Determining boundaries for subsurface features through pixel-wise inferencing of inversion images

Abstract

A computer implemented super-resolution system is described herein for generating high-resolution subsurface modelling results by utilizing machine learning super-resolution techniques. A subsurface feature can be modelled with a low-resolution, physics-based simulation performed on a coarse simulation grid. Low-resolution property images for one or more properties of the subsurface feature can be generated from the low-resolution model that indicate one or more subsurface properties of the subsurface feature. Utilizing the low-resolution property images as input, simulated high-resolution property images are generated with a machine learning model that is trained to predict property values for pixels at a high resolution for representing the subsurface property of the subsurface feature. The simulated high-resolution property images can be provided for characterizing the subsurface feature, such as by generating a simulated high-resolution model of the subsurface feature.

Description

Docket No. IS23.0585-WO-PCTDETERMINING BOUNDARIES FOR SUBSURFACE FEATURES THROUGH PIXEL-WISE INFLUENCING OF INVERSION IMAGESBACKGROUND OF THE DISCLOSURE

[0001] Many natural resources are located underground, including water reservoirs and hydrocarbon reservoirs, such as natural gas and oil. To access these resources, downhole drilling systems drill a wellbore from a surface, and along a trajectory, to a target location, formation, or geological feature. Modelling, simulation, analytical, and / or other tools are often implemented in connection with modem drilling systems for characterizing and understanding the subsurface. These tool use measurement data to simulate or model geological features in relation to the trajectory.

[0002] However, in many cases it may be difficult and time-consuming to obtain images and / or models of subsurface features in high resolution in order to adequately ascertain the location, shape, orientation, boundaries, etc., of subsurface features. Accordingly, given the shortcomings described above, a solution is needed to overcome the disadvantages of conventional techniques and to provide high-quality representations of subsurface features in a high resolution.BRIEF DESCRIPTION OF THE DRAWINGS

[0003] In order to describe the manner in which the above-recited and other features of the disclosure can be obtained, a more particular description will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings. For better understanding, the like elements have been designated by like reference numbers throughout the various accompanying figures. While some of the drawings may be schematic or exaggerated representations of concepts, at least some of the drawings may be drawn to scale. Understanding that the drawings depict some example embodiments, the embodiments will be described and explained with additional specificity and detail through the use of the accompanying drawings in which:

[0004] FIG. 1 is an example of a downhole system, according to at least one embodiment of the present disclosure;Docket No. IS23.0585-WO-PCT

[0005] FIG. 2A illustrates an example environment in which a subsurface superresolution system is implemented, according to at least one embodiment of the present disclosure;

[0006] FIG. 2B illustrates an example implementation of a subsurface super-resolution system as described herein, according to at least one embodiment of the present disclosure;

[0007] FIG. 3A illustrates an example workflow of generating property images from a downhole model, according to at least one embodiment of the present disclosure;

[0008] FIG. 3B illustrates an example workflow for training the subsurface superresolution machine learning model, according to at least one embodiment of the present disclosure;

[0009] FIG. 3C illustrates an example workflow for using the subsurface superresolution machine learning model to generate simulated high-resolution property images, according to at least one embodiment of the present disclosure;

[0010] FIG. 3D illustrates an example workflow for generating a simulated high- resolution model from simulated high-resolution property image, according to at least one embodiment of the present disclosure;

[0011] FIG. 4 illustrates an example workflow for generating a geological model over a geological time period, according to at least one embodiment of the present disclosure;

[0012] FIG. 5 illustrates a flow diagram for a method or a series of acts for modelling a subsurface feature as described herein, according to at least one embodiment of the present disclosure;

[0013] FIG. 6 illustrates certain components that may be included within a computer system.DETAILED DESCRIPTION

[0014] This disclosure describes a drilling system that uses a subsurface superresolution system to generate images and / or models of subsurface features in high resolution in a timely and efficient manner. For instance, subsurface models and images corresponding to subsurface properties of subsurface features may conventionally be generated based on performing physics-based simulations, such as forward or inverse modelling simulations. Such simulations may be time consuming and computationallyDocket No. IS23.0585-WO-PCT expensive. For example, obtaining high-resolution models from fine-grid physics-based simulations may be slow, costly, and inefficient. While such simulations may be performed on a coarse grid of cells to achieve faster, more efficient results, such results come at the expense of a lack of detail, less accuracy, lower resolution etc.

[0015] The subsurface super-resolution system described herein generates high- resolution images of subsurface properties based on coarse-grid simulation results. For example, after performing a more efficient, faster simulation on a coarse grid of cells and obtaining a lower-resolution subsurface model, the subsurface super-resolution system implements a subsurface super-resolution machine learning model to upscale the resolution of the lower-resolution model. For instance, the subsurface super-resolution system generates property images each associated with a specific property of a subsurface feature as indicated in the low-resolution subsurface model. The subsurface super-resolution machine learning model is trained to generate simulated high-resolution property images with an upscaled, higher resolution by predicting property values for pixels of the simulated high-resolution property images that correspond to the subsurface feature as indicated in a target low-resolution property image. In this way, the simulated high-resolution property images may simulate a high-resolution subsurface model as generated based on a fine-grid physics-based simulation, but without the significant time and computational burden. For example, the subsurface super-resolution system may assemble several simulated high- resolution property images to generate a 3-dimensional subsurface model indicating may different subsurface properties.

[0016] As will be discussed in further detail below, the present disclosure includes a number of practical applications having features described herein that provide benefits and / or solve problems associated with identifying subsurface features and / or boundaries of subsurface features. Some example benefits are discussed herein in connection with various features and functionalities provided by a subsurface super-resolution system implemented on one or more computing devices. It will be appreciated that benefits explicitly discussed in connection with one or more embodiments described herein are provided by way of example and are not intended to be an exhaustive list of all possible benefits of the subsurface super-resolution system.Docket No. IS23.0585-WO-PCT

[0017] For example, the subsurface super-resolution system described herein can be implemented to provide high-resolution subsurface modelling results with substantially the same accuracy, precision, and detail to those of a fine-grid physics-based simulation, but with a significant increase in speed. For instance, by performing a coarse-grid physicsbased simulation, which takes a fraction of the time of a fine-grid simulation, and utilizing the subsurface super-resolution machine learning model to upscale the resolution of the coarse-grid subsurface model to a high-resolution subsurface model, which can be performed nearly instantly, the subsurface super-resolution system can provide significant time savings in characterizing, analyzing, and understanding various subsurface properties of a subsurface geology. Indeed, the techniques described herein of the subsurface superresolution system can be performed in a matter of seconds or minutes, in contrast to delaying hours or days for conventional fine-grid physics-based simulations.

[0018] The time-saving benefits of the subsurface super-resolution system may be realized in addition to an increase in computational efficiency. For example, full scale physics-based simulations may consume a considerable amount of computing resources, which is magnified as resolution is increased. Thus, generating a model at a high resolution with a physics-based simulation may come at the cost of a significant computational expense. In contrast, by performing a physics-based simulation on a coarse grid, which utilizes significantly fewer computing resources, and upscaling resolution with the subsurface super-resolution system, which can be implemented with as few as 1 computing core, the techniques described herein can provide a significant improvement to the efficiency and operation of the computing devices on which it is implemented.

[0019] Further, the subsurface super-resolution system can be implemented with respect to property images that are derived from a low-resolution model. A physics-based simulation may simulate, model, calculate, and otherwise provide results for many different properties, characteristics, behaviors, etc. of a subsurface feature. Depending on a given application, only some of the modelled properties may be significant, critical, or otherwise of interest for characterizing and understanding a subsurface feature with respect to the application. By generating property images from the subsurface model that correspond to properties of interest, the techniques described herein provide valuable adaptability and flexibility for focusing the analysis and modelling of a subsurface featureDocket No. IS23.0585-WO-PCT to those properties which are relevant to the application at hand. This focusing of subsurface properties additionally adds to the time and computing efficiency benefits.

[0020] As illustrated in the following discussion, this disclosure uses a variety of terms to describe the features and advantages of one or more implementations described in this disclosure. Additional details are provided to clarify the meaning of some of these terms, while details regarding other terms may be provided later in the document.

[0021] As used herein, a “feature” such as a geological feature, downhole feature, or a subsurface feature may be any element of a geological formation. A geological feature may include a reservoir, pay zone, subterranean target, or any other underground feature for which it may be desirable to know its location, orientation, position, etc. For instance, a geological feature may include a geological structure, such as a formation. A feature may include the entire geological structure. A downhole feature may include a volume of space, including one or more structures, rock types, material types, and so forth. In some embodiments, a downhole feature may include a specific structure of a set of structures, such as a fluid reservoir. A feature may be three-dimensional. For example, a feature may include a three-dimensional surface having variations in latitude, longitude, and depth. In some embodiments, a feature may be a reservoir, pay zone, or underground resource, such as an oil, gas, or water reservoir, a source of geothermal energy, or any other subterranean target.

[0022] A subsurface feature may have and / or may be characterized by one or more “properties.” Properties, such as subsurface properties or feature properties, may refer to any physical property of a subsurface feature. For example, a property may refer to a rock property of a feature. Rock properties may include one or more of a porosity, a rock stress such as a vertical stress and / or a maximum horizontal stress, a permeability, or a capillary pressure of some or all of a feature. In some cases, a property may be static (e.g., not dynamic or fluid). For instance, a static rock property may include static properties of the rock or material of a feature which may not change over the short term, but which may change (e.g., slowly) over a substantial geological period. For instance, static properties (e.g., unlike fluid properties) may not be continually changing, and may not change over the course of days, weeks, years, or even millennia. Rock properties, however, may exhibit slow changes over the course of 100s to millions of years.Docket No. IS23.0585-WO-PCT

[0023] As used herein, “machine-learning model” refers to a computer model or computer representation that may be trained (e.g., optimized) based on inputs to approximate unknown functions. For instance, a machine-learning model may include, but is not limited to, a neural network or deep neural network (e.g., a convolutional neural network (CNN), graph neural network, transformer, etc ), a classification and regression tree (e.g., a gradient-boosted decision tree), a linear regression model, a logistic regression model.

