PREDICTING OIL AND GAS RESERVOIR PRODUCTION

Abstract

A method of predicting an output of oil and gas production in a hydrocarbon reservoir of a current and future producing well using a neural network model, includes receiving a data set comprising a plurality of parameters of the hydrocarbon reservoir at a wellsite. The method also includes using the data set to generate a plurality of simulation curves of the hydrocarbon reservoir, each parameter of the plurality of parameters has a range, and the range is adjustable, and the wellsite includes a wellbore penetrating a subterranean formation to extract reserves from the hydrocarbon reservoir. The method also includes performing a simulation, based on the range of each said parameter of the plurality of parameters, of the hydrocarbon reservoir. The method also includes downloading the plurality of simulation curves into a local server to prepare training data for training the neural network model.

Claims

1. A method of predicting an output of oil and gas production in a hydrocarbon reservoir using a neural network model, comprising: receiving a data set comprising a plurality of parameters of the hydrocarbon reservoir at a wellsite, wherein the wellsite comprises a wellbore penetrating a subterranean formation to extract reserves from the hydrocarbon reservoir; generating, using the data set, a plurality of simulation curves of the hydrocarbon reservoir, wherein each parameter of the plurality of parameters has a range, and the range is adjustable, and performing a simulation, based on the range of each said parameter of the plurality of parameters, of the hydrocarbon reservoir to generate a plurality of simulation curves; downloading the plurality of simulation curves into a local server to prepare training data for training the neural network model; calculating a plurality of key factors from a neighboring well for at least two wellsites from the simulation; combining the plurality of key factors with the plurality of parameters to define input features of the neural network model; and defining each of the plurality of simulation curves as output features of the neural network model.

2. The method of claim 1, further comprising tuning the neural network model using a set of hidden layers between the input features and the output features, wherein the tuning comprises a plurality of tunings.

3. The method of claim 2, further comprising retrieving, for each said tuning of the plurality of tunings, selected data, shuffling and splitting the selected data with K-fold cross validation, and scaling the selected data using a scaler to obtain scaled data.

4. The method of claim 3, further comprising searching for a plurality of hyperparameters by fitting the scaled data in each said tuning of the plurality of tunings.

5. The method of claim 3, further comprising applying an early stop function to prevent the training from overfitting the selected data to the neural network model.

6. The method of claim 2, further comprising processing each said tuning of the plurality of tunings and calculating an average error of the neural network model, and saving the average error as a result.

7. The method of claim 6, further comprising searching for a plurality of hyperparameters by fitting the scaled data in each said tuning of the plurality of tunings, and comparing a result and selecting an optimal set of hyperparameters of the plurality of hyperparameters belonging to the neural network model having a lowest validation error.

8. The method of claim 7, further comprising: further training the neural network model having the lowest validation error with the optimal set of hyperparameters to obtain an optimized neural network model; and uploading the optimized neural network model to a virtual server or a virtual private cloud.

9. The method of claim 1, further comprising: uploading from a client firewall, by an analytical module, actual wellsite data comprising actual wellsite production data, actual wellsite pressure data, and actual wellsite parameter data; and inputting the wellsite parameter data and selecting the range of each said parameter of the plurality of parameters to display the plurality of simulation curves generated from the neural network model in a virtual server or a virtual private cloud.

10. The method of claim 9, further comprising: matching simulation production data and simulation pressure data from the plurality of simulation curves generated from the neural network model in the virtual server or the virtual private cloud with the actual wellsite production data and the actual wellsite pressure data to obtain a plurality of matching simulation curves; displaying an outcome of the plurality of matching simulation curves on a display; and storing the plurality of matching simulation curves and the plurality of parameters.

11. The method of claim 9, further comprising: creating a plurality of hydrocarbon development scenarios in the analytical module user interface for drilling operation in an area of interest; assigning the stored plurality of parameters to the wellsite in a hydrocarbon development scenarios of the plurality of hydrocarbon development scenarios; and displaying the plurality of simulation curves generated from the neural network model in the virtual server or the virtual private cloud using the plurality of hydrocarbon development scenarios.

12. The method of claim 11, further comprising calculating a probability distribution for an outcome of the plurality of simulation curves.

13. The method of claim 12, further comprising: creating a plurality of decline curve models with an outcome of calculated probability distribution; matching an outcome of the plurality of probability simulation curves to the plurality of decline curve models by adjusting a plurality of decline curve parameters; and exporting the adjusted plurality of decline curve models for the current and the future producing wells into a user format for economic analysis.

14. The method of claim 11, further comprising: re-selecting the range of each said parameter of the plurality of parameters and re-adjusting the hydrocarbon development scenarios for adjusting the probability distribution for the current and the future producing wells until achieving an optimal economic result; and using the adjusted probability distribution to select a location to perform a drilling operation to drill another wellbore at the hydrocarbon reservoir.

15. The method of claim 1, wherein the range has a low variable and a high variable.

16. The method of claim 1, further comprising using a simulation module user interface to: adjust each said range of the plurality of parameters for the simulation to obtain an outcome of the base case simulation; display a plurality of hydrocarbon producing wells and the reserve based on the adjusted range of the plurality of parameters and the outcome of the simulation; display an outcome of the simulation in the plurality of simulation curves on a display in the simulation module; and export and store the plurality of simulation curves into a database in a virtual server or a virtual private cloud.

17. The method of claim 1, wherein the plurality of key factors comprises: neighboring well quantities and influence, spacing differences, timing differences, or FDI factors.

