter experts, and reservoir management) as captured in the historical fracking jobs, rock fabric data, and well drilling data.

During data preparation, data analytics is performed, where patterns and/clusters of intervals with frackability issues are identified. Geomechanics is decoupled from the workflow, and instead the relationship between frackability and the rock fabric properties such as porosity, permeability, resistivity, mineralogy, etc. is used to predict frackable intervals. The data analytics includes assessing the input parameters against the discretized historical frackability results, where a value of 0 indicates not frackable and a value of 1 indicates frackable. Exploratory data analysis describes data statistically to identify patterns that are used in predicting frackable intervals. The statistically insignificant variables are removed from consideration, reducing the relevant variables from 119 to 13. In some embodiments, statistical significance is determined according to a predetermined threshold, such as 5%. The significance level describes a likelihood of the variable influencing the predicted frackable intervals. In some embodiments, trained machine learning models, such as decision trees, that output the relevant contribution of each variable on the prediction of frackable intervals. An optimization or parameter reduction is performed to reach maximum accuracy of the predicted frackable intervals with minimum input properties to build parsimonious and robust models that predict frackable intervals.

shows a cross plot of variables 1-4.shows a cross plot of variables 5-8.show data clustering where frackability issues were encountered. Identified clusters with frackability issues are enclosed in dashed polygons on each respective cross plot.

After the most relevant variables were selected, the data-frame was divided into two data sets randomly keeping same proportion of frackable and non-frackable intervals, one set for training and another for testing. Data modeling using machine learningtrains a machine learning model to predict the frackability of at least one interval. Multiple machine learning algorithms were evaluated to assess the predictability and accuracy of the respective model, among them binary classification, linear logistic regression, support vector model, anomaly detection, neural network and random forest.

During the modeling stage, the importance (e.g., statistical significance) of each variable in prediction model is estimated.shows variable importances. In the example of, each bar corresponds to a final input variable used in the model after the variable reduction workflow. The y-axis ofrefers to the importance contribution to the total prediction in fraction. The sum of the respective importances for each variable is equal to 1. Once the model is selected and tuned, the model is tested against the testing data set. The model outputs discretized intervals with high probability of being successfully frac. In examples, variables are selected for training the machine learning model with a cumulative importance that satisfy a predetermined threshold.shows the cumulative importances of variables selected for training or testing the machine learning model.

shows the identification of fracking location at various intervals of a well. In particular, fracking flags,, andare shown at various intervals. Fracking flagsandindicate that the corresponding interval is suitable for fracking. Fracking flagindicates that the corresponding interval is not suitable for fracking. In examples, the present techniques yield a +95% accuracy in predicting frackable and unfrackable intervals. In examples, the validated model is executed for blind testing with additional intervals that were not included in the initial analysis. The model predicted frackable and un-frackable with the same accuracy as observed during the training and testing stages (+95%).

shows model results during a training test at reference number, during a blind test at reference number, and during deployment at reference number. In the example of, the cluster number is defined as the number of perforations where stimulation was attempted. Additionally, the number of wells is defined as total well count included for a respective scenario (e.g., training test, blind test, or deployment). “Frac” refers to the interval count without frackability issues. “No Frac” refers to the interval count with frackability issues. In some embodiments, the trained model to predict frackable intervals has a greater than 95% successful prediction rate for frackable and unfrackable clusters. This is shown in the deployment at reference numberwhere there are no unfrackable clusters the trained machine learning model was deployed.

In an example the data driven solution workflowofwas deployed in the field on data obtained from newly drilled wells with accurate output. Table 1 is an example of predictions made for wells of a formation. As shown by Table 1, the perforated intervals selected using the workflow were successfully fracked and therefore provided significant cost saving and efficient operations. The workflowofwas also implemented in different gas fields with good results as shown in Table 1.

is a process flow diagram of a processthat enables predicting frackable intervals.