[0024] As used herein, a “model,” such as downhole model, a subsurface model, and the like, refers to a representation of a subsurface feature, formation, or other geological area. A model may be generated based on performing a simulation of the subsurface geology including various factors and / or measurements. For example, the simulation may be a physics-based simulation that may incorporate various equations and other physics models for representing and / or determining characteristics of the subsurface geology. In some cases, the simulation may be based on forward modelling, inversion modelling, finite element analysis, and / or any other analytical method. A model may have a visual component, for example for visually representing one or more properties or other characteristics of a feature. For instance, a model may be a 2- or 3-dimensional representation of a subsurface feature. A model may have an analytical component, for example, for representing various measurements, values, or other characteristics of a feature (e.g., in a form that is not strictly visual or illustrative). A model may include and / or may incorporate many different properties of a feature. For example, in some cases a model determines and / or represents many rock properties of a geological area of interest. For example, a model may indicate a rock porosity, vertical stress, and horizontal stress of a feature.

[0025] A model may be characterized by a resolution representing a certain quantity of measurements or samples per unit length, area, or volume. For instance, a model with a higher density of information may be considered a high-resolution model, and a model with a lower information density may be considered a low-resolution model. High- resolution models may be generated based on fine grid simulations. For example, high- resolution models may have cell sizes of 400 meters by 400 meters, 200 meters by 200 meters, or less. Low-resolution models may be generated based on coarse-grid simulations.Docket No. IS23.0585-WO-PCTFor instance, low-resolution models may have cell sizes of 600 meters by 600 meters, 800 meters by 800 meters, or more. In this way, high-resolution models may include information at a higher resolution that low-resolution models based on a higher resolution of an underlying simulation. As discussed herein, high-resolution models and fine-grid simulations may take a significant amount of time to generate / perform as compared to low- resolution models.

[0026] As used herein, an “image” refers to a representation of information in a matrix form. For example, an image may be a 1-, 2-, or 3-dimensional table, array, or other matrix that may contain information at some or all locations or entries of the matrix. In some cases, an image matrix may include entries comprising pixels with pixel data (e.g., red, green, blue, and intensity information) for visually representing underlying information. In some cases, an image matrix may include entries comprising numbers, measurements, symbols, or other values at one or more positions. In some embodiments of the present disclosure, an image may be a property image for representing a property of a feature in matrix form. For example, the entries of a property image may be property values representing a porosity, rock stress, or any other property of a subsurface feature. The property values of a property image may be represented as pixels in a visual image matrix. In some cases, a property image may represent a single property of an underground feature. As described herein, property images may be generated or derived from a model of a subsurface feature, for instance, slicing, filtering, and / or cropping the subsurface model.

[0027] Additional details will now be provided regarding systems described herein in relation to illustrative figures portraying example implementations. For example, FIG. 1 shows one example of a downhole system 100 for drilling an earth formation 101 to form a wellbore 102. The downhole system 100 includes a drill rig 103 used to turn a drilling tool assembly 104 which extends downward into the wellbore 102. The drilling tool assembly 104 may include a drill string 105, a bottomhole assembly (“BHA”) 106, and a bit 110, attached to the downhole end of the drill string 105.

[0028] The drill string 105 may include several joints of drill pipe 108 connected end- to-end through tool joints 109. The drill string 105 transmits drilling fluid through a central bore and transmits rotational power from the drill rig 103 to the BHA 106. In some embodiments, the drill string 105 further includes additional downhole drilling tools and / orDocket No. IS23.0585-WO-PCT components. The drill pipe 108 provides a hydraulic passage through which drilling fluid is pumped from the surface to the bit 110.

[0029] The BHA 106 may include other downhole drilling tools and components. Examples of additional BHA components include measurement-while-drilling (“MWD”) tools, logging-while-drilling (“LWD”) tools, and measurement sensors.

[0030] The BHA 106 may further include a rotary steerable system (RSS). The RSS may include directional drilling tools that change a direction of the bit 110, and thereby the trajectory of the wellbore. At least a portion of the RSS may maintain a geostationary position relative to an absolute reference frame, such as one or more of gravity, magnetic north, or true north. Using measurements obtained with the geostationary position, the RSS may locate the bit 110, change the course of the bit 110, and direct the directional drilling tools on a projected trajectory. The RSS may steer the bit 110 in accordance with or based on a trajectory for the bit 110. For example, a trajectory may be determined for directing the bit 110 toward one or more subterranean targets such as an oil or gas reservoir.

[0031] The downhole system 100 may include or may be associated with a client device 112 with a subsurface super-resolution system 120 implemented thereon (e g., or with a client application implemented thereon for accessing the subsurface super-resolution system 120 as described herein). The subsurface super-resolution system 120 may facilitate identifying and / or characterizing subsurface features, for example, to facilitate directing or steering the bit 110 in relation to a subsurface feature.

[0032] FIG. 2A illustrates an example environment 200 in which a subsurface superresolution system 120 is implemented in accordance with one or more embodiments described herein. As shown in FIG. 2A, the environment 200 includes a server device 114. The server device 114 may include one or more computing devices (e.g., including processing units, data storage, etc.) organized in an architecture with various network interfaces for connecting to and providing data management and distribution across one or more client systems.

[0033] As shown in FIG. 2A, the server device 114 may be connected to and may communicate with (either directly or indirectly) a client device 112 through a network 116. The network 116 may include one or multiple networks and may use one or more communication platforms and / or technologies suitable for transmitting data. The networkDocket No. IS23.0585-WO-PCT116 may refer to any data link that enables transport of electronic data between devices of the environment 200. The network 116 may refer to a hardwired network, a wireless network, or a combination of a hardwired network and a wireless network. In one or more embodiments, the network 116 includes the internet. The network 116 may be configured to facilitate communication between the various computing devices via well-site information transfer standard markup language (WITSML) or similar protocol, or any other protocol or form of communication.

[0034] The client device 112 may be representative of one or multiple client devices, and may refer to various types of computing devices. For example, the client device 112 may include a mobile device such as a mobile telephone, a smartphone, a personal digital assistant (PDA), a tablet, a laptop, or any other portable device. Additionally, or alternatively, the client device 112 may include one or more non-mobile devices such as a desktop computer, server device, surface or downhole processor or computer (e.g., associated with a sensor, system, or function of the downhole system), or other nonportable device. In one or more implementations, the client device 112 includes graphical user interfaces (GUI) thereon (e.g., a screen of a mobile device). In addition, or as an alternative, one or more of the client device 112 may be in communication with (e.g., wired or wirelessly) a display device having a graphical user interface thereon for providing a display of system content. The server device 114 may similarly refer to various types of computing devices. Each of the devices of the environment 200 may include features and / or functionalities described below in connection with FIG. 6.

[0035] As shown in FIG. 2A, the environment 200 may include a subsurface superresolution system 120 implemented on the server device 114. While shown on the server device 114, the subsurface super-resolution system 120 may be implemented wholly or in part on the client device 112, across the server device 114 and the client device 112, or on or across one or more additional devices, such that different portions or components of the subsurface super-resolution system 120 are implemented on different computing devices in the environment 200.

[0036] The client device 112 may include a client application 118. The client application 118 may include an application or interface for interacting with and / or receiving the features of the subsurface super-resolution system 120 as described herein.Docket No. IS23.0585-WO-PCTIn some embodiments, one or more of the functionalities or features of the subsurface super-resolution system 120 may be carried out or performed on or by the client application 118. In this way, the environment 200 may be a cloud computing environment, and the subsurface super-resolution system 120 may be implemented across one or more devices of the cloud computing environment in order to leverage the processing capabilities, memory capabilities, connectivity, speed, etc., of such cloud computing environments in order to facilitate the features and functionalities described herein.

[0037] FIG. 2A illustrates an example implementation of the subsurface superresolution system 120 as described herein, according to at least one embodiment of the present disclosure. The subsurface super-resolution system 120 may include a subsurface model manager 122, a property image engine 124, an image resolution manager 126 implementing one or more subsurface super-resolution machine learning models 129, and a model simulator 128. The subsurface super-resolution system 120 may also include a data storage 130 having subsurface models 132, property images 134 stored thereon. While one or more embodiments described herein describe features and functionalities performed by specific components 122-128 of the subsurface super-resolution system 120, it will be appreciated that specific features described in connection with one component of the subsurface super-resolution system 120 may, in some examples, be performed by one or more of the other components of the subsurface super-resolution system 120.

[0038] By way of example, the simulation and / or modelling functionalities of the subsurface model manager 122 may be delegated to other components of the subsurface super-resolution system 120. As another example, while property images may be generated by the property image engine 124, in some instances, some or all of these features may be performed by the image resolution manager 126 (or other component of the subsurface super-resolution system 120). Indeed, it will be appreciated that some or all of the specific components may be combined into other components and specific functions may be performed by one or across multiple components 122-128 of the subsurface superresolution system 120.

[0039] Additionally, while FIG. 1, for example, depicts the subsurface superresolution system 120 implemented on a client device 112 of the downhole system, it should be understood that some or all of the features and functionalities of the subsurfaceDocket No. IS23.0585-WO-PCT super-resolution system 120 may be implemented on or across multiple client devices 1 12 and / or server devices 114. For example, simulations may be performed and models generated by the subsurface model manager 122 on a (e.g., local) client device, and high- resolution property images and / or subsurface models may be generated based on upscaling resolution with the subsurface super-resolution machine learning models on one or more of a remote, server, or cloud device. Indeed, it will be appreciated that some or all of the specific components 122-128 may be implemented on or across multiple client devices 112 and / or server devices 114, including individual functions of a specific component being performed across multiple devices.