18. The method of claim 2, wherein the tuning further comprises using: a number of nodes, activation functions, optimizer functions, learning rates, dropout rates, and regularization.

19. A method of predicting an output of oil and gas production in a hydrocarbon reservoir using a neural network model, comprising: receiving, by a data collection module, a data set comprising a plurality of parameters of the hydrocarbon reservoir at a wellsite, wherein the wellsite comprises a wellbore penetrating a subterranean formation to extract reserves from the hydrocarbon reservoir; generating, using the data set, a plurality of simulation curves of the hydrocarbon reservoir, wherein each parameter of the plurality of parameters has a range, and the range is adjustable; performing a simulation, by a simulation module, based on the range of each said parameter of the plurality of parameters, of the hydrocarbon reservoir to generate a plurality of simulation curves; downloading the plurality of simulation curves into a local server to prepare training data for training the neural network model; calculating a plurality of key factors from a neighboring well for at least two wellsites from the simulation; combining the plurality of key factors with the plurality of parameters to define input features of the neural network model; and defining each of the plurality of simulation curves as output features of the neural network model.

20. A computer device of predicting an output of oil and gas production in a hydrocarbon reservoir using a neural network model, comprising: a non-transitory computer readable medium configured to store computer executable instructions; at least one processor, wherein in response to executing the computer executable instructions, the processor is configured to: receive a data set, using a graphic user interface (GUI), comprising a plurality of parameters of the hydrocarbon reservoir at a wellsite, wherein the wellsite comprises a wellbore penetrating a subterranean formation to extract reserves from the hydrocarbon reservoir; generate, using the data set, a plurality of simulation curves of the hydrocarbon reservoir on the GUI, wherein each parameter of the plurality of parameters has a range, and the range is adjustable using the GUI; perform a simulation, using the GUI, based on the range of each said parameter of the plurality of parameters, of the hydrocarbon reservoir; download the plurality of simulation curves into a database to prepare training data for training the neural network model; calculate a plurality of key factors from a neighboring well for at least two wellsites from the simulation; combine the plurality of key factors with the plurality of parameters to define input features of the neural network model; and define each of the plurality of simulation curves as output features of the neural network model.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0024] For an understanding of embodiments of the disclosure, reference is now made to the following description taken in conjunction with the accompanying drawings, in which:

[0025] FIG. 1 is a diagram of a wellsite for conventional and unconventional oil and gas, represented by a geological image, in accordance with at least one embodiment of the present disclosure.

[0026] FIG. 2 is a diagram of a wellsite for conventional and unconventional oil and gas, in accordance with at least one embodiment of the present disclosure.

[0027] FIG. 3 is a diagram of a geological image of directional drilling, in accordance with at least one embodiment of the present disclosure.

[0028] FIG. 4 is a diagram of a user interface for a subsurface parameters' analysis, in accordance with at least one embodiment of the present disclosure.

[0029] FIG. 5 is a diagram of a user interface for shale assessment/future type wells analysis, in accordance with at least one embodiment of the present disclosure.

[0030] FIG. 6 is a diagram of components of cloud computing, a data processing system, in accordance with at least one embodiment of the present disclosure.

[0031] FIG. 7 is a flowchart of a method for predicting oil and gas reservoir production in current producing wells, in accordance with at least one embodiment of the present disclosure.

[0032] FIG. 8 is a flowchart of a method for predicting oil and gas reservoir production in future producing wells, in accordance with at least one embodiment of the present disclosure.

[0033] FIG. 9 is a flowchart of a method for predicting oil and gas reservoir production, to automate forecasting for current and future producing wells, in accordance with at least one embodiment of the present disclosure.

[0034] FIG. 10 is a flowchart of a method for predicting oil and gas reservoir production using algorithms to teach a deep learning and/or neural network models from pre-run simulation results, in accordance with at least one embodiment of the disclosure.

[0035] FIG. 11 is a diagram of a user interface of performing machine learning for well and/or reservoir analysis, in accordance with at least one embodiment of the present disclosure.

[0036] In the Figures, the same reference numerals are used for components which are identical or similar, even if a repeated description is superfluous for reasons of simplicity.

DETAILED DESCRIPTION

[0037] The following detailed description is merely exemplary in nature and is not intended to limit the disclosure or the application and uses of the disclosure. As used herein, the word “exemplary” means “serving as an example, instance, or illustration.” Thus, any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments. Furthermore, there is no intention to be bound by any expressed or implied theory presented in the preceding technical field, background, brief summary, or the following detailed description.

[0038] For ease of reference, certain terms used in this application and their meanings as used in this context are set forth. To the extent a term used herein is not defined below, it should be given the broadest definition persons in the pertinent art have given that term as reflected in at least one printed publication or issued patent. Further, the present disclosure is not limited by the usage of the terms shown below, as all equivalents, synonyms, new developments, and terms or methods that serve the same or a similar purpose are considered to be within the scope of the present claims.

[0039] In this description, reference is made to the drawings, wherein like parts are designated with like reference numerals throughout. As used in the description herein and throughout, the meaning of “a,” “an,” “the,” and “said” includes plural reference unless the context clearly dictates otherwise. Also, as used in the description herein, the meaning of “in” includes “into” and “on” unless the context clearly dictates otherwise.

[0040] “Analytical software” refers to data analysis software. An example pertinent to the present disclosure includes but is not limited to SpotfireTM. The analytical software includes a parameters window, wherein the user is able to define the ranges of the specified parameters in the parameters window.