At blockrock fabric data is extracted from well logs. In some embodiments, well log data comprising parameters is assessed for each perforation interval from different disciplines against the discretized historical frackability results to minimize parameters integrated with historical performance data (or to select the most relevant parameters).

At block, rock fabric data from well logs is integrated with well drilling data and corresponding historical performance data (e.g., data preparation) to create labeled rock fabric data. In some embodiments, patterns or clusters of intervals where frackability issues were encountered are identified. Labeled rock fabric data is created by assessing well log variables/parameters for each perforation interval from different disciplines and labeling respective variables/parameters as frackable or unfrackable. In some embodiments, the machine learning model is trained to predict a probability of being an unsuccessful frac for discretized intervals.

At block, a machine learning model is trained (e.g., data modeling) to predict a probability of being a successful fracture for at least one interval using the labeled rock fabric data. In some embodiments, a stimulation operation is automatically executed in an interval when a probability of being a successful frac satisfies a predetermined threshold. In some embodiments, the stimulation operation automatically bypasses intervals when a probability of being an unsuccessful frac satisfies a predetermined threshold. In some embodiments, new, unseen well log data is input to the trained machine learning model to predict a probability of being a successful frac for discretized intervals.

illustrates hydrocarbon production operationsthat include both one or more field operationsand one or more computational operations, which exchange information and control exploration for the production of hydrocarbons. In some implementations, outputs of techniques of the present disclosure can be performed before, during, or in combination with the hydrocarbon production operations, specifically, for example, either as field operationsor computational operations, or both.

Examples of field operationsinclude forming/drilling a wellbore, hydraulic fracturing, producing through the wellbore, injecting fluids (such as water) through the wellbore, to name a few. In some implementations, methods of the present disclosure can trigger or control the field operations. For example, the methods of the present disclosure can generate data from hardware/software including sensors and physical data gathering equipment (e.g., seismic sensors, well logging tools, flow meters, and temperature and pressure sensors). The methods of the present disclosure can include transmitting the data from the hardware/software to the field operationsand responsively triggering the field operationsincluding, for example, generating plans and signals that provide feedback to and control physical components of the field operations. Alternatively or in addition, the field operationscan trigger the methods of the present disclosure. For example, implementing physical components (including, for example, hardware, such as sensors) deployed in the field operationscan generate plans and signals that can be provided as input or feedback (or both) to the methods of the present disclosure.

Examples of computational operationsinclude one or more computer systemsthat include one or more processors and computer-readable media (e.g., non-transitory computer-readable media) operatively coupled to the one or more processors to execute computer operations to perform the methods of the present disclosure. The computational operationscan be implemented using one or more databases, which store data received from the field operationsand/or generated internally within the computational operations(e.g., by implementing the methods of the present disclosure) or both. For example, the one or more computer systemsprocess inputs from the field operationsto assess conditions in the physical world, the outputs of which are stored in the databases. For example, seismic sensors of the field operationscan be used to perform a seismic survey to map subterranean features, such as facies and faults. In performing a seismic survey, seismic sources (e.g., seismic vibrators or explosions) generate seismic waves that propagate in the earth and seismic receivers (e.g., geophones) measure reflections generated as the seismic waves interact with boundaries between layers of a subsurface formation. The source and received signals are provided to the computational operationswhere they are stored in the databasesand analyzed by the one or more computer systems.

In some implementations, one or more outputsgenerated by the one or more computer systemscan be provided as feedback/input to the field operations(either as direct input or stored in the databases). The field operationscan use the feedback/input to control physical components used to perform the field operationsin the real world.

For example, the computational operationscan process the seismic data to generate three-dimensional (3D) maps of the subsurface formation. The computational operationscan use these 3D maps to provide plans for locating and drilling exploratory wells. In some operations, the exploratory wells are drilled using logging-while-drilling (LWD) techniques which incorporate logging tools into the drill string. LWD techniques can enable the computational operationsto process new information about the formation and control the drilling to adjust to the observed conditions in real-time.