[0040] As just mentioned, the subsurface super-resolution system 120 includes a subsurface model manager 122. The subsurface model manager 122 may facilitate accessing one or more subsurface models of a subsurface geology, feature, etc. For example, in some cases, the subsurface model manager 122 may perform or run one or more physics-based simulations, such as a forward modelling simulation, for analytically modelling a subsurface feature. In some cases, the subsurface model manager 122 may receive a (e.g., completed) subsurface model and / or the results of a simulation of a subsurface feature, for example, from another component or analytical tool associated with the subsurface super-resolution system 120.

[0041] FIG. 3A illustrates a workflow for generating or deriving property images from subsurface models, according to at least one embodiment of the present disclosure. As just mentioned, the subsurface model manager 122 may access one or more subsurface models 132. For instance, in some embodiments, the subsurface model manager 122 accesses a low-resolution model 132L. In some embodiments, the subsurface model manager 122 accesses a high-resolution model 132H.

[0042] The subsurface models 132 (e.g., the low-resolution model 132L and high- resolution model 132H) may be generated based on physics-based equations, laws, and other principles. For example, the subsurface models 132 may be generated from a simulation that may be a computational model that predicts the geological and physical properties of the earth’s subsurface based on data from seismic surveys, well logs, geological maps, etc., in relation to physical laws, principles, and other physical observations about the subsurface. For instance, complex mathematical equations mayDocket No. IS23.0585-WO-PCT describe and simulate physical characteristics and / or processes of the underground geology such as fluid flow, heat transfer, mechanical deformation, rock properties, etc. The simulation may take into account geological layers, fault systems, and other relevant features. Once established, the simulation may be performed (e.g., through numerical methods) to simulate a variety of information about a subsurface geology including a subsurface feature of interest. For example, the subsurface models 132 may indicate information regarding the structural framework, shape and size, lithology, fluid content, rock properties, or other characteristics of a subsurface feature. The subsurface models 132 may generally be 3-dimensional (e.g., cube) models, however, in some cases the subsurface models 132 may be 2-dimensional.

[0043] In this way, the subsurface models 132 may describe, indicate, or otherwise characterize any number of physical properties of the subsurface geology. For instance, in some cases, the subsurface models 132 may be complex, robust models that may indicate many properties and characteristics of a subsurface feature.

[0044] The low-resolution model 132L and high-resolution model 132H may correspond to each other, or in other words, may model a similar subsurface feature but at different resolutions. For instance, the low-resolution model 132L and high-resolution model 132H may each be generated from the same simulation as performed on different coarseness or resolution of simulation grids. For instance, the high-resolution model 132H may be generated based on a simulation of a subsurface feature of interest with a fine-grid of cells of the simulation, and the low-resolution model 132L may be generated based on the same simulation, but as run with a coarse-grid or lower-resolution of cells of the simulation. In this way, the low-resolution models 132L may characterize, indicate, or otherwise represent similar information to that of the high-resolution model 132H, but with a lower density of information.

[0045] In some embodiments, the subsurface models 132 may represent an entire subsurface basin, oilfield, or reservoir. For instance, the simulation may be performed on a basin scale. However, the subsurface models 132 may be of any size, scale, or magnitude, such as spanning one or multiple subsurface features, a portion of a subsurface feature, some or all of a basin, etc.Docket No. IS23.0585-WO-PCT

[0046] As mentioned above in connection with FIG. 2B, the subsurface superresolution system 120 includes a property image engine 124. The property image engine 124 facilitates generating property images 134 for use with training and implementing one or more subsurface super-resolution machine models.

[0047] The property image engine 124 may generate the property images 134 to represent one or more properties of a subsurface feature as indicated in a corresponding subsurface model. For example, in some cases, the property images 134 may indicate a single property of interest. As an example, the property images 134 may each represent a single rock property, such as porosity, vertical rock stress, or maximum horizontal rock stress. In some cases, the property images 134 may indicate permeability or capillary pressure. The property image engine 124 may generate the property images 134 indicating the (e.g., single) property based on filtering the associated subsurface model and / or by removing some of the information indicated in the associated subsurface model that does not correspond to the property of interest. In this way, the property images 134 may be a simplified representation of a property of a subsurface feature.

[0048] The property images 134 may represent the property based on property values 140 indicated in image or matrix form. For example, the subsurface model may indicate certain property values 140 (e.g., numbers, values, pixels, or other information) at specific locations in a 1-, 2-, or 3-dimensional array or matrix, and the property images 134 may similarly represent those property values at corresponding locations.

[0049] In some embodiments, the property images 134 may have the same dimensionality as the subsurface model 132. For instance, the subsurface models 132 may be 3-dimensional and the property images 134 may also be 3 -dimensional. In some embodiments, the property images 134 may have a lower dimensionality than the subsurface models 132. For example, the property image engine 124 may slice or crosssection the associated subsurface model one or more times to generate a property image 134. The slice may be a perpendicular slice or another transverse (e.g., non-perpendicular) slice. In some embodiments, the property images 134 may be an entire cross-section of an associated subsurface model. For example, the property image engine 124 may slice the subsurface model and may utilize the slice as a property image 134. For instance, the property image 134 may have the same dimensions as the subsurface model. In someDocket No. IS23.0585-WO-PCT embodiments, one or more property images 134 may be less than all of a cross-section of an associated subsurface model. For example, the property image engine 124 may slice the subsurface model and may crop (e.g., before or after slicing) some or all of the cross-section to form the property image 134. For instance, property images 134 may have smaller dimensions than the cross-section and / or the subsurface model. In some embodiments, the property images 134 may have square or cubed dimensions.

[0050] In some embodiments, the property values 140 in the property images 134 may be represented by pixels having color and / or intensity values corresponding to a certain degree or measure of the property. In this way, the property images 134 may indicate a form or shape of the property of interest and / or of the subsurface feature. In some cases, the property values 140 may be indicated by a number or value of a measurement or unit of a property for one or more entries of the matrix. In this way, the present techniques may be implemented with property images that may not necessarily be images in the traditional sense, with pixels having color and intensity values, but may be implemented with respect to any assortment of information in a tabulated or matrix form.

[0051] In some embodiments, the property image engine 124 generates low-resolution property images 136 from the low-resolution model 132L. In some cases, the low- resolution property images 136 may have a resolution of 50 by 50 (e.g., pixels or entries), 25 by 25, or lower. The property image engine 124 may generate high-resolution property images 138 from the high-resolution model 132H. The high-resolution property images 138 may have a resolution of 100 by 100 (e.g., pixels or entries), 200 by 200, or higher. In this way, the resolution of the property images 134 may be based on or derived from the resolution of the subsurface model from which the property images 134 are generated.

[0052] In some embodiments, the property image engine 124 generates high-resolution property images 138 that correspond to the low-resolution property images 136. For instance, for each low-resolution property image 136, the property image engine 124 may generate a high-resolution property image that corresponds to and / or spans a same geological space. In this way, each pair of low-resolution property images 136 and high- resolution property images 138 may depict or represent the same physical space of a geological feature, but may indicate the associated property with a different level of fidelity, resolution, or information density. As described herein, the property image engineDocket No. IS23.0585-WO-PCT124 may generate pairs of low- and high-resolution property images for the purpose of training a subsurface super-resolution machine learning model. For example, the low- resolution property images 136 and corresponding high-resolution property images 138 may be implemented as training data. In some cases, the property image engine 124 generates low-resolution property images 136 without generating a corresponding high- resolution property image 138, for example, for inferencing with the (e.g., trained) subsurface super-resolution machine learning model to upscale the resolution of the low- resolution property images 136 as described herein.

[0053] In some embodiments, the property image engine 124 generates several sets of property images 134 corresponding to several different subsurface properties. For example, the property image engine 124 may generate a first set of low-resolution property images 136-1 corresponding with a first property, a second set of low-resolution property images 136-2 corresponding with a second property, and so on to an nth set of low-resolution property images 136-n corresponding with an nth property. For instance, the first set of low-resolution property images 136-1 may represent the porosity of a subsurface feature. The second set of low-resolution property images 136-2 may represent the vertical rock stress of the subsurface feature. The nth set of low-resolution property images 136-n may represent the maximum vertical stress of the subsurface feature. Similarly, the property image engine 124 may generate a first set of high-resolution property images 138-1 corresponding with the first property, a second set of high-resolution property images 138- 2 corresponding with the second property, and an nth set of high-resolution property images 138-n corresponding with the nth property. In this way, the property image engine 124 may generate any number of sets of property images 134 for representing any number of different properties of a subsurface feature. Additionally, each set of property images may include any number of property images.

[0054] FIG. 3B illustrates an example workflow 300 for training a subsurface superresolution machine learning model 129, according to at least one embodiment of the present disclosure. The workflow 300 may be performed by the subsurface superresolution system 120. The subsurface super-resolution system 120 may facilitate training the subsurface super-resolution machine learning model 129, for example, by implementing the image resolution manager 126.Docket No. IS23.0585-WO-PCT

[0055] The subsurface super-resolution system 120 may train the subsurface superresolution machine learning model 129 based on training data 142. For example, the training data 142 may include the low-resolution property images 136, and the high- resolution property images 138 which correspond to the low-resolution property images 136.

[0056] The subsurface super-resolution machine learning model 129 may be trained to generate predicted high-resolution property images 144 based on the low-resolution property images 136. The predicted high-resolution property images 144 include predicted property values 146 for one or more pixels (e.g., or matrix entries) of the predicted high- resolution property images 144 based on the property values 140 of the low-resolution property images 136. For example, the subsurface super-resolution machine learning model 129 may predict, for each pixel of the predicted high-resolution property images 144, the predicted property values 146 based on a property of a subsurface feature as indicated by the property values 140 of the low-resolution property images 136 to represent the property and subsurface feature in a higher resolution.