[0041] As used in this description, the terms “component,” “database,” “module,” “system,” and the like are intended to broadly capture a computer-related entity, either hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component, in some instances, is a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. An example of a database pertinent to the present disclosure includes but is not limited to a Relational Database System.

[0042] “Decline curve model” refers to employing the graphical procedure of decline curve analysis. An example pertinent to the present disclosure includes but is not limited to Arps decline curve analysis.

[0043] “Decline curve parameters” refer to decline rate, exponential, b factor, Arps, (super) hyperbolic, harmonic, and terminal decline rate. In at least some embodiments, the generated decline curve is exponential. In at least some embodiments, the generated decline curve is hyperbolic. In at least some embodiments, the generated decline curve is harmonic. In at least some embodiments, the generated decline curve includes one or more curve segments, and each curve segment includes unique decline curve parameters. An example of decline curve parameters pertinent to the present disclosure includes but is not limited to Arps parameters.

[0044] “Areas of Interest” refers to a geological area which warrants drilling, based on specific parameter values over which the user is able to control.

[0045] “Outcome” includes a goal or objective of an optimization process. In at least some embodiments, an outcome includes a set of simulation codes and/or algorithms. In at least some embodiments, an outcome includes the errors or uncertainty in predictions of future production, including specific parameter values over which the user is able to control. In at least some embodiments, the outcome determines one or more actions to be applied to the operation of the system, in which the operation is adjusted to perform in a manner that most closely meets the goals or objectives of the user.

[0046] “History matching” refers to the process of adjusting unknown parameters, such as the ones described below, of a reservoir model until the predictions of the model resemble the past production of the reservoir as closely as possible. The more historical data in the base case that is provided for history matching, the more reliable the “simulation curve” of the present disclosure will be, which serves as a basis for history matching error determination and the reliability of future performance predictions. History matching is extremely time consuming and highly dependent on the skill and knowledge of a reservoir engineer.

[0047] “Geological model” is a computer-based representation of a subsurface earth structure, representative of the structure and the behavior thereof. Geological models are used in the optimization and development of a reservoir to determine structural and petrophysical properties of a reservoir.

[0048] Examples of geological model parameters pertinent to the present disclosure include but are not limited to the following: stratigraphic surfaces, flooding surfaces, structural surfaces, boundaries, well data, lithofacies, porosity, permeability, sequence interfaces, fluid contacts, fluid saturation, seismic trace data, subsurface faults, bounding surfaces, and facies variations.

[0049] “Production Data” refers to any values that are able to be measured over the life of the field. Examples include rates of production of oil, gas, and water from individual producing wells, pressure measured vs. depth for specified wells at specified times, pressure at a specified depth measured in a specified well vs. time, seismic response measured at a specified time over a specified area, fluid compositions vs. time in specified wells, flow rate vs. depth for a specified well at specified times.

[0050] “Reserves” refers to the estimated quantities of oil and gas to be produced from the current date to the end of life of the well, which geological and engineering data demonstrate with reasonable certainty to be recoverable in future years from known reservoirs.

[0051] “Reservoir simulation model,” “simulation model,” “simulation curves” and the like refer to a mathematical representation of a hydrocarbon reservoir and the fluids, wells, and facilities associated with the hydrocarbon reservoir. Simulation curves are used to conduct numerical experiments regarding future performance of the hydrocarbon reservoir to determine the most profitable operating strategy. A petroleum engineer managing a hydrocarbon reservoir is able to create many different simulation models to quantify the past performance of the reservoir and predict future performance of the reservoir.

[0052] “Wellsite” refers to a wellbore penetrating a subterranean formation for extracting fluid from an underground reservoir therein.

[0053] In analysis methods according to at least one embodiment of the present disclosure, production forecast models are generated using reservoir simulation software such as Computer Modelling Group™ reservoir simulation software or Petrel Reservoir Engineering Eclipse™ simulation software. In at least some embodiments, different production forecast models are able to be used; such other production forecast models utilize substitution of or modification of some or all of the below listed attributes for the respective production forecast model's specific parameters.

[0054] In analysis methods according to at least some embodiments of the present disclosure, specified parameters, also called attributes are defined. Examples of specified parameters pertinent to the present disclosure include but are not limited to the following: initial reservoir pressure, reservoir depth, bottom-hole flowing pressure, bubble point pressure, dew point pressure, shear stress gradient, pressure gradient, reservoir temperature, reservoir thickness, oil density, gas gravity, rock matrix and natural fracture permeability, non-fracture zone matrix permeability multiplier, vertical and horizontal permeability multipliers, rock matrix/natural fracture porosity, natural fracture spacing, rock matrix/hydraulic fracture initial water saturation, water-oil contact depth, matrix/natural fracture compressibility, well lateral length, cluster spacing, well spacing, number of clusters, hydraulic fracture half-length/height/width/conductivity/permeability, number of fracture stages, hydraulic fracture compaction/relative permeability tables, and Pressure-Volume-Temperature (PVT) tables. The ranges of the specified parameters comprise a low and high variable, varied by source.

[0055] These data are collected from a variety of public or private sources and are used in the generation or prediction of decline curves as described by embodiments herein. Examples of data sources pertinent to the present disclosure include but are not limited to the following: Google®, Drilling Info, IHS Markit™, Society of Petroleum Engineer Publications™, Wolfcamp, Niobrara, Bonespring, Avalon, Lower Spraberry Shale, Jo Mill, Middle Spraberry, Cline, Tuscaloosa, Mancos, Eagle Ford, Bakken, Avalon, Scoop/Stack, Marcellus, Haynesville, Utica, Fayetteville, Barnett, Woodford, and Woodford-Barnett.

