Altering Natural Gas Composition In-situ During Production

This specification generally relates to producing natural gas from subsurface formations, particularly using both injection and production wells.

Enhanced gas recovery can enable production of natural gas from formations where it would otherwise not be economical. Such formations can include shale and tight gas systems. Enhanced gas recovery techniques include hydraulic, pneumatic, and thermal fracturing, carbon dioxide (CO) injection, mechanical cutting of shale formations, and enhanced bacterial methanogenesis. In some situations, injecting COinto a formation can enhance gas recovery while also providing geologic storage of the CO.

This specification describes an approach to producing natural gas from subsurface formations while altering gas composition in-situ during production. This approach uses a system that predicts the required gas injection volume to alter the subsurface gas composition. By altering gas composition in-situ during production, this approach helps produce a defined gas composition from the wells and the field, even when dealing with, for example, natural gas containing high hydrogen sulfide (HS) concentrations.

In some implementation, this system uses real-time pressure, volume, temperature (PVT) from downhole samples and well logs with a machine learning model to predict injection volumes that will produce a high quality gas mix on surface. An example implementation of this approach can start by adding a production well and an associated injection well to a field. The production and injection wells can be equipped with sensors (e.g., a gas chromatograph, a downhole pressure gauge, a downhole temperature gauge, and a fluid density meter) to monitor the current downhole conditions downhole for the training of the machine learning model. The gas chromatograph analyzes the gas sample to determine its composition, while the pressure gauge and temperature gauge provide the pressure and temperature data to be used for PVT analysis. The fluid density meter is used to determine the density of the injected fluid.

The approach disclosed in this specification can provide one or more of the following advantages.

Gas brown fields (i.e., previously produced gas fields) can be challenging to produce due to high production of impurity byproducts (e.g., HS and CO). Some fields remained undeveloped due to their high capital and operating costs. The major cost drivers for such fields are the high capital of anti-corrosive facilities, operating cost to maintain the facilities safely integrated, and/or costs related to safeguarding the environment and reducing COemissions. This approach can provide an assessment of the impacts of the injection parameters, sweep patterns and compositional changes in subsurface to improve, for example, carbon sequestration and enhanced gas recovery (CSEGR) that is not limited by the experimental control and accuracy. In particular, this approach can assess the compositional changes in real time while injecting/producing using a machine learning model that is fed with the monitoring data and used to make predictions to improve the injection/production parameters and reach the desired gas composition produced.

This approach has the potential to reduce the capital investment required to develop brown fields and other reservoirs with high HS concentrations. It can monitor subsurface compositional changes and injection gas sweep in real time, reduce production of HS or other undesired components, and enhance injection/production cycles. For example, HS is toxic and corrosive in nature posing significant risk to human health, safety, and environment (e.g. assets damaged by corrosion/leakage or personnel injured or death). By injecting gas (e.g., methane or CO) into a formation to adjust the composition of gas produced from the formation, this approach can reduce or eliminate the need for measures to deal with HS at the surface (e.g., HS detection, HS monitoring, HS removal and treatment, and anti-corrosion metals/pipelines). These measures increase as HS concentrations climb above 10% and are particularly expensive as HS concentrations climb above 25% as the cost of assets, pipes, infrastructure increases drastically. Monitoring PVT parameters in real time from both production and injection wells enables a machine learning model to make predictions based on real time changes in PVT in contrast to simulation model-based approaches that depend on historical data updates. This approach is also much more computationally efficient than simulation model-based approaches.

The details of one or more embodiments of these systems and methods are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of these systems and methods will be apparent from the description and drawings, and from the claims.

Like reference symbols in the various drawings indicate like elements.

This specification describes an approach to producing natural gas from subsurface formations while altering gas composition in-situ during production and/or enhancing COsequestration. In some implementations, this approach uses a system that predicts the required gas injection volume to alter the subsurface gas composition. By altering gas composition in-situ during production, this approach helps produce a defined gas composition from the wells and the field, even when dealing with, for example, natural gas containing high hydrogen sulfide (HS) concentrations. In some implementations, the system uses real-time pressure, volume, temperature (PVT) from downhole samples and well logs with a machine learning model to predict injection volumes that will enhance a COsequestration process. For example, this approach can start by adding an injection well for sequestration and an associated observation well to a field. The observation and injection wells can be equipped with sensors (e.g., a gas chromatograph, a downhole pressure gauge, a downhole temperature gauge, and a fluid density meter) to monitor the current downhole conditions downhole for the training of the machine learning model. The gas chromatograph analyzes the gas sample to determine its composition, while the pressure gauge and temperature gauge provide the pressure and temperature data to be used for PVT analysis. The fluid density meter is used to determine the density of the injected fluid.