[0057] The subsurface super-resolution machine learning model 129 may be agnostic to any particular type or form of machine learning model, and may be implemented as any of a variety of machine learning models and / or artificial intelligence techniques. For instance, the subsurface super-resolution machine learning model 129 may be a neural network, a tree- or decision-based model, a diffusion model, a generative model, or any type of machine learning model suitable for implementing the present techniques of upscaling resolution of low-resolution property images.

[0058] In accordance with at least one embodiment of the present disclosure, the subsurface super-resolution machine learning model 129 includes a neural network architecture, such as higher and lower neural network layers. For example, the subsurface super-resolution machine learning model 129 is an image processing neural network, similar to one or more of the machine-learning model types described above. In various instances, the subsurface super-resolution machine learning model 129 is a U-Net neural network, or a U-Net++ neural network. In some instances, the subsurface super-resolution machine learning model 129 is a large generative model, or a multi-modal image-to-image neural network. In some cases, the subsurface super-resolution machine learning modelDocket No. IS23.0585-WO-PCT129 is a latent-variable generative model such as a diffusion probability model. In some cases, the subsurface super-resolution machine learning model 129 is a lightweight model such as an enhanced deep super-resolution model.

[0059] To elaborate, in some instances, the subsurface super-resolution machine learning model 129 is a convolutional neural network (CNN) that includes several neural network layers, such as lower neural network layers that form an encoder, and higher neural network layers that form a decoder. For example, the encoder maps or encodes input images into feature vectors (i.e., latent object feature maps or latent object feature vectors) by processing each input image through various neural network layers (e.g., convolutional, rectified linear unit (ReLU), and / or pooling layers) to encode pixel data from the input images into feature vectors (e.g., a string of numbers in vector space representing the encoded image data). For instance, the encoder of a subsurface super-resolution machine learning model processes input images to encode image features corresponding to subsurface property values or measurements from the low-resolution property images 136.

[0060] Additionally, in various implementations, the subsurface super-resolution machine learning model 129 includes higher neural network layers that form a decoder, which may include fully connected layers and / or a classifier function (e.g., a SoftMax or a sigmoid function). In these implementations, the decoder processes the feature vectors to decode detected resistivity change interfaces, horizons, and / or reservoir boundaries in an encoded image vector and generate a predicted high-resolution property image.

[0061] In some implementations, the subsurface super-resolution machine learning model 129 includes multiple super-resolution machine learning models (e.g., of the same or different types of models) each corresponding to a particular property type of a subsurface feature. For example, each super-resolution machine learning model may be trained to upscale the resolution of a particular set of low-resolution property images indicating a given property type. Multiple super-resolution machine learning models may accordingly be trained to upscale low-resolution property images 136 indicating multiple different types of subsurface properties.

[0062] In some embodiments, the subsurface super-resolution machine learning model 129 may be (e.g., a single model) trained to process low-resolution property images of any type and generate corresponding predicted high-resolution property images of that type.Docket No. IS23.0585-WO-PCTFor instance, the subsurface super-resolution machine learning model 129 may be trained to identify a property type of an input image and generate the predicted high-resolution property images 144 based on the identified property type.

[0063] In some cases, the low-resolution property images 136 may be normalized in order to facilitate the subsurface super-resolution machine learning model 129 operating with respect to any property type. For instance, the property values (e.g., multiple different types of properties) of the low-resolution property images 136 may be normalized to a common or generic unit (or unitless) value such that the subsurface super-resolution machine learning model 129 may be trained and implemented without respect to a particular unit or measure of a given property. In some embodiments, the subsurface superresolution machine learning model 129 may perform this normalization, or another component of the subsurface super-resolution machine learning model 129 may perform normalization. In some embodiments, the predicted high-resolution property images 144 may be generated to have a native unit or measure of the underlying subsurface property. For instance, the subsurface super-resolution system 120 (or other component of the subsurface super-resolution system 120) may de-normalize an output image to generate the predicted high-resolution property images 144 with respect to the units of the subsurface property as indicated in the associated low-resolution property image 136.

[0064] In some embodiments, the subsurface super-resolution machine learning model 129 may be suited to operate with respect to property images of many different subsurface properties as just descried based on the various select properties of the subsurface feature exhibiting a similar form or shape. For example, in many cases, rock properties such as porosity, rock stress, permeability, etc. may each take the shape or form of channel-like features. For instance, the shape and / or form of these properties may be similarly manifest as bends, fingers, meanders, and other similar bodies or volumes. In some embodiments, the shape or form of each property may not necessarily resemble each other, but rather may be manifest through features, shapes, and / or forms, that are similarly channel-like in nature.

[0065] In some embodiments, the subsurface super-resolution machine learning model 129 is trained based on a loss model 151. The loss model 151 may have one or more loss functions (e.g., cross-entropy loss and / or image similarity loss). In various implementations, the subsurface super-resolution system 120 uses the loss model 151 toDocket No. IS23.0585-WO-PCT determine an error or loss amount, which the subsurface super-resolution system 120 provides back to the subsurface super-resolution machine learning model 129 as label feedback 153 to train and fine-tune the subsurface super-resolution machine learning model 129.

[0066] To further elaborate, in various implementations, the subsurface superresolution system 120 compares the high-resolution property images 138 to the predicted high-resolution property images 144 using the loss model 151 to generate the label feedback 153 indicating an error or loss amount. The predicted high-resolution property images 144 is generated by the subsurface super-resolution machine learning model 129 based on low-resolution property images 136 corresponding to the high-resolution property images 138.

[0067] Additionally, in one or more implementations, the subsurface super-resolution system 120 uses the label feedback 153 to train, optimize, and / or fine-tune the layers of the subsurface super-resolution machine learning model 129 through techniques like backpropagation and / or end-to-end learning. In some implementations, the subsurface super-resolution system 120 uses an optimizer algorithm such as the Adam optimizer and / or another optimization algorithm for stochastic gradient descent (SGD) to train the machine learning models. Furthermore, the subsurface super-resolution system 120 may iteratively fine-tune and train the layers of the subsurface super-resolution machine learning model 129 until they converge, for a set number of iterations, until the training data is exhausted, or until a satisfactory level of accuracy is achieved.

[0068] In various implementations, the subsurface super-resolution system 120 uses one or more data augmentation techniques and / / -fold ensembles. For instance, the subsurface super-resolution system 120 may augment the training data 142 to generate additional or synthetic training data. For example, the subsurface super-resolution system 120 may create instances of the training data 142 that are horizontally and / or vertically flipped, randomly rotated up to 90 degrees, randomly adjusted for brightness and contrast levels, subjected to col or jittering, and / or modified based on random Gaussian noise values. Modification of the training data 142 and / or augmenting the training data 142 in this way may increase the robustness of the subsurface super-resolution machine learning model 129.Docket No. IS23.0585-WO-PCT

[0069] Once trained, in various implementations, the subsurface super-resolution system 120 uses the subsurface super-resolution machine learning model 129 to automatically generate, from low-resolution property images, high-resolution property images that simulate high resolution property images generated from a fine-grid simulation. FIG. 3C illustrates an example workflow 301 for using the subsurface superresolution machine learning model 129 to generate simulated high-resolution property images 150, according to at least one embodiment of the present disclosure.

[0070] In FIG. 3C, the subsurface super-resolution machine learning model 129 represents a trained model with tuned (e.g., neural network) layers and other trained components. The subsurface super-resolution machine learning model 129 generates a simulated high-resolution property image 150 from a target low-resolution property image 152.

[0071] In some embodiments, the target low-resolution property images 152 represent property images generated or derived from a target low-resolution model 154. For example, the target low-resolution model 154 may be a model generated from a simulation on a coarse-grid scale. Accordingly, the target low-resolution model 154, and the target low- resolution property images 152, may have a lower resolution that could otherwise have been achieved by performing the simulation on a finer-grid scale. The target low-resolution model 154 may be a subsurface model that is generated without generating a corresponding high-resolution model. For instance, a subsurface feature, such as a basin or reservoir may be modeled only by the target low-resolution model without any corresponding high- resolution model for providing access to finer detail. Accordingly, the target low-resolution property images may represent various properties of interest of a subsurface feature, albeit at low resolution, and may be implemented in connection with the subsurface superresolution machine learning model 129 to generate corresponding simulated high- resolution property images 150.

[0072] The simulated high-resolution property images 150 may be utilized for facilitating various downstream applications 156. For example, the simulated high- resolution property images 150 may be utilized in various downstream models and workflows. For instance, the simulated high-resolution property images 150 may facilitate improving the accuracy of other models that map subsurface geological features. ForDocket No. IS23.0585-WO-PCT instance, the subsurface super-resolution system 120 may provide the simulated high- resolution property images 150 as an input to a system that performs other geophysical modelling. In one particular example, the subsurface super-resolution system 120 may be particularly suited for upscaling low-resolution models and property images that indicated various static and / or rock properties of a subsurface feature. In some cases, the simulated high-resolution property images 150 may facilitate performing one or more fluid or dynamic analysis or models of a subsurface feature for characterizing various dynamic or fluid subsurface properties based on the (e.g., static) rock properties as indicated by the simulated high-resolution property images 150.