[0056] The present disclosure provides a user the ability to generate thousands of simulations from the integration of numerical and neural network models. In these simulations, there are various parameters associated with a single well or plurality of wells. A single well is one that has no adjacent wells. A plurality of wells, in some instances, is called a family, a family having at least one parent well and child well. The various parameters include aforementioned actual wellsite parameters which are able to be selected for a single well or family of wells to determine the estimated ultimate recovery (EUR) of each well. By already having actual wellsite parameters, however, new parameters are created based on relationships between wells. Relationship between wells refers to the spatial distance/position, or well spacing, between at least two wells, well interference, timing and pressure communications.

[0057] The new parameters include “NumberTopWells”, “AvgTopDistance”, “AvgTopTiming”, “FdiTop”, “NumberBottomWells”, “AvgBotDistance”, “AvgBotTiming”, “FdiBottom”, “LeftWellDistance”, “LeftWellTimingDiff”, “LeftWellFdi”, “RightWellDistance”, “RightWellTimingDiff”, and “RightWellFdi”. “NumberTopWells” is expressed in units of well counts and describes the number of wells closest to the top within a family of wells. “AvgTopDistance” is expressed in units of feet and describes the average distance of nearest top wells. “AvgTopTiming” is expressed in units of months and describes the average timing difference of wells nearest the top. “FdiTop” is expressed in units of square feet times hydraulic fracture permeability and describes how top wells affect the EUR.

[0058] “NumberBottomWells” is expressed in units of well counts and describes the number of wells closest to the bottom within a family of wells. “AvgBotDistance” is expressed in units of feet and describes the average distance of nearest bottom wells. “AvgBotTiming” is expressed in units of months and describes average timing difference of wells nearest the bottom. “FdiBottom” is expressed in units of square feet times hydraulic fracture permeability and describes how bottom wells affect the EUR. “LeftWellDistance” is expressed in units of feet and describes the distance between the target well and the left closest well of the target well. “LeftWellTimingDiff” is expressed in units of months and describes the timing difference of the left closest well of the target well. “LeftWellFdi” is expressed in units of square feet times hydraulic fracture permeability and describes how the left closest well affects the EUR.

[0059] “RightWellDistance” is expressed in units of feet and describes the distance between the target well and the right closest well of the target well. “RightWellTimingDiff” is expressed in units of months and describes the timing difference of the right closest well of the target well. “RightWellFdi” is expressed in units of square feet times hydraulic fracture permeability and describes how the right closest well affects the EUR. In addition to the new parameters, the pressure drop per hour (“PDPH”) for a well is able to be calculated.

[0060] Another group of parameters referred to as Neighboring Well Influence (“NWI”) parameters are able to be derived from the parameters referred to above using the following equation:

[00001]$\mathrm{NWI}\left(\mathrm{neighbor}\right)={.Math.}_{n=1}^{\#\mathrm{NumWells}}\left(\frac{\mathrm{Lateral}*\mathrm{Frac}\mathrm{Height}*\mathrm{Xf}*\mathrm{HFPerm}*\mathrm{Pi}*\mathrm{PDPH}}{\mathrm{BHPi}*\mathrm{BHP}\min *\mathrm{Distance}*\mathrm{Time}\mathrm{Difference}}\right)$

Lateral refers to well lateral length in units of feet. Frac Height refers to well fracture height in units of feet. Xf refers to fracture half-length horizontally in units of feet. HFPerm refers to hydraullic fracture permeability in units of millidarcy. Pi refers to initial reservoir pressure in units of pounds per square inch. PDPH refers to pressure drop per hour in units pounds per square inch per hour. BHPi refers to initial bottomhole pressure in units of pounds per square inch. BHPmin refers to minimum bottomhole pressure in units of pounds per square inch. Distance refers to horizontal or vertical spacing distance to neighbor well(s) in units of feet. TimeDifference refers to age differences between primary and infill wells in units of years. “TopNWI” is expressed in units of millidarcy times square feet per square hour and describes how neighboring top wells physically affect the EUR vertically. “BottomNWI” is expressed in units of millidarcy times square feet per square hour and describes how neighboring bottom wells physically affect the EUR vertically. “LeftNWI” is expressed in units of millidarcy times square feet per square hour and describes how neighboring left wells physically affect the EUR horizontally. “RightNWI” is expressed in units of millidarcy times square feet per square hour and describes how neighboring right wells physically affect the EUR horizontally.

[0061] Another group of parameters referred to as Fracture Driven Interactions (“FDI”) parameters represent how a server fracture interference affects the future production for a given well of interest and are measured in overlapped volume percentages. These FDI parameters are able to be calculated based on the following equation:

[00002]$\mathrm{FDI}\mathrm{factor}=\frac{\mathrm{Intersected}\mathrm{Volume}}{\mathrm{Well}\mathrm{of}\mathrm{Interest}\mathrm{Volume}}$

“TopFDI Factor” is expressed in units of percentage and describes the level of FDI's influence from the top wells. Intersected Volume refers to volume of intersected rectangular prism in units of feet cubed. Well of Interest Volume refers to total stimulated rock volume in units of feet cubed. “BotFDI Factor” is expressed in units of percentage and describes the level of FDI's influence from the bottom wells. “RightFDl Factor” is expressed in units of percentage and describes the level of FDI's influence from the right wells. “LeftFDI Factor” is expressed in units of percentage and describes the level of FDI's influence from the left wells.