In some implementation, this system uses real-time pressure, volume, temperature (PVT) from downhole samples and well logs with a machine learning model to predict injection volumes that will produce high quality gas mix on surface. An example implementation of this approach can start by adding a production well and an associated injection well to a field. The production and injection wells can be equipped with sensors (e.g., a gas chromatograph, a downhole pressure gauge, a downhole temperature gauge, and a fluid density meter) to monitor the current downhole conditions downhole for the training of the machine learning model. The gas chromatograph analyzes the gas sample to determine its composition, while the pressure gauge and temperature gauge provide the pressure and temperature data to be used for PVT analysis. The fluid density meter is used to determine the density of the injected fluid.

is a schematic view illustrating hydrocarbon exploration and production activities in a subsurface formation. For example, exploration activities including a seismic survey and well logging are illustrated. Illustrated production activities include the production of natural gas using an injection welland associated production well. The injection welland the production wellhave downhole sensorsoperable to monitor gas conditions and compositions in the subsurface formation. A pumpcan be to inject fluids (e.g., CO) into the subsurface formation to enhance the flow of gas to the production well. Controlling the timing and volume of these injections can also alter gas composition in-situ during production to help produce a defined gas composition.

The subsurface formationincludes a layer of impermeable cap rocksat the surface. Facies underlying the impermeable cap rocksinclude layers,, and. A fault lineextends across the layerand the layer.

Oil and gas tend to rise through permeable reservoir rock until further upward migration is blocked, for example, by the layer of impermeable cap rock. Exploration activities attempt to identify locations where interaction between layers of the subsurface formationare likely to trap oil and gas by limiting this upward migration. For example,shows an anticline trap, where the layer of impermeable cap rockhas an upward convex configuration, and a fault trap, where the fault linemight allow oil and gas to flow in with clay material between the walls traps the petroleum. Other traps include salt domes and stratigraphic traps.

A seismic source(for example, a seismic vibrator or an explosion) generates seismic waves that propagate in the earth. Although illustrated as a single component in, the source or sourcesare typically a line or an array of sources. The generated seismic waves include seismic body wavesthat travel into the ground and seismic surface wavestravel along the ground surface and diminish as they get further from the surface.

The seismic body wavesreflected boundaries between layers are received by a sensor or sensors. Although illustrated as a single component in, the sensor or sensorsare typically a line or an array of sensorsthat generate an output signal in response to received seismic waves including waves reflected by the horizons in the subsurface formation. The sensorscan be geophone-receivers that produce electrical output signals transmitted as input data, for example, to a computeron a seismic control truck. Based on the input data, the computermay generate a seismic data output such as, for example, a seismic two-way response time plot.

The seismic surface wavestravel more slowly than seismic body waves. Analysis of the time it takes seismic surface wavesto travel from source to sensor can provide information about near surface features.

A control centercan be operatively coupled to the seismic control truckand other data acquisition and wellsite systems. The control centermay have computer facilities for receiving, storing, processing, and analyzing data from the seismic control truckand other data acquisition and wellsite systems that provide additional information about the subsurface formation. For example, the control centercan receive data from a computerassociated with a well logging unit.

Computer systemscan be located in or at a different location than the control center. Some computer systems are provided with functionality for manipulating and analyzing the data, such as performing seismic interpretation or borehole resistivity image log interpretation to identify geological surfaces in the subsurface formation or performing simulation, planning, and optimization of production operations of the wellsite systems.

In some embodiments, a wellborethat has been drilled in the subsurface formationis logged in a well logging operation. The wellboreextends downhole from a wellhead. The wellboreis a vertical wellbore but well logging can also be performed in other wellbores, for example, slanted or horizontal wellbores. In the well logging operation, the wellborepenetrates through three layers,, andof a subsurface formation. A control trucklowers a logging tooldown the wellboreon a wireline.