[0073] In another instance, the simulated high-resolution property images 150 may be utilized to provide improved information about the lithology, reservoir properties, structure, etc. expected in a section of the geological area being (or to be) drilled, for instance, as part of a planning stage of a wellbore. For example, the various properties indicated in the simulated high-resolution property images 150 may facilitate determining, with a higher fidelity, chance-of-success maps and / or risk maps for drilling, forming, or operating a wellbore to reach, access, or produce from an underground reservoir. For instance, such maps may facilitate understanding the chance-of-success or failure with accessing a reservoir at a certain location or at a certain angle, stimulating the reservoir with one or more techniques, producing underground resources from a reservoir, etc. As such, it may be desirable that these chance-of-success and / or risk maps be produced with as high a resolution as reasonably possible. Accordingly, the super-resolution techniques described herein may facilitate generating these maps with an increased resolution that may be difficult, costly, or otherwise not possible through traditional, physics-based modelling techniques.

[0074] In another example, the simulated high-resolution property images 150 may be utilized to facilitate modelling a subsurface geology over a geological time period. For instance, certain time intervals and / or geological events may be more relevant, interesting, and / or critical than others, and as such may be modeled with a high-resolution physicsbased simulation to capture detail with a high fidelity. Other events and time intervals may be less critical or of less interest, and such events and time intervals may be represented by utilizing the super-resolution techniques described herein to provide a similar level of detailDocket No. IS23.0585-WO-PCT and resolution for at least some properties of a subsurface feature, but without incurring the time and computing expenses associated with performing a full-scale simulation.

[0075] In another example, the simulated high-resolution property images 150 may be utilized to inform the steering of a wellbore, such as via automated geosteering. For instance, the simulated high-resolution property images 150 may be utilized in various other workflows for identifying subsurface features and / or relevant properties of subsurface features for use in adjusting a drill bit’s trajectory to steer the drill bit in relation to a subsurface feature.

[0076] In some implementations, the simulated high-resolution property images 150 may be viewed and / or reviewed by a user. For example, the subsurface super-resolution system 120 may present the simulated high-resolution property images 150 (e.g., and / or other products generated based on the simulated high-resolution property images) via a graphical user interface to a user.

[0077] In this way, the subsurface super-resolution system 120, and in particular the subsurface super-resolution machine learning model(s) 129, may facilitate generating property images that exhibit a high resolution and / or a high fidelity based on a simulation that is performed on only a coarse grid of cells. This is advantageous in that generating the simulated high-resolution property images 150 as described herein can be performed significantly faster than is possible by performing a fine-grid simulation to generate high- resolution property images. For instance, in some cases, high-resolution modelling based on a fine-grid simulation can delay from several hours up to several days in completing the simulation and generating a corresponding high-resolution model (e.g., high-resolution property images).

[0078] Additionally, such a high-resolution simulation can often occupy significant computing resources to complete such a task, for example, by utilizing up to 20 or more computing cores. In contrast, by performing the (e.g., same) simulation on a lower- resolution, coarse grid of cells to generate a low-resolution model, and utilizing the subsurface super-resolution system 120 to upscale the resolution as described herein, corresponding high resolution property images can be simulated in a matter of minutes, and such can be performed by utilizing significantly fewer computing resources, such as only one computing core. Further, because of the significant cost of performing a fine-gridDocket No. IS23.0585-WO-PCT simulation, a practical limit exists for the size and / or robustness of a model that can be generated based on fine-grid simulations. For instance, generating a high-resolution model of a substantial size, for a large basin for example, may not be practical through conventional, fine-grid physics-based simulation techniques. However, through use of the subsurface super-resolution system 120 as described herein, such a large-scale basin may be effectively modelled by performing a coarse-grid simulation and upscaling the resolution with the techniques described herein. Accordingly, basin-modelling with the subsurface super-resolution system as described herein may not have practical limits on model size, or at least may have practical limits that are significantly larger than with conventional techniques.

[0079] Thus, the subsurface super-resolution system 120 can obtain a substantially equivalent result to a high-resolution simulation or physics-based model, while significantly lowering the time and computing expense of doing so. Indeed, the simulated high-resolution property images may correctly represent the subsurface feature and subsurface properties with substantially the same precision and accuracy as do high- resolution property images generated from a high-resolution simulation. For instance, in some cases, the simulated high-resolution property images are up to 95%, up to 99%, or more accurate to the true high-resolution property results of a physics-based simulation.

[0080] As described herein, the subsurface super-resolution system 120 includes a model simulator 128. The model simulator may facilitate simulating a high-resolution model of a subsurface feature or geological area based on various simulated high-resolution property images 150 generated by the subsurface super-resolution system 120 as described herein. FIG. 3D illustrates an example workflow 302 for generating a simulated high- resolution model 158 from simulated high-resolution property images 150, according to at least one embodiment of the present disclosure.

[0081] As described above, the subsurface super-resolution system 120 may generate simulated high-resolution property images 150 for a variety of different subsurface properties of a subsurface feature as modeled and represented by a low-resolution subsurface model. For example, various sets of simulated high-resolution property images may be generated for various, different properties. For instance, the simulated high- resolution property images 150 may include a first set of simulated high-resolutionDocket No. IS23.0585-WO-PCT property images 150-1 associated with a first property, a second set of simulated high- resolution property images 150-2 associated with a second property, and so on to an nth set of simulated high-resolution property images 150-n associated with an nth property. For instance, the various properties of the simulated high-resolution property images 150 may be various, different rock properties of a subsurface feature, such as porosity, vertical rock stress, maximum horizontal rock stress, permeability, capillary pressure, etc. Based on the various sets of simulated high-resolution property images 150-1 to 150-n, the model simulator 128 may generate a simulated high-resolution model 158.

[0082] For instance, the model simulator 128 may assemble several of simulated high- resolution property images 150 associated with multiple different types of subsurface properties in order to generate the simulated high-resolution model 158 to represent the multiple different types of subsurface properties in a singular object or model. For instance, the simulated high-resolution property images 150 may include property images associated with a porosity, a vertical rock stress, and a maximum horizontal rock stress of a subsurface feature, and the model simulator 128 may generate the simulated high-resolution model 158 to represent the subsurface feature by indicating each of these properties in the simulated high-resolution model 158.

[0083] In another example, the model simulator 128 may assemble several of the simulated high-resolution property images 150 to generate the simulated high-resolution model 158 in a higher dimension than the simulated high-resolution property images 150. For instance, the model simulator 128 may assemble a plurality of 2-dimensional simulated high-resolution property images 150 to form a 3-dimensional (e.g., cube) model as the simulated high-resolution model 158. For example, the model simulator 128 may stack or otherwise assemble consecutive images of the simulated high-resolution property images 150 as adjacent slices or layers of a 3-dimensional model, or in some instances, may interpolate between one or more images. Similarly, in cases where the property images are 1 dimensional matrix representations, the simulated high-resolution property images 150 may be assembled to form a 2-dimensional simulated high-resolution model 158.

[0084] In some cases, the model simulator 128 may operate in this way to both generate a high-dimensional model and to incorporate multiple subsurface properties into the simulated high-resolution model 158. The simulated high-resolution model 158 mayDocket No. IS23.0585-WO-PCT facilitate any of the downstream applications of the simulated high-resolution property images 150 as described herein. In this way, the simulated high-resolution model 158 may provide a similar tool or resource to that of a full-scale, physics based high-resolution modelling of a downhole feature, but may be achieved through the time and resourceefficient machine learning techniques described herein.

[0085] FIG. 4 illustrates an example workflow 400 for generating a geological model 460, according to at least one embodiment of the present disclosure. The geological model 460 may be a subsurface model of a geological feature, area, basin, or landscape, and may span a geological time period 462.

[0086] For example, it may be of interest to understand how a geological landscape has been formed and / or changed over a (e.g., long) period, and the geological model 460 may be generated to characterize this long-term change. For instance, the geological time period 462 may be a time period ranging 100’s of years, millennia, or even millions of years. The geological model 460 may indicate a progression or change of various geological aspects over the geological time period 462. For instance, the geological model 460 may include a static subsurface model as described herein, which may be generated or iterated several times periodically over the geological time period 462, such as every 100 years, every 1000 years, every million years, etc. In this way, the geological model 460 may provide several instances or snapshots of a geological landscape over the course of the geological time period 462 in order that changes or developments in the geological landscape may be observed based on comparing the various instances of the subsurface model.

[0087] In many cases, it may be burdensome, time consuming, and costly to generate the geological model 460 in this way (e.g., iterating many times over the geological time period 462) through physics-based simulations as described herein. For instance, such simulations can often be characterized by significant delays in producing a result and through use of significant computing resources. This cost is magnified when considering the numerous iterations that must be performed in order to generate the geological model 460 to span the geological time period 462.

[0088] In some embodiments, the subsurface super-resolution system 120 may be implemented to generate the geological model 460 in a more timely and efficient manner. For instance, the geological time period 462 may be analyzed (e.g., for the geologicalDocket No. IS23.0585-WO-PCT landscape of interest) and one or more geological events of interest may be identified that are significant, more critical, or otherwise of more interest for characterizing, observing, understanding, etc. For instance, an event of interest may be a subsurface carbon dioxide migration. Events of interest may be identified through user input. Accordingly, one or more primary time intervals 464 may be identified that correspond to these more significant events.

[0089] In some cases, the geological time period 462 may be analyzed to identify (e.g., through user input) one or more events in the geological history that are less significant, less critical, or otherwise of less interest. For instance, such events may be identified through user input. Accordingly, one or more secondary time intervals 466 may be identified that correspond to these less significant events. For instance, in some cases, the secondary time intervals 466 may correspond with any remaining time intervals after the primary time intervals 464 are identified.

[0090] In some embodiments, a combination of modelling through physics-based simulations, and through the generating simulated high-resolution models as described herein may be implemented to generate the geological model 460. For example, for the more significant, primary time intervals 464, a physics-based, fine-grid simulation may be performed to generate a high-resolution model of these events. For instance, as described herein, a high-resolution model as generated from a fine-grid, physics-based simulation may include more information (e.g., more modelled properties, characteristics, etc ), may include more robust information, and / or may be more accurate than the simulated high- resolution models as described herein. For example, the simulated high-resolution models generated by the subsurface super-resolution system as described may be generated for only select properties, which may be limited to static and / or rock properties, and may be less accurate (e.g., albeit only slightly) than a full-scale physics-based simulation on a coarse grid. Accordingly, for the primary time intervals 464, a conventional, physics-based high-resolution model may be implemented to represent these more-significant events in the geological time period 462 for the geological model 460 as accurately and as detailed as possible.