[0062] The new parameters are utilized as input features in a neural network model, which determines the output, which is a cumulative oil output projection up to a period of 360 months. The cumulative oil output is able to be segmented into cumulative oil outputs for each month starting at month 1 to consecutive months, and up to month 360. In some embodiments, cumulative outputs for secondary phases such as water and natural gas are determined using a neural network model, as well.

[0063] The neural network model is used to build a deep learning model. To build a deep learning model, a computer programming language is used, such as the Python programing language. Keras is a deep learning Application Programming Interface (“API”) written in Python, and runs on top of a machine learning platform. A machine learning platform compatible with Python is, for example, TensorFlow. Using Keras, hypothetical or training parameters, or hyperparameters are tuned to train a sequential model in order to build an optimal model.

[0064] Tuning a parameter refers to training or optimizing a model's performance without overfitting the data. The training parameters are entered in an input layer, the input layer having up to 27 nodes representing up to 20 to 50 input features; an output layer having up to 359 nodes representing up to 359 months of EUR; and hidden layers to find the optimal number(s) of nodes in each of the layers. In some embodiments, other parameters are tuned for training purposes, including optimizer functions, activation functions, learning rates, dropout rates, and regularization.

[0065] After building a sequential model, a dataset is then adapted to fit the sequential model. In adapting the dataset to fit the sequential model, cross-validation is performed for every tuning. K-fold cross-validation refers to evaluating a model(s) using a limited sample to estimate how the model is expected to perform in general when used to make predictions on data not used during the training of the model(s). Different combinations of parameters are adapted to fit the model, and the model is able to be trained multiple times, for example, a model is trained ten times (K=10, the data will be split 10 times into a training data set, validation data set, and test data set) using training data, the model undergoing k-fold cross-validation, then tested for accuracy using test data. 90% of the data being adapted to fit the model is training data. The remaining 10% of the data is actual test data.

[0066] After running the test data through the model, an average validation-loss is determined (MAE or MSE). With training the model ten times, and tuning the parameters 1000 times, for example, a result of 10,000 combinations are used to train the model. An early-stop function is added to prevent overfitting from occurring, while still obtaining a model with the lowest possible average-validation loss. Overfitting refers to a model that models the training data too well, such that the model learns too much detail or noise, ultimately having a negative impact on the model's ability to generalize. An early stop function is a type of regularization which is used to avoid overfitting when training a learning model repetitively.

[0067] After fitting a dataset to a model, a final deep learning model with the lowest possible average-validation loss is generated. The deep learning model is saved and uploaded to a cloud server or virtual machine.

[0068] When a user wants to perform an analysis, a request is sent to the system with the user's defined well parameters, well count, landing targets, vertical spacing and lateral spacing. The measures for vertical spacing and lateral spacing are represented in exact values or in a range of values. Based on what the user provides for the defined well parameters, calculations are performed to generate new parameters to match the inputs of the deep learning model. The model is loaded from the cloud server/virtual machine with the user's defined parameters, and a result is returned. Decline curve analysis is performed on the result, where the curves are drawn using an application, such as Spotfire© referred to above.

[0069] On the user side, the system has a graphical user interface. On the graphical user interface, a user interacts with a homepage. From the homepage, for example, a user downloads simulated cases saved in a database. Each simulated case is a 30 years' time-series of information associated with a well saved in a database. To download the time-series information associated with a well, a user picks what kind of model type for various areas of interest, the model type being “Single” or “Multiple”.

[0070] For a “Single” type model, single-type input parameters are utilized, these parameters including “Formation Name”, “Lateral Length”, “GOR”, “Pi”, “Matrixporo”, “Matrixperm”, “EUR”, and “SWI”.

[0071] For a “Multiple” type model, single-type input parameters are utilized in addition to the following multiple-type input parameters so that a Three-Dimensional model of the user's model is generated. These parameters include “Well Count”, “Landing Target”, “Horizontal Spacing”, “Vertical Spacing”, and “Timing”.

[0072] After building a Three-Dimensional model, a user inputs the range parameters of the model to download all cases of the model. When all the utilized inputs are filled, the SBF software sends an object request to an API stored in the cloud. The API reads through the object request, then finds the matching cases, and returns the matching cases to the software as a data file, such as a json file. In addition to json files, the software reads other data file types. The software converts the json file into the data to be stored in the data table. Using the data file, a History Matching page is selected from the graphical user interface, and on the History Matching page, matching is performed to fit the curves to their actual wells. These matched cases are saved, and a record is exported or used as the parameter range to do prediction analysis for a new model.

[0073] To build forecasting cases from the integrated neural network and simulation models, parameter variables are selected, such as “Landing Target”, “Landing Distance”, “Well Count”, and “Well Spacing”. To build various drilling scenarios, the position of each well is adjustable by moving the well model in up, down, left, or right directions to specific coordinates. The model is shaped by staggering the floors of the wells or adding/deleting a selected well from the model.

[0074] After the new model is designed, input parameters for each well are selected for the model. A user is able to select two options for the input parameters: (i) recorded simulation cases or (ii) type-in input parameters. The following parameters, at least some of which have been referred to above, are selected for a model, and include Lateral Length, Well Spacing, Pb, Pi, Xf, Swi, HFSwi, HFPerm, Fracture Penetration Up, Fracture Penetration Down, Matrixporo, Matrixperm, Perfcluster Spacing, Timing in Months and PDPH (Pressure drop per hour).