The computer systemsin the control centercan be configured to analyze, model, control, optimize, or perform management tasks of field operations associated with development and production of resources such as oil and gas from the subsurface formation. For example, an injection welland a production wellextend into layerof the subsurface formation. Based on data gathered by the exploratory field operations, the computer systemscan generate models such as a reservoir model for portions of the subsurface formation. These models can simulate the effects of production field operations (e.g., injecting water or carbon dioxide through the injection wellto increase the production of hydrocarbons through the production well). The simulations can be used to plan and, in some instances, control field operations (e.g., the operation of pumpsassociated with the injection welland the production well).

In some embodiments, results generated by the computer systemsmay be displayed for user viewing using local or remote monitors or other display units. One approach to analyzing seismic data is to associate the data with portions of a seismic cube representing the subsurface formation. The seismic cube can also display results of the analysis of the seismic data associated with the seismic survey.

is a schematic illustrating an example systemfor producing natural gas from subsurface formations while altering gas composition in-situ during production. The systemincludes a control systemwith a modeland a pump control module. The control systemis in communication a database, one or more injection wells, and one or more production wells. The control systemand the databasecan be implemented on computer systems in a control center as described with reference to.

The databasecan be used to store reservoir and operational data generated, for example, by the hydrocarbon exploration and production activities described with reference to. The databaseis also in communication with the one or more injection wellsand the one or more production wells. The databasereceives data about conditions in the subsurface formation from the sensors associated with the wells (e.g., a gas chromatograph, a downhole pressure gauge, a downhole temperature gauge, and a fluid density meter). The databasealso receives operational data from the pumps associated with the wells.

The modelis in communication with the pump control module. The modelpredicts the required gas injection volume to alter the subsurface gas composition based on real-time pressure, volume, temperature from downhole samples and well logs to predict injection volumes that will provide keep produced gas high quality. In this implementation, the modelis a machine learning model. Based on the injection volumes predicted by the model, the pump control module sends signals to the wells,controlling the operation of associated pumps.

illustrates an implementation of the system. In this implementation, the systemis configured for producing gas from a subsurface formation through a production well. The systemis illustrated with an injection well, a production well, and an observation welldrilled into a gas-bearing reservoir. Non-gas bearing layersare present above and below the gas-bearing reservoir. The sensorsand pumpsassociated with the wells are in communication with the control centerand the associated control systems.

As described with respect to, the systempredicts the required gas injection volume to alter the subsurface gas composition in the gas-bearing reservoirto help produce a defined gas composition from the production wellto ameliorate the issues associated high HS concentrations found in the formation before adjustment. In some implementations, the systemuses real-time pressure, volume, temperature (PVT) from downhole samples and well logs with a machine learning model to predict injection volumes that will adjust the gas composition being produced from the gas-bearing reservoir. For example, this approach can start by adding a injection well and an associated observation well to a field that already includes production wells. The observation and injection wells can be equipped with sensors (e.g., a gas chromatograph, a downhole pressure gauge, a downhole temperature gauge, and a fluid density meter) to monitor the current downhole conditions downhole for the training of the machine learning model. The gas chromatograph analyzes the gas sample to determine its composition, while the pressure gauge and temperature gauge provide the pressure and temperature data to be used for PVT analysis. The fluid density meter is used to determine the density of the injected fluid.

Although illustrated with one injection well, one production well,, and one observation well, such systems will typically have multiple wells of each type. Some implementations do not include observation wellsand only rely on sensors in the production wellsand/or injection wells.

Implementations of systemcan also be configured for enhancing COsequestration. These implementations do not require production wellsif gas is not being extracted from the reservoir. In such implementations, the system uses real-time pressure, volume, temperature (PVT) from downhole samples and well logs with a machine learning model to predict injection volumes that will enhance a COsequestration process. For example, this approach can start by adding a injection well for sequestration and an associated observation well to a field. The observation and injection wells can be equipped with sensors (e.g., a gas chromatograph, a downhole pressure gauge, a downhole temperature gauge, and a fluid density meter) to monitor the current downhole conditions downhole for the training of the machine learning model. The gas chromatograph analyzes the gas sample to determine its composition, while the pressure gauge and temperature gauge provide the pressure and temperature data to be used for PVT analysis. The fluid density meter is used to determine the density of the injected fluid.