[0091] In some embodiments, for the less significant events of the one or more secondary time intervals 466, a simulated high-resolution model may be generated asDocket No. IS23.0585-WO-PCT described herein, by performing a coarse-grid simulation and upscaling the resolution with the subsurface super-resolution system 120. In this way, the geological model 460 may be generated in a high resolution, and the time and computational burden of doing so may be reduced based on implementing the super-resolution techniques as described herein, notwithstanding the geological time period 462 being substantial. Additionally, an upmost level of detail, precision, and accuracy may be provided for one or more geological events of interest in the geological time period 462 baby performing a full scale, fine grid physicsbased simulation for these time intervals.

[0092] FIG. 5 illustrates a flow diagram for a method 500 or a series of acts for modelling a subsurface feature as described herein, according to at least one embodiment of the present disclosure. While FIG. 5 illustrates acts according to one embodiment, alternative embodiments may add to, omit, reorder, or modify any of the acts of FIG. 5. In some embodiments, the method 500 may be performed by a system, in some embodiments, the method 500 may be implemented as instructions stored on a computer-readable storage medium.

[0093] In some embodiments, the method 500 includes an act 510 of receiving a low- resolution model of the subsurface feature generated from a physics-based forward model of the subsurface feature.

[0094] In some embodiments, the subsurface feature is a hydrocarbon reservoir. In some embodiments, the low-resolution model is a 3-dimensional model.

[0095] In some embodiments, the method includes an act 520 of generating, from the low-resolution model, one or more low-resolution property images representing one or more rock properties of the subsurface feature.

[0096] For example, the rock properties may include one or more of porosity, vertical stress, maximum horizontal stress, or capillary pressure. In some embodiments, the low- resolution model indicates a plurality of rock properties, and the one or more low- resolution property images each including one rock property of the plurality of rock properties. In some embodiments, the low-resolution model indicated one or more additional subsurface properties.

[0097] In some embodiments, the one or more low-resolution property images are 2- dimensional images, and generating the one or more low-resolution property imagesDocket No. IS23.0585-WO-PCT includes slicing cross-sections of the 3-dimensional model. In some embodiments, the one or more low-resolution property images are 3 -dimensional images. In some embodiments, each of the plurality of rock properties appear as similar, channel-like features in the one or more low-resolution property images.

[0098] In some embodiments, the method 500 includes an act 530 of generating one or more simulated high-resolution property images using a subsurface super-resolution machine learning model. For example, the act 530 may include generating one or more simulated high-resolution property images representing the one or more rock properties of the subsurface feature using a subsurface super-resolution machine learning model that is trained to process target images indicating low-resolution property values of a static subsurface property and generate high-resolution output images having high-resolution property values that simulate a fine-grid forward modelling of the static subsurface property.

[0099] In some embodiments, the low-resolution property values correspond to at least a 600 meter by 600 meter cell size, and the high-resolution property values correspond to at most a 400 meter by 400 meter cell size. In some embodiments, the one or more low- resolution property images have a low resolution of at most 50 by 50 pixels, and the one or more simulated high-resolution property images have a high-resolution of at least 100 by 100 pixels.

[0100] In some embodiments, the one or more low-resolution property images indicate a plurality of rock properties of the subsurface feature, each image of the one or more low- resolution property images indicating one rock property of the plurality of rock properties, and generating the one or more simulated high-resolution property images includes normalizing the one or more low-resolution property images of the plurality of rock properties.

[0101] In some embodiments, the method 500 includes an act 540 of providing the one or more simulated high-resolution property images for characterizing the subsurface feature. For example, characterizing the subsurface feature may include generating a change or success map or a risk map for a subsurface geology including the subsurface feature based on the one or more simulated high-resolution property images. In some embodiments, characterizing the subsurface features includes performing a fluid flowDocket No. IS23.0585-WO-PCT simulation of a subsurface reservoir based on the one or more simulated high-resolution property images.

[0102] In some embodiments, providing the one or more simulated high-resolution property images includes generating a simulated high-resolution property model indicating a plurality of rock properties of the one or more rock properties of the subsurface feature based on assembling a plurality of simulated high-resolution property images of the one or more simulated high-resolution property images, and providing the simulated high- resolution property model for characterizing the subsurface feature. For example, providing the one or more simulated high-resolution property images may include generating a simulated 3-dimensional high-resolution property model based on assembling a plurality of simulated high-resolution property images of the one or more simulated high- resolution property images, and providing the simulated 3-dimensional high-resolution property model for characterizing the subsurface feature.

[0103] In some embodiments, the method 500 includes identifying a low-resolution model of a subsurface feature; generating low-resolution training images from the low- resolution model, the low-resolution training images each indicating low-resolution property values for a rock property of the subsurface feature; identifying a high-resolution model of the subsurface feature; generating high-resolution training images from the high- resolution model, the high-resolution training images corresponding to the low-resolution training images and each indicating high-resolution property values for the rock property of the subsurface feature; and training a subsurface super-resolution machine learning model based on the low-resolution training images and the high-resolution training images to process target images indicating low-resolution property values of a static subsurface property and generate simulated high-resolution output images having high-resolution property values that simulate a fine-grid forward modelling of the static subsurface property.

[0104] In some embodiments, the low-resolution model and the high-resolution model each indicate a plurality of rock properties of the subsurface feature, the low-resolution training images and corresponding high-resolution training images indicate the plurality of rock properties of the subsurface feature, each training image of the low-resolution training images and corresponding high-resolution training images indicating one rock property ofDocket No. IS23.0585-WO-PCT the plurality of rock properties, and the subsurface super-resolution machine learning model is trained to process target images indicating low-resolution property values for any rock property of the plurality of rock properties and generate corresponding simulated high-resolution output images for an associated rock property

[0105] In some embodiments, the method 500 includes identifying a geological time period for simulating a subsurface feature; identifying one or more primary time intervals corresponding to one or more geological events of interest in the geological time period; for the one or more primary time intervals, generating a high-resolution model of the subsurface feature, including: performing a fine-grid simulation for the subsurface feature using a physics-based forward model; for one or more secondary time intervals of the geological time period, generating a simulated high-resolution model of the subsurface feature, including performing a coarse-grid simulation for the subsurface feature using the physics-based forward model; generating, from the coarse-grid simulation, a plurality of low-resolution property images for one or more rock properties of the subsurface feature; generating a plurality of simulated high-resolution property images for the one or more rock properties of the subsurface feature using a subsurface super-resolution machine learning model that is trained to process target images indicating low-resolution property values of a static subsurface property and generate simulated high-resolution output images having high-resolution property values that simulate a fine-grid simulation of the static subsurface property with the physics-based forward model; and assembling the plurality of simulated high-resolution property images into the simulated high-resolution model of the subsurface feature; and generating a geological model of the subsurface feature for the geological time period including the high-resolution model and the simulated high- resolution model. For example, the one or more geological events of interest may correspond to a subsurface carbon dioxide migration.

[0106] Turning now to FIG. 6, this figure illustrates certain components that may be included within a computer system 600. One or more computer systems 600 may be used to implement the various devices, components, and systems described herein.

[0107] The computer system 600 includes a processor 601. The processor 601 may be a general-purpose single- or multi-chip microprocessor (e.g., an Advanced RISC (Reduced Instruction Set Computer) Machine (ARM)), a special purpose microprocessor (e g., aDocket No. IS23.0585-WO-PCT digital signal processor (DSP)), a microcontroller, a programmable gate array, etc. The processor 601 may be referred to as a central processing unit (CPU). Although just a single processor 601 is shown in the computer system 600 of FIG. 6, in an alternative configuration, a combination of processors (e.g., an ARM and DSP) could be used.

[0108] The computer system 600 also includes memory 603 in electronic communication with the processor 601. The memory 603 may include computer-readable storage media and can be any available media that can be accessed by a general purpose or special purpose computer system. Computer-readable media that store computerexecutable instructions are non-transitory computer-readable media (device). Computer- readable media that carry computer-executable instructions are transmission media. Thus, by way of example and not limitations, embodiment of the present disclosure can comprise at least two distinctly different kinds of computer-readable media: non-transitory computer-readable media (devices) and transmission media.

[0109] Both non-transitory computer-readable media (devices) and transmission media may be used temporarily to store or carry software instructions in the form of computer readable program code that allows performance of embodiments of the present disclosure. Non-transitory computer-readable media may further be used to persistently or permanently store such software instructions. Examples of non-transitory computer- readable storage media include physical memory (e.g., RAM, ROM, EPROM, EEPROM, etc ), optical disk storage (e.g., CD, DVD, HDDVD, Blu-ray, etc.), storage devices (e g., magnetic disk storage, tape storage, diskette, etc.), flash or other solid-state storage or memory, or any other non-transmission medium which can be used to store program code in the form of computer-executable instructions or data structures and which can be accessed by a general purpose or special purpose computer, whether such program code is stored or in software, hardware, firmware, or combinations thereof.

[0110] Instructions 605 and data 607 may be stored in the memory 603. The instructions 605 may be executable by the processor 601 to implement some or all of the functionality disclosed herein. Executing the instructions 605 may involve the use of the data 607 that is stored in the memory 603. Any of the various examples of modules and components described herein may be implemented, partially or wholly, as instructions 605 stored in memory 603 and executed by the processor 601. Any of the various examples ofDocket No. IS23.0585-WO-PCT data described herein may be among the data 607 that is stored in memory 603 and used during execution of the instructions 605 by the processor 601.