[0075] In some embodiments, a new model is built based on an existing model. To build a new model based on an existing model, parameter variables are selected, such as “Landing Target”, “Landing Distance”, “Well Count”, and “Well Spacing”. After selecting the parameter variables, the parameter variables are assigned from each matched case to correspond with each premium well in the model. A range of the parameters is then set from the premium model for new wells that the user wants to add on to the current model. Prediction is then performed, and trigger API functions discussed above to request deep learning models to predict the type curve outcomes. Outcomes are displayed as a family of curves or individual curves.

[0076] FIG. 1 is a diagram of a wellsite 150 for conventional 179 and unconventional 178 oil and gas, represented by a geological image. Drilling rigs 155 are pieces of equipment used to create holes or wellbores 156 in the earth's surface 153. Conventional non-associated gas 159, gas already in the reservoir, does not accumulate with conventional oil 151. Conventional associated gas 152 accumulates in conjunction with the conventional oil 151. The conventional gas accumulations 152, 159 occurs when gas migrates from oil and gas rich shale 157 into sandstone formation 140, which then becomes trapped by an overlying impermeable formation, called a seal 154. Tight sand gas accumulations 158 occur when gas migrates from a source rock into the sandstone formation 140 but is unable to migrate upward due to the permeability in the sandstone. Coalbed methane 141 is generated during the transformation of organic material to coal.

[0077] FIG. 2 is a diagram of another geological image of the wellsite 150 displaying the conventional 179 and the unconventional 178 methods of drilling oil and gas. The surface is an onshore or offshore facility producing conventional or unconventional hydrocarbons from a subterranean reservoir. The drilling rigs 155 are machines on the surface used to drill the wellbores 156. The conventional 179 method is the traditional way of drilling oil and gas, extracted by natural pressure, to access the conventional non-associated gas 159. The unconventional 178 method is drilling down the wellbore 156 horizontally, causing fracking 177, in order to access the oil and gas rich shale 157.

[0078] FIG. 3 is a diagram of a geological image of directional drilling 175. The drilling rigs 155 allow the oil and gas rich shale 157 to be accessed via horizontal drilling techniques from the wellbore 156.

[0079] FIG. 4 is a diagram of a graphic user interface for a subsurface parameters' analysis 180, according to the present disclosure. The subsurface parameters' analysis 180 shows the decline curve analysis that appears to a user on his or her display. A window for selected well information 120 appears in the upper left of the screen. The user has the ability to select different variables, available to the user, such as reservoir properties 123, rock and fluid properties 124, well completion specification data 125, and planar hydraulic fracture specification 126. The graphs appearing to the right of the window for the selected well information 120 includes a graph of an oil rate vs. time simulation 121 and a graph of a cumulative oil production vs. time simulation 122, in accordance with an exemplary embodiment of the present disclosure.

[0080] FIG. 5 is a diagram of a graphic user interface for shale assessment/future type wells analysis 190, according to the present disclosure. In the upper left of the screen appears a window for case specific data 130 that includes the ranges of the specified parameters for the reservoir properties 123 and the well completion specification data 125. The ranges of the specified parameters comprise a low and high variable, varied by source. Available graphs appearing in the upper right of the screen are a graph of an oil cumulative production oil rate simulation 132 and a graph of a gas cumulative production gas rate simulation 133. Graphs appearing in the lower right of the screen are a graph of an oil cumulative production vs. time simulation 134 and a graph of a gas cumulative production vs. time simulation 135. In the lower left of the screen appears a window for a list of available cases with desired reservoir and well characterization 131.

[0081] FIG. 6 is a flowchart of at least one embodiment of the components of cloud computing, a data processing system 160, according to the present disclosure. The data processing system 160 includes one or more computers 168, one or more databases 161, and one or more networks 163. The one or more databases 161 contains a plurality of simulation curves 162. The plurality of simulation curves 162 is matched to actual wellsite data 167 using high performance computing, containing a virtual server 166 and a virtual private cloud 165. The desired outcome is uploaded to the one or more networks 163 and stored in the one or more computers 168. A client firewall 169 contains the actual wellsite data 167 uploaded locally by a user. User input parameters and development scenarios and request neural network model 164 to generate simulated type curves and display on the one or more computers 168. The data processing system 160 has associated therewith the one or more databases 161, the plurality of simulation curves 162, the neural network model 164, the virtual server 166, and the virtual private cloud 165, according to the data processing methodology of FIG. 6.

[0082] FIG. 7, FIG. 8, and FIG. 9 are flowcharts of a block diagram of a method for predicting oil and gas reservoir production. A set of data is collected 200 to generate ranges of specified parameters for one or more oil and gas reservoirs in order to create a base case 201. The base case is created 201 in a simulation software using the set of data collected 200. Simulation is run on the base case 202. The specified parameters for the base case are then adjusted 203. The ranges of the specified parameters and the base case are used to display a plurality of fluid production and reserves in the simulation software 204. The ranges of the specified parameters are adjusted to obtain an outcome 205. The outcome is displayed in a plurality of simulation curves 206. The plurality of simulation curves is exported to a database 207 and stored in the database 208 for future use.