is a flow chart illustrating a methodof producing natural gas from subsurface formations while altering gas composition in-situ during production. Initially a target reservoir is identified. For example, the reservoir can be a reservoir in which gas composition alteration is needed due to factors such as high concentrations of HS or other undesired components (e.g., COor, in some cases, water vapor). After the reservoir is identified, monitoring equipment is installed in at least two wells, including a production well and an injection well. The equipment typically include sensors to monitor PVT properties such as pressure, temperature, and composition. The data from these sensors will be used to train the machine learning model. Data from the monitoring equipment, including real-time PVT measurements and well logs, is collected and processed.

The pressure, temperature, and composition of fluids in the subsurface formation are monitored using sensors installed downhole in an injection well or wells (step). In some implementations, the pressure, the temperature, and the composition of fluids in the subsurface formation are continuously monitored. Similarly, the pressure, temperature, and composition of fluids in the subsurface formation are monitored using sensors installed downhole in an production well or wells (step) and, in some implementations, are monitored continuously.

A volume of gas injection required to alter the composition of fluids in the subsurface formation to provide a specific fluid composition in the production well is predicted without use of a simulation model (step). In some implementations, a machine learning model is trained using the pressure, the temperature, and the composition of fluids in the subsurface formation using sensors installed downhole in an injection well and installed downhole in the production well to identify the impact of changes in injection strategies. Predicting the volume of gas injection using the machine learning model is more computationally efficient than using a simulation model.

The predicted volume of gas is injected through the injection well(s) using pump(s) associated with the injection well(s) (step). For example, injecting the predicted volume of gas can include continuously monitor the composition of the fluids downhole and adjusting injection volume and frequency based on the machine learning model. Gases injected can include CO, ethane, and nitrogen (N). If the alteration is not as effective as expected or the PVT is not changing after some injection, additional monitoring can be carried out to determine the cause. This can help to identify any issues with the wellbore or the reservoir and determine if additional actions need to be taken. For example, issues that might arise include: wellbore/surface leakage, formation plugging with a decline in injectivity, and inadequate data. The machine learning model can be trained to classify these issues for further review and/or suggest remediation (e.g., additional pumps), or suggest a retraining of the model. If the PVT is not changing as expected, adjustments can be made to the injection parameters such as volume and frequency of injection. This can help to optimize the injection process and achieve the desired PVT change. In conjunction with the injection, fluids in the subsurface formation are produced to the surface (step).

It is anticipated that the monitoring and modeling for a reservoir will initially be based on a pair of wells. Typically, additional injection and production wells will be incorporated over time. More monitoring equipment to other wells can be added incrementally in the field to monitor the change in composition across the field. This will help to adjust the machine learning model on the amount to inject per well or area. As more wells are drilled, monitoring equipment will be added in these wells to collect data and adjust the machine learning model accordingly.

is a diagram illustrating an example computer systemconfigured to execute a machine learning model. Generally, the computer systemis configured to process data indicating downhole conditions in a reservoir. The systemincludes computer processors. The computer processorsinclude computer-readable memoryand computer readable instructions. The systemalso includes a machine learning system. The machine learning systemincludes a machine learning model. The machine learning modelcan be separate from or integrated with the computer processors.

The computer-readable medium(or computer-readable memory) can include any data storage technology type which is suitable to the local technical environment, including but not limited to semiconductor based memory devices, magnetic memory devices and systems, optical memory devices and systems, fixed memory, removable memory, disc memory, flash memory, dynamic random-access memory (DRAM), static random-access memory (SRAM), electronically erasable programmable read-only memory (EEPROM) and the like. In an embodiment, the computer-readable mediumincludes code-segment having executable instructions.

In some implementations, the computer processorsinclude a general purpose processor. In some implementations, the computer processorsinclude a central processing unit (CPU). In some implementations, the computer processorsinclude at least one application specific integrated circuit (ASIC). The computer processorscan also include general purpose programmable microprocessors, graphic processing units, special-purpose programmable microprocessors, digital signal processors (DSPs), programmable logic arrays (PLAs), field programmable gate arrays (FPGA), special purpose electronic circuits, etc., or a combination thereof. The computer processorsare configured to execute program code means such as the computer-executable instructionsand configured to execute executable logic that includes the machine learning model.