[0111] A computer system 600 may also include one or more communication interfaces 609 for communicating with other electronic devices. The communication interface(s) 609 may be based on wired communication technology, wireless communication technology, or both. Some examples of communication interfaces 609 include a Universal Serial Bus (USB), an Ethernet adapter, a wireless adapter that operates in accordance with an Institute of Electrical and Electronics Engineers (IEEE) 802.11 wireless communication protocol, a Bluetooth® wireless communication adapter, and an infrared (IR) communication port.

[0112] The communication interfaces 609 may connect the computer system 600 to a network. A “network” or “communications network” may generally be defined as one or more data links that enable the transport of electronic data between computer systems and / or modules, engines, or other electronic devices, or combinations thereof. When information is transferred or provided over a communication network or another communications connection (either hardwired, wireless, or a combination of hardwired or wireless) to a computing device, the computing device properly views the connection as a transmission medium. Transmission media can include a communication network and / or data links, carrier waves, wireless signals, and the like, which can be used to carry desired program or template code means or instructions in the form of computer-executable instruction or data structures and which can be accessed by a general purpose or special purpose computer.

[0113] A computer system 600 may also include one or more input devices 611 and one or more output devices 613. Some examples of input devices 611 include a keyboard, mouse, microphone, remote control device, buttonjoystick, trackball, touchpad, and light pen. Some examples of output devices 613 include a speaker and a printer. One specific type of output device that is typically included in a computer system 600 is a display device 615. Display devices 615 used with embodiments disclosed herein may utilize any suitable image projection technology, such as liquid crystal display (LCD), light-emitting diode (LED), gas plasma, electroluminescence, or the like. A display controller 617 may also beDocket No. IS23.0585-WO-PCT provided, for converting data 607 stored in the memory 603 into one or more of text, graphics, or moving images (as appropriate) shown on the display device 615.

[0114] The various components of the computer system 600 may be coupled together by one or more buses, which may include one or more of a power bus, a control signal bus, a status signal bus, a data bus, other similar components, or combinations thereof. For the sake of clarity, the various buses are illustrated in FIG. 6 as a bus system 619.

[0115] The techniques described herein may be implemented in hardware, software, firmware, or any combination thereof, unless specifically described as being implemented in a specific manner. Any features described as modules, components, or the like may also be implemented together in an integrated logic device or separately as discrete but interoperable logic devices. If implemented in software, the techniques may be realized at least in part by a non-transitory processor-readable storage medium comprising instructions that, when executed by at least one processor, perform one or more of the methods described herein. The instructions may be organized into routines, programs, objects, components, data structures, etc., which may perform particular tasks and / or implement particular data types, and which may be combined or distributed as desired in various embodiments.

[0116] Further, upon reaching various computer system components, program code in the form of computer-executable instructions or data structures can be transferred automatically or manually from transmission media to non-transitory computer-readable storage media (or vice versa). For example, computer executable instructions or data structures received over a network or data link can be buffered in memory (e.g., RAM) within a network interface module (NIC), and then eventually transferred to computer system RAM and / or to less volatile non-transitory computer-readable storage media at a computer system. Thus, it should be understood that non-transitory computer-readable storage media can be included in computer system components that also (or even primarily) utilize transmission media.INDUSTRIAL APPLICABILITY

[0117] The following description from [

[0118] —

[0137] includes various embodiments that, where feasible, may be combined in any permutation. For example, theDocket No. IS23.0585-WO-PCT embodiment of

[0118] may be combined with any or all embodiments of the following paragraphs. Embodiments that describe acts of a method may be combined with embodiments that describe, for example, systems and / or devices. Any permutation of the following paragraphs is considered to be hereby disclosed for the purposes of providing “unambiguously derivable support” for any claim amendment based on the following paragraphs. Furthermore, the following paragraphs provide support such that any combination of the following paragraphs would not create an “intermediate generalization.”

[0118] In some embodiments, a method of modelling a subsurface feature, includes receiving a low-resolution model of the subsurface feature generated from a physics-based forward model of the subsurface feature, generating, from the low-resolution model, one or more low-resolution property images representing one or more rock properties of the subsurface feature, generating one or more simulated high-resolution property images representing the one or more rock properties of the subsurface feature using a subsurface super-resolution machine learning model that is trained to process target images indicating low-resolution property values of a static subsurface property and generate high-resolution output images having high-resolution property values that simulate a fine-grid forward modelling of the static subsurface property, and providing the one or more simulated high- resolution property images for characterizing the subsurface feature.

[0119] In some embodiments, the one or more rock properties include one or more of porosity, vertical stress, maximum horizontal stress, or capillary pressure.

[0120] In some embodiments, the subsurface feature is a hydrocarbon reservoir.

[0121] In some embodiments, the low-resolution model indicates a plurality of rock properties, and generating the one or more low-resolution property images includes generating each low-resolution property images to include one rock property of the plurality of rock properties.

[0122] In some embodiments, the low-resolution model indicates one or more additional subsurface properties.

[0123] In some embodiments, the low-resolution model is a 3-dimensional model.Docket No. IS23.0585-WO-PCT

[0124] In some embodiments, the one or more low-resolution property images are 2- dimensional images, and generating the one or more low-resolution property images includes slicing cross-sections of the 3-dimensional model.

[0125] In some embodiments, the one or more low-resolution property images are 3- dimensional images.

[0126] In some embodiments, characterizing the subsurface feature includes generating one or more of a chance-of-success map or a risk map for a subsurface geology including the subsurface feature based on the one or more simulated high-resolution property images.

[0127] In some embodiments, characterizing the subsurface feature includes performing a fluid flow simulation of a subsurface reservoir based on the one or more simulated high-resolution property images.

[0128] In some embodiments, providing the one or more simulated high-resolution property images includes generating a simulated high-resolution property model indicating a plurality of rock properties of the one or more rock properties of the subsurface feature based on assembling a plurality of simulated high-resolution property images of the one or more simulated high-resolution property images, and providing the simulated high- resolution property model for characterizing the subsurface feature.

[0129] In some embodiments, providing the one or more simulated high-resolution property images includes generating a simulated 3-dimensional high-resolution property model based on assembling a plurality of simulated high-resolution property images of the one or more simulated high-resolution property images, and providing the simulated 3- dimensional high-resolution property model for characterizing the subsurface feature.

[0130] In some embodiments, the low-resolution property values correspond to at least a 600 meter by 600 meter cell size, and the high-resolution property values correspond to at most a 400 meter by 400 meter cell size.

[0131] In some embodiments, the one or more low-resolution property images have a low resolution of at most 50 by 50 pixels, and the one or more simulated high-resolution property images have a high-resolution of at least 100 by 100 pixels.

[0132] In some embodiments, the one or more low-resolution property images indicate a plurality of rock properties of the subsurface feature, each image of the one or more low-Docket No. IS23.0585-WO-PCT resolution property images indicating one rock property of the plurality of rock properties, and generating the one or more simulated high-resolution property images includes normalizing the one or more low-resolution property images of the plurality of rock properties.

[0133] In some embodiments, each of the plurality of rock properties appear as similar, channel-like features in the one or more low-resolution property images.

[0134] In some embodiments, a system includes at least one processor, memory in electronic communication with the at least one processor, and instructions stored in the memory, the instructions being executable by the at least one processor to identify a low- resolution model of a subsurface feature, generate low-resolution training images from the low-resolution model, the low-resolution training images each indicating low-resolution property values for a rock property of the subsurface feature, identify a high-resolution model of the subsurface feature, generate high-resolution training images from the high- resolution model, the high-resolution training images corresponding to the low-resolution training images and each indicating high-resolution property values for the rock property of the subsurface feature, and train a subsurface super-resolution machine learning model based on the low-resolution training images and the high-resolution training images to process target images indicating low-resolution property values of a static subsurface property and generate simulated high-resolution output images having high-resolution property values that simulate a fine-grid forward modelling of the static subsurface property.

[0135] In some embodiments, the low-resolution model and the high-resolution model each indicate a plurality of rock properties of the subsurface feature, the low-resolution training images and corresponding high-resolution training images indicate the plurality of rock properties of the subsurface feature, each training images of the low-resolution training images and corresponding high-resolution training images indicating one rock property of the plurality of rock properties, and the subsurface super-resolution machine learning model is trained to process target images indicating low-resolution property values for any rock property of the plurality of rock properties and generate corresponding simulated high-resolution output images for an associated rock property.Docket No. IS23.0585-WO-PCT

[0136] In some embodiments, a computer-readable storage medium includes that, when executed by at least one processor, cause the processor to identify a geological time period for simulating a subsurface feature, identify one or more time intervals of interest corresponding to one or more geological events of interest in the geological time period, for the one or more time intervals of interest, generate a high-resolution model of the subsurface feature, including performing a fine-grid simulation for the subsurface feature using a physics-based forward model, for one or more remaining time intervals of the geological time period, generate a simulated high-resolution model of the subsurface feature, including, performing a coarse-grid simulation for the subsurface feature using the physics-based forward model, generating, from the coarse-grid simulation, a plurality of low-resolution property images for one or more rock properties of the subsurface feature, generating a plurality of simulated high-resolution property images for the one or more rock properties of the subsurface feature using a subsurface super-resolution machine learning model that is trained to process target images indicating low-resolution property values of a static subsurface property and generate simulated high-resolution output images having high-resolution property values that simulate a fine-grid simulation of the static subsurface property with the physics-based forward model, and assemble the plurality of simulated high-resolution property images into the simulated high-resolution model of the subsurface feature, and generate a geological model of the subsurface feature for the geological time period including the high-resolution model and the simulated high- resolution model.

[0137] In some embodiments, the one or more geological events of interest correspond to a subsurface carbon dioxide migration.