[0083] FIG. 7 is a flowchart of additional steps for predicting oil and gas reservoir production in current producing wells 250. For current producing wells, a set of actual wellsite production data, actual wellsite pressure data is uploaded to an analytical software 209. The actual wellsite parameter data is inputted, and a user selects a plurality of matching simulation curves 210. Simulation production data and simulation pressure data from the plurality of matching simulation curves is matched with the actual wellsite production data and the actual wellsite pressure data 211. The outcome 218 is displayed containing the plurality of matching simulation curves 212. If the user is unsatisfied with the outcome, the user selects the plurality of simulation curves from the outcome 218 and adjusts the plurality of simulation curves using the actual wellsite parameter data 219. Simulation is then run for the plurality of simulation curves to optimize the outcome 220 and the optimized outcome from the plurality of simulation curves is uploaded to the analytical software 221.

[0084] However, if the user is satisfied with the outcome displayed containing the matching simulation curves 212, the user proceeds by selecting a set of production cutoff ranges 213. The outcome is displayed containing the plurality of matching simulation curves 214. A plurality of decline curve models is created containing the outcome 215. The outcome from the plurality of matching simulation curves is matched to the plurality of decline curve models by adjusting a plurality of decline curve parameters 216. The user then exports the plurality of decline curve models results into a user format 217.

[0085] FIG. 8 is a flowchart of additional steps for predicting oil and gas reservoir production in future producing wells 260. In future producing wells, a set of actual wellsite production data for a plurality of wells is uploaded to an analytical software by the user 222. The ranges of the specified parameters are inputted for user defined areas of interest 223. The plurality of simulation curves is displayed within the ranges of the specified parameters previously input 224. Probability distribution of the plurality of simulation curves and the actual wellsite production data for the plurality of wells is calculated 225 and then compared by adjusting the ranges of the specified parameters until the outcome is reached 226. Probabilistic type wells are calculated from the outcome 227 and matched to the plurality of simulation curves 228. The plurality of simulation curves and the probabilistic type wells results are displayed 229. A plurality of decline curve models is then created with the outcome 230. The plurality of simulation curves and the probabilistic type wells are matched to the plurality of decline curve models by adjusting a plurality of decline curve parameters 231. The plurality of decline curve models' results is then exported into a user format 232.

[0086] FIG. 9 is a flowchart of additional steps for predicting oil and gas reservoir production, to automate forecasting for current and future producing wells 270. To automate forecasting for current and future producing wells, a set of actual wellsite production data, actual wellsite pressure data, and the ranges of the specified parameters for actual wellsite parameter data is uploaded to an analytical software for a plurality of wells 233. A set of production cutoff ranges is selected 234. The user then inputs a specified number of a plurality of matching simulation curves for one or more actual wellsites 235. The ranges of the specified parameters from the plurality of simulation curves are matched with the ranges of the specified parameters of the actual wellsite parameter data to generate a plurality of matching simulation curves 236. The simulation production data and simulation pressure data from the plurality of simulation curves is matched with the actual wellsite production data and the actual wellsite pressure data 237.

[0087] Steps 236 and 237 are repeated until the outcome is obtained for the plurality of wells 238. The plurality of matching simulation curves for the plurality of wells is displayed 239. A plurality of decline curve models is displayed with the outcome 240. The plurality of simulation curves is matched to the plurality of decline curve models by adjusting a plurality of decline curve parameters for the plurality of wells 241. The plurality of decline curve models results for the plurality of wells are exported into a user format 242. Matches for the plurality of wells are grouped for user defined areas of interest 243. The ranges of the specified parameters from the matches, the plurality of matching simulation curves, and the plurality of decline curve models are displayed for the user defined areas of interest 244. The ranges of the specified parameters are adjusted to optimize the outcome 245. Probabilistic type wells are calculated from the plurality of matching simulation curves and the plurality of decline curve models 246 and then matched by adjusting the plurality of decline curve parameters for the user defined areas of interest 247. The plurality of decline curve parameters for the plurality of type wells are then exported into a user format 248.

[0088] FIG. 10 is a flow chart of additional steps for predicting oil and gas reservoir production, using pre-run simulations to build a deep learning and/or neural network model(s). As discussed above, the plurality of simulation curves is exported to a database and stored for future use. The stored plurality of simulation curves and up to 30 years' time-series data is downloaded as a “Single” or “Multiple” model type to a systems folder 249. After downloading the model type, the simulated new parameters are selected as input features 250. The input features and up to 30 years' time-series data is entered on one line for model training 251. The model is trained using a multi-regression algorithm 252. The model is trained and optimized when the model has the lowest number of initial average mean square errors for the whole data set 253. The models are trained using different sample percentages, for example, 90% of the data is training data, and 10% of the data is test data 254. The most accurate models are selected 255 and uploaded onto a cloud server or virtual machine 256. The models are configured to scale 257. A user inputs ranges of parameters of the model to download all production and pressure curves cases generated from the model 258. The user receives production and pressure curves from the models 259, and then either selects the production and pressure curves data in an analytical software tool 209 to obtain productivity on a current producing well, or modifies the production and pressure curves data in an analytical software tool 222 to obtain productivity on a future producing well.

[0089] FIG. 11 is a diagram of a user interface 1100 of performing machine learning for well and/or reservoir analysis, in accordance with at least one embodiment of the present disclosure. The Landing Target feature 1101 of the user interface 1100 permits a user to enter a number indicating a quantity of landing targets for the simulation model representative of an actual wellsite arrangement. The landing targets are able to be viewed along the Y-axis of the chart adjacent thereto, which displays an array of wellsites. A user is able to input any number indicating the quantity of landing targets, and in the example depicted in FIG. 11 the quantity of landing targets is 3, and in response to the input of 3 landing targets 3 wellsites are viewable in a row along the Y-axis of said chart.