The one or more computer systemscan update the 3D maps of the subsurface formation as information from one exploration well is received and the computational operationscan adjust the location of the next exploration well based on the updated 3D maps. Similarly, the data received from production operations can be used by the computational operationsto control components of the production operations. For example, production well and pipeline data can be analyzed to predict slugging in pipelines leading to a refinery and the computational operationscan control machine operated valves upstream of the refinery to reduce the likelihood of plant disruptions that run the risk of taking the plant offline.

In some implementations of the computational operations, customized user interfaces can present intermediate or final results of the above-described processes to a user. Information can be presented in one or more textual, tabular, or graphical formats, such as through a dashboard. The information can be presented at one or more on-site locations (such as at an oil well or other facility), on the Internet (such as on a webpage), on a mobile application (or app), or at a central processing facility.

The presented information can include feedback, such as changes in parameters or processing inputs, that the user can select to improve a production environment, such as in the exploration, production, and/or testing of petrochemical processes or facilities. For example, the feedback can include parameters that, when selected by the user, can cause a change to, or an improvement in, drilling parameters (including drill bit speed and direction) or overall production of a gas or oil well. The feedback, when implemented by the user, can improve the speed and accuracy of calculations, streamline processes, improve models, and solve problems related to efficiency, performance, safety, reliability, costs, downtime, and the need for human interaction.

In some implementations, the feedback can be implemented in real-time, such as to provide an immediate or near-immediate change in operations or in a model. The term real-time (or similar terms as understood by one of ordinary skill in the art) means that an action and a response are temporally proximate such that an individual perceives the action and the response occurring substantially simultaneously. For example, the time difference for a response to display (or for an initiation of a display) of data following the individual's action to access the data can be less than 1 millisecond (ms), less than 1 second(s), or less than 5 s. While the requested data need not be displayed (or initiated for display) instantaneously, it is displayed (or initiated for display) without any intentional delay, taking into account processing limitations of a described computing system and time required to, for example, gather, accurately measure, analyze, process, store, or transmit the data.

Events can include readings or measurements captured by downhole equipment such as sensors, pumps, bottom hole assemblies, or other equipment. The readings or measurements can be analyzed at the surface, such as by using applications that can include modeling applications and machine learning. The analysis can be used to generate changes to settings of downhole equipment, such as drilling equipment. In some implementations, values of parameters or other variables that are determined can be used automatically (such as through using rules) to implement changes in oil or gas well exploration, production/drilling, or testing. For example, outputs of the present disclosure can be used as inputs to other equipment and/or systems at a facility. This can be especially useful for systems or various pieces of equipment that are located several meters or several miles apart, or are located in different countries or other jurisdictions.

is a schematic illustration of an example controller(or control system) for that enables predicting frackable intervals. For example, the controllermay be operable according to the workflowofor the processof. In some embodiments, the controlleris the same as or similar to the computer systemsof. The controlleris intended to include various forms of digital computers, such as printed circuit boards (PCB), processors, digital circuitry, or otherwise parts of a system for supply chain alert management. Additionally the system can include portable storage media, such as, Universal Serial Bus (USB) flash drives. For example, the USB flash drives may store operating systems and other applications. The USB flash drives can include input/output components, such as a wireless transmitter or USB connector that may be inserted into a USB port of another computing device.

The controllerincludes a processor, a memory, a storage device, and an input/output interfacecommunicatively coupled with input/output devices(for example, displays, keyboards, measurement devices, sensors, valves, pumps). Each of the components,,, andare interconnected using a system bus. The processoris capable of processing instructions for execution within the controller. The processor may be designed using any of a number of architectures. For example, the processormay be a CISC (Complex Instruction Set Computers) processor, a RISC (Reduced Instruction Set Computer) processor, or a MISC (Minimal Instruction Set Computer) processor.

In one implementation, the processoris a single-threaded processor. In another implementation, the processoris a multi-threaded processor. The processoris capable of processing instructions stored in the memoryor on the storage deviceto display graphical information for a user interface on the input/output interface.