The computer processorsare configured to receive data including: data about conditions in the subsurface formation from the sensors associated with the wells (e.g., a gas chromatograph, a downhole pressure gauge, a downhole temperature gauge, and a fluid density meter) as well as operational data from the pumps associated with the wells. The machine learning modelis capable of processing the data to predict the volume of gas injection required to alter the subsurface gas composition to a desired level, while minimizing the use of simulation models.

The system relies on accurate predictions of gas composition and subsurface conditions. However, unexpected changes in the subsurface (e.g., variations in rock properties or gas flow) sometimes occur and can affect the accuracy of the predictions. Outlier prediction models can be used to identify if the predicted valued started to become significantly different than the previous data. This will help in flagging the data for review of the data received or model's accuracy.

The machine learning systemis capable of applying machine learning techniques to train the machine learning model. As part of the training of the machine learning model, the machine learning systemforms a training set of input data by identifying a positive training set of input data items that have been determined to have the property in question, and, in some embodiments, forms a negative training set of input data items that lack the property in question.

The machine learning systemextracts feature values from the input data of the training set, the features being variables deemed potentially relevant to whether or not the input data items have the associated property or properties. An ordered list of the features for the input data is herein referred to as the feature vector for the input data. In one embodiment, the machine learning systemapplies dimensionality reduction (e.g., via linear discriminant analysis (LDA), principle component analysis (PCA), or the like) to reduce the amount of data in the feature vectors for the input data to a smaller, more representative set of data.

In some implementations, the machine learning systemuses supervised machine learning to train the machine learning modelswith the feature vectors of the positive training set and the negative training set serving as the inputs. Different machine learning techniques-such as linear support vector machine (linear SVM), boosting for other algorithms (e.g., AdaBoost), neural networks, logistic regression, naïve Bayes, memory-based learning, random forests, bagged trees, decision trees, boosted trees, or boosted stumps—may be used in different embodiments. The machine learning model, when applied to the feature vector extracted from the input data item, outputs an indication of whether the input data item has the property in question, such as a Boolean yes/no estimate, or a scalar value representing a probability.

In some embodiments, a validation set is formed of additional input data, other than those in the training sets, which have already been determined to have or to lack the property in question. The machine learning systemapplies the trained machine learning modelto the data of the validation set to quantify the accuracy of the machine learning model. Common metrics applied in accuracy measurement include: Precision=TP/(TP+FP) and Recall=TP/(TP+FN), where precision is how many the machine learning model correctly predicted (TP or true positives) out of the total it predicted (TP+FP or false positives), and recall is how many the machine learning model correctly predicted (TP) out of the total number of input data items that did have the property in question (TP+FN or false negatives). The F score (F-score=2*PR/(P+R)) unifies precision and recall into a single measure. In one embodiment, the machine learning module iteratively re-trains the machine learning model until the occurrence of a stopping condition, such as the accuracy measurement indication that the model is sufficiently accurate, or a number of training rounds having taken place.

In some implementations, the machine learning modelis a convolutional neural network (CNN). A CNN can be configured based on a presumption that inputs to the CNN correspond to image pixel data for an image or other data that includes features at multiple spatial locations. For example, sets of inputs can form a multi-dimensional data structure, such as a tensor, that represent color features of an example digital image (e.g., a biological image of biological tissue). In some implementations, inputs to the CNN correspond to a variety of other types of data, such as data obtained from different devices and sensors of a vehicle, point cloud data, audio data that includes certain features or raw audio at each of multiple time steps, or various types of one-dimensional or multiple dimensional data. A convolutional layer of the CNN can process the inputs to transform features of the image that are represented by inputs of the data structure. For example, the inputs are processed by performing dot product operations using input data along a given dimension of the data structure and a set of parameters for the convolutional layer.