[0138] The embodiments of the subsurface super-resolution system have been primarily described with reference to wellbore drilling operations; the subsurface superresolution system described herein may be used in applications other than the drilling of a wellbore. In other embodiments, the subsurface super-resolution system according to the present disclosure may be used outside a wellbore or other downhole environment used for the exploration or production of natural resources. For instance, the subsurface superresolution system of the present disclosure may be used in a borehole used for placement of utility lines. Accordingly, the terms “wellbore,” “borehole” and the like should not beDocket No. IS23.0585-WO-PCT interpreted to limit tools, systems, assemblies, or methods of the present disclosure to any particular industry, field, or environment.

[0139] One or more specific embodiments of the present disclosure are described herein. These described embodiments are examples of the presently disclosed techniques. Additionally, in an effort to provide a concise description of these embodiments, not all features of an actual embodiment may be described in the specification. It should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous embodiment-specific decisions will be made to achieve the developers’ specific goals, such as compliance with system-related and business-related constraints, which may vary from one embodiment to another. Moreover, it should be appreciated that such a development effort might be complex and time consuming, but would nevertheless be a routine undertaking of design, fabrication, and manufacture for those of ordinary skill having the benefit of this disclosure.

[0140] Additionally, it should be understood that references to “one embodiment” or “an embodiment” of the present disclosure are not intended to be interpreted as excluding the existence of additional embodiments that also incorporate the recited features. For example, any element described in relation to an embodiment herein may be combinable with any element of any other embodiment described herein. Numbers, percentages, ratios, or other values stated herein are intended to include that value, and also other values that are “about” or “approximately” the stated value, as would be appreciated by one of ordinary skill in the art encompassed by embodiments of the present disclosure. A stated value should therefore be interpreted broadly enough to encompass values that are at least close enough to the stated value to perform a desired function or achieve a desired result. The stated values include at least the variation to be expected in a suitable manufacturing or production process, and may include values that are within 5%, within 1%, within 0.1%, or within 0.01% of a stated value.

[0141] A person having ordinary skill in the art should realize in view of the present disclosure that equivalent constructions do not depart from the spirit and scope of the present disclosure, and that various changes, substitutions, and alterations may be made to embodiments disclosed herein without departing from the spirit and scope of the present disclosure. Equivalent constructions, including functional “means-plus-function” clausesDocket No. IS23.0585-WO-PCT are intended to cover the structures described herein as performing the recited function, including both structural equivalents that operate in the same manner, and equivalent structures that provide the same function. There is no intention to invoke means-plus- function or other functional claiming for any claim except for those in which the words ‘means for’ appear together with an associated function. Each addition, deletion, and modification to the embodiments that falls within the meaning and scope of the claims is to be embraced by the claims.

[0142] The terms “approximately,” “about,” and “substantially” as used herein represent an amount close to the stated amount that is within standard manufacturing or process tolerances, or which still performs a desired function or achieves a desired result. For example, the terms “approximately,” “about,” and “substantially” may refer to an amount that is within less than 5% of, within less than 1% of, within less than 0.1% of, and within less than 0.01% of a stated amount. Further, it should be understood that any directions or reference frames in the preceding description are merely relative directions or movements. For example, any references to “up” and “down” or “above” or “below” are merely descriptive of the relative position or movement of the related elements. Additionally, as used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items.

[0143] The present disclosure may be embodied in other specific forms without departing from its spirit or characteristics. The described embodiments are to be considered as illustrative and not restrictive. The scope of the disclosure is, therefore, indicated by the appended claims rather than by the foregoing description. Changes that come within the meaning and range of equivalency of the claims are to be embraced within their scope.

Claims

Docket No. IS23.0585-WO-PCTCLAIMSWhat is claimed is:

1. A method of modelling a subsurface feature, comprising: receiving a low-resolution model of the subsurface feature generated from a physics-based forward model of the subsurface feature; generating, from the low-resolution model, one or more low-resolution property images representing one or more rock properties of the subsurface feature; generating one or more simulated high-resolution property images representing the one or more rock properties of the subsurface feature using a subsurface super-resolution machine learning model that is trained to process target images indicating low-resolution property values of a static subsurface property and generate high-resolution output images having high-resolution property values that simulate a fine-grid forward modelling of the static subsurface property; and providing the one or more simulated high-resolution property images for characterizing the subsurface feature.

2. The method of Claim 1, wherein the one or more rock properties include one or more of porosity, vertical stress, maximum horizontal stress, or capillary pressure.

3. The method of Claim 1, wherein the subsurface feature is a hydrocarbon reservoir.

4. The method of Claim 1, wherein the low-resolution model indicates a plurality of rock properties, and generating the one or more low-resolution property images includes generating each low-resolution property images to include one rock property of the plurality of rock properties.

5. The method of Claim 1, wherein the low-resolution model indicates one or more additional subsurface properties.Docket No. IS23.0585-WO-PCT6. The method of Claim 1 , wherein the low-resolution model is a 3-dimensional model.

7. The method of Claim 6, wherein the one or more low-resolution property images are 2-dimensional images, and generating the one or more low-resolution property images includes slicing cross-sections of the 3-dimensional model.

8. The method of Claim 6, wherein the one or more low-resolution property images are 3 -dimensional images.

9. The method of Claim 1, wherein characterizing the subsurface feature includes generating one or more of a chance-of-success map or a risk map for a subsurface geology including the subsurface feature based on the one or more simulated high- resolution property images.

10. The method of Claim 1, wherein characterizing the subsurface feature includes performing a fluid flow simulation of a subsurface reservoir based on the one or more simulated high-resolution property images.

11. The method of Claim 1, wherein providing the one or more simulated high- resolution property images includes generating a simulated high-resolution property model indicating a plurality of rock properties of the one or more rock properties of the subsurface feature based on assembling a plurality of simulated high-resolution property images of the one or more simulated high-resolution property images, and providing the simulated high-resolution property model for characterizing the subsurface feature.

12. The method of Claim 1, wherein providing the one or more simulated high- resolution property images includes generating a simulated 3 -dimensional high- resolution property model based on assembling a plurality of simulated high- resolution property images of the one or more simulated high-resolution propertyDocket No. IS23.0585-WO-PCT images, and providing the simulated 3-dimensional high-resolution property model for characterizing the subsurface feature.

13. The method of Claim 1, wherein the low-resolution property values correspond to at least a 600 meter by 600 meter cell size, and the high-resolution property values correspond to at most a 400 meter by 400 meter cell size.

14. The method of Claim 1, wherein the one or more low-resolution property images have a low resolution of at most 50 by 50 pixels, and the one or more simulated high-resolution property images have a high-resolution of at least 100 by 100 pixels.

15. The method of Claim 1, wherein the one or more low -resolution property images indicate a plurality of rock properties of the subsurface feature, each image of the one or more low-resolution property images indicating one rock property of the plurality of rock properties, and generating the one or more simulated high- resolution property images includes normalizing the one or more low-resolution property images of the plurality of rock properties.

16. The method of Claim 15, wherein each of the plurality of rock properties appear as similar, channel-like features in the one or more low-resolution property images.

17. A system, comprising: at least one processor; memory in electronic communication with the at least one processor; and instructions stored in the memory, the instructions being executable by the at least one processor to: identify a low-resolution model of a subsurface feature; generate low-resolution training images from the low-resolution model, the low-resolution training images each indicating low-resolution property values for a rock property of the subsurface feature;Docket No. IS23.0585-WO-PCT identify a high-resolution model of the subsurface feature; generate high-resolution training images from the high-resolution model, the high-resolution training images corresponding to the low-resolution training images and each indicating high-resolution property values for the rock property of the subsurface feature; and train a subsurface super-resolution machine learning model based on the low-resolution training images and the high-resolution training images to process target images indicating low-resolution property values of a static subsurface property and generate simulated high-resolution output images having high- resolution property values that simulate a fine-grid forward modelling of the static subsurface property.

18. The system of Claim 17, wherein: the low-resolution model and the high-resolution model each indicate a plurality of rock properties of the subsurface feature, the low-resolution training images and corresponding high-resolution training images indicate the plurality of rock properties of the subsurface feature, each training images of the low-resolution training images and corresponding high- resolution training images indicating one rock property of the plurality of rock properties, and the subsurface super-resolution machine learning model is trained to process target images indicating low-resolution property values for any rock property of the plurality of rock properties and generate corresponding simulated high-resolution output images for an associated rock property.

19. A non-transitory computer-readable storage medium including instructions that, when executed by at least one processor, cause the processor to: identify a geological time period for simulating a subsurface feature; identify one or more time intervals of interest corresponding to one or more geological events of interest in the geological time period;Docket No. IS23.0585-WO-PCT for the one or more time intervals of interest, generate a high-resolution model of the subsurface feature, including: performing a fine-grid simulation for the subsurface feature using a physics-based forward model; for one or more remaining time intervals of the geological time period, generate a simulated high-resolution model of the subsurface feature, including: performing a coarse-grid simulation for the subsurface feature using the physics-based forward model; generating, from the coarse-grid simulation, a plurality of low- resolution property images for one or more rock properties of the subsurface feature; generating a plurality of simulated high-resolution property images for the one or more rock properties of the subsurface feature using a subsurface super-resolution machine learning model that is trained to process target images indicating low-resolution property values of a static subsurface property and generate simulated high-resolution output images having high-resolution property values that simulate a fine-grid simulation of the static subsurface property with the physics-based forward model; and assemble the plurality of simulated high-resolution property images into the simulated high-resolution model of the subsurface feature; and generate a geological model of the subsurface feature for the geological time period including the high-resolution model and the simulated high-resolution model.

20. The non-transitory computer-readable storage medium of Claim 19, wherein the one or more geological events of interest correspond to a subsurface carbon dioxide migration.

Images