The memorystores information within the controller. In one implementation, the memoryis a computer-readable medium. In one implementation, the memoryis a volatile memory unit. In another implementation, the memoryis a nonvolatile memory unit.

The storage deviceis capable of providing mass storage for the controller. In one implementation, the storage deviceis a computer-readable medium. In various different implementations, the storage devicemay be a floppy disk device, a hard disk device, an optical disk device, or a tape device.

The input/output interfaceprovides input/output operations for the controller. In one implementation, the input/output devicesincludes a keyboard and/or pointing device. In another implementation, the input/output devicesincludes a display unit for displaying graphical user interfaces.

There can be any number of controllersassociated with, or external to, a computer system containing controller, with each controllercommunicating over a network. Further, the terms “client,” “user,” and other appropriate terminology can be used interchangeably, as appropriate, without departing from the scope of the present disclosure. Moreover, the present disclosure contemplates that many users can use one controllerand one user can use multiple controllers.

According to some non-limiting embodiments or examples, provided is a computer-implemented method of training a neural network for predicting frackable intervals including: extracting, using at least one hardware processor, rock fabric data from well logs; integrating, using the at least one hardware processor, rock fabric data from well logs with well drilling data and corresponding historical performance data to create labeled rock fabric data; and training, using the at least one hardware processor, a machine learning model to predict a probability of being a successful fracture for at least one interval using the labeled rock fabric data.

According to some non-limiting embodiments or examples, provided is an apparatus including a non-transitory, computer readable, storage medium that stores instructions that, when executed by at least one processor, cause the at least one processor to perform operations including: extracting rock fabric data from well logs; integrating rock fabric data from well logs with well drilling data and corresponding historical performance data to create labeled rock fabric data; and training a machine learning model to predict a probability of being a successful fracture for at least one interval using the labeled rock fabric data.

According to some non-limiting embodiments or examples, provided is a system, including: one or more memory modules; one or more hardware processors communicably coupled to the one or more memory modules, the one or more hardware processors configured to execute instructions stored on the one or more memory modules to perform operations including: extracting rock fabric data from well logs; integrating rock fabric data from well logs with well drilling data and corresponding historical performance data to create labeled rock fabric data; and training a machine learning model to predict a probability of being a successful fracture for at least one interval using the labeled rock fabric data.

Further non-limiting aspects or embodiments are set forth in the following numbered embodiments:

Embodiment 1: A computer-implemented method of training a neural network for predicting frackable intervals including: extracting, using at least one hardware processor, rock fabric data from well logs; integrating, using the at least one hardware processor, rock fabric data from well logs with well drilling data and corresponding historical performance data to create labeled rock fabric data; and training, using the at least one hardware processor, a machine learning model to predict a probability of being a successful fracture for at least one interval using the labeled rock fabric data.

Embodiment 2: The computer implemented method of any preceding embodiment, including identifying patterns or clusters of intervals where frackability issues were encountered, and training the machine learning model using data corresponding to the clusters.

Embodiment 3: The computer implemented method of any preceding embodiment, wherein labeled rock fabric data is created by assessing well log parameters for each perforation interval from different disciplines and labeling respective parameters as frackable or unfrackable.

Embodiment 4: The computer implemented method of any preceding embodiment, including automatically bypassing intervals when a probability of being an unsuccessful frac satisfies a predetermined threshold.

Embodiment 5: The computer implemented method of any preceding embodiment, wherein the machine learning model is trained to predict a probability of being an unsuccessful frac for discretized intervals.

Embodiment 6: The computer implemented method of any preceding embodiment, including assessing well log data including parameters for each perforation interval from different disciplines against discretized historical frackability results to minimize parameters integrated with historical performance data or to select the most relevant parameters.

Embodiment 7: The computer implemented method of any preceding embodiment, including inputting new, unseen well log data to the trained machine learning model to predict a probability of being a successful frac for discretized intervals.