Performing computations for a convolutional layer can include applying one or more sets of kernels to portions of inputs in the data structure. The manner in which CNN performs the computations can be based on specific properties for each layer of an example multi-layer neural network or deep neural network that supports deep neural net workloads. A deep neural network can include one or more convolutional towers (or layers) along with other computational layers. In particular, for example computer vision applications, these convolutional towers often account for a large proportion of the inference calculations that are performed. Convolutional layers of a CNN can have sets of artificial neurons that are arranged in three dimensions, a width dimension, a height dimension, and a depth dimension. The depth dimension corresponds to a third dimension of an input or activation volume and can represent respective color channels of an image. For example, input images can form an input volume of data (e.g., activations), and the volume has dimensions 32×32×3 (width, height, depth respectively). A depth dimension of 3 can correspond to the RGB color channels of red (R), green (G), and blue (B).

In general, layers of a CNN are configured to transform the three dimensional input volume (inputs) to a multi-dimensional output volume of neuron activations (activations). For example, a 3D input structure of 32×32×3 holds the raw pixel values of an example image, in this case an image of width 32, height 32, and with three color channels, R,G,B. A convolutional layer of a CNN of the machine learning modelcomputes the output of neurons that may be connected to local regions in the input volume. Each neuron in the convolutional layer can be connected only to a local region in the input volume spatially, but to the full depth (e.g., all color channels) of the input volume. For a set of neurons at the convolutional layer, the layer computes a dot product between the parameters (weights) for the neurons and a certain region in the input volume to which the neurons are connected. This computation may result in a volume such as 32×32×12, where 12 corresponds to a number of kernels that are used for the computation. A neuron's connection to inputs of a region can have a spatial extent along the depth axis that is equal to the depth of the input volume. The spatial extent corresponds to spatial dimensions (e.g., x and y dimensions) of a kernel.

A set of kernels can have spatial characteristics that include a width and a height and that extends through a depth of the input volume. Each set of kernels for the layer is applied to one or more sets of inputs provided to the layer. That is, for each kernel or set of kernels, the machine learning modelcan overlay the kernel, which can be represented multi-dimensionally, over a first portion of layer inputs (e.g., that form an input volume or input tensor), which can be represented multi-dimensionally. For example, a set of kernels for a first layer of a CNN may have size 5×5×3×16, corresponding to a width of 5 pixels, a height of 5 pixel, a depth of 3 that corresponds to the color channels of the input volume to which to a kernel is being applied, and an output dimension of 16 that corresponds to a number of output channels. In this context, the set of kernels includes 16 kernels so that an output of the convolution has a depth dimension of 16.

The machine learning modelcan then compute a dot product from the overlapped elements. For example, the machine learning modelcan convolve (or slide) each kernel across the width and height of the input volume and compute dot products between the entries of the kernel and inputs for a position or region of the image. Each output value in a convolution output is the result of a dot product between a kernel and some set of inputs from an example input tensor. The dot product can result in a convolution output that corresponds to a single layer input, e.g., an activation element that has an upper-left position in the overlapped multi-dimensional space. As discussed above, a neuron of a convolutional layer can be connected to a region of the input volume that includes multiple inputs. The machine learning modelcan convolve each kernel over each input of an input volume. The machine learning modelcan perform this convolution operation by, for example, moving (or sliding) each kernel over each input in the region.

The machine learning modelcan move each kernel over inputs of the region based on a stride value for a given convolutional layer. For example, when the stride is set to 1, then the machine learning modelcan move the kernels over the region one pixel (or input) at a time. Likewise, when the stride is 2, then the machine learning modelcan move the kernels over the region two pixels at a time. Thus, kernels may be shifted based on a stride value for a layer and the machine learning modelcan repeatedly perform this process until inputs for the region have a corresponding dot product. Related to the stride value is a skip value. The skip value can identify one or more sets of inputs (2×2), in a region of the input volume, that are skipped when inputs are loaded for processing at a neural network layer. In some implementations, an input volume of pixels for an image can be “padded” with zeros, e.g., around a border region of an image. This zero-padding is used to control the spatial size of the output volumes.

As discussed previously, a convolutional layer of CNN is configured to transform a three dimensional input volume (inputs of the region) to a multi-dimensional output volume of neuron activations. For example, as the kernel is convolved over the width and height of the input volume, the machine learning modelcan produce a multi-dimensional activation map that includes results of convolving the kernel at one or more spatial positions based on the stride value. In some cases, increasing the stride value produces smaller output volumes of activations spatially. In some implementations, an activation can be applied to outputs of the convolution before the outputs are sent to a subsequent layer of the CNN.