[0090] The Landing Distance feature 1102 of the user interface 1100 permits a user to enter a number indicating a distance between each wellsite along the Y-axis of said chart. Distance is able to be expressed in a plurality of standard and metric units, and in the embodiment depicted in FIG. 11 each wellsite is placed 50 feet from an adjacent wellsite along the Y-axis respective the plurality of columns on said chart.

[0091] The Well Count feature 1103 of the user interface 1100 permits a user to a enter a number indicating a total count of wellsites within the model at a given location, and the given location is representative of an actual location. In response to a user entering the landing target 1101 and the well count 1103, the system automatically fits the corresponding number of rows to the columns of evenly distributed wellsites along the X-axis.

[0092] In the example depicted in FIG. 11, in response to a user inputting 3 landing targets 1101 and 15 wells in the well count 1103, the system automatically fits the corresponding number of wells in each row, 5 wells in each row. The Well Spacing feature 1104 of the user interface 1100 permits a user to a enter a number indicating a distance between each wellsite along the X-axis. In this embodiment, a user has entered the number 1,000 indicating the distance between each wellsite along the X-axis. Similarly to the landing distance 1102 measurements, the measurements of the well spacing 1104 is able to be expressed in standard or metric units, and in this example, the wells are spaced apart from each other by a distance of 1,000 feet along X-axis. Using the graphic icons 1105, 1106, and 1107, a user is able to adjust, create, or delete any highlighted well location by assigning X and Y coordinated accordingly.

[0093] The Wellsite feature 1109 is a graphical icon of a wellsite representative of a currently producing well or a future producing well. In this example, a user is able to view 15 individual wellsites. A user is able to hover over or click on each wellsite, and in response to hovering over or clicking on the particular wellsite, a user is able to view information unique to the properties of the particular reservoir and wellsite penetrating into the reservoir in the Reservoir Parameters feature/table 1117 and Well Parameters feature/table 1118, respectively. A variety of information relating to a reservoir and a wellsite are available for viewing under these tables/features, the specific information discussed above in this disclosure.

[0094] The adjustment tool feature 119 permits a user to adjust the spacing of a wellsite. A user is able to input a distance and manipulate the wellsite horizontally along the X-axis or vertically along the Y-axis to adjust the selected wellsite to be further from or closer to another an adjacent wellsite. For each particular case input in the Selected feature 1111 a user is able to view a plurality of parameter features 1112. For each parameter of the plurality of parameters, a user is able to input a minimum range in a Min. feature 1113 and a maximum range in a Max. feature 1114. A user is also able to search and select a saved case using graphical search feature 1108. After inputting the minimum and the maximum for a range for each parameter of the plurality of parameters, a user is then able to save the Saved Case 1111, and access at a later time.

[0095] After setting the ranges for the plurality of parameters a user is then able to visualize an Area Model type 1116, such as a “Delaware Basin” model. Other types of area models include Midland Basin, Willistion Basin, Powder River Basin, etc. . . After the area model feature/type is selected, then the number of models to be generated is entered in the Number of Model feature 1115. In this example, 100 models will be generated based on the input, and each one of these models is able to be visualized as a unique simulation curve upon selection of the Prediction feature 1120. Each simulation curve 133 and 134 of FIG. 5 is able to be individually selected by a user, the respective reservoir and wellsite parameters are also able to be viewed upon selection of the individually selected simulation curve.

[0096] A method of predicting an output of oil and gas production in a hydrocarbon reservoir of a current and future producing well using a neural network model, includes receiving a data set comprising a plurality of parameters of the hydrocarbon reservoir at a wellsite. The method also includes using the data set to generate a plurality of simulation curves of the hydrocarbon reservoir, each parameter of the plurality of parameters has a range, and the range is adjustable, and the wellsite includes a wellbore penetrating a subterranean formation to extract reserves from the hydrocarbon reservoir. The method also includes performing a simulation, based on the range of each said parameter of the plurality of parameters, of the hydrocarbon reservoir. The method also includes downloading the plurality of simulation curves into a local server to prepare training data for training the neural network model.

[0097] The method also includes calculating a plurality of key factors from a neighboring well for at least two wellsites from the simulation. The method also includes combining the plurality of key factors with the plurality of simulated parameters to define as input features of the neural network model. The method also includes defining the plurality of simulation curves as output features of the neural network model. The method also includes tuning the neural network model using a set of hidden layers between the input features and the output features, wherein the tuning comprises a plurality of tunings.

[0098] The method also includes retrieving, for each said tuning, selected data, shuffling and splitting the selected data with K-fold cross validation, and scaling the selected data using a scaler to obtain scaled data. The method also includes searching for a plurality of hyperparameters by fitting the scaled data in each said tuning. The method also includes applying an early stop function to prevent the training from overfitting the selected data to the neural network model. The method also includes processing each said tuning and calculating an average error of the neural network model, and saving the average error as a result.

[0099] The method also includes searching for a plurality of hyperparameters by fitting the scaled data in each said tuning, and comparing the result and selecting an optimal set of hyperparameters of the plurality of hyperparameters belonging to the neural network model having a lowest validation error. The method also includes further training the neural network model having the lowest validation error with the optimal set of hyperparameters to obtain an optimized neural network model, and upload the optimized neural network model to a virtual server or a virtual private cloud. The method also includes uploading from a client firewall, by an analytical module, actual wellsite data comprising actual wellsite production data, actual wellsite pressure data, and actual wellsite parameter data. The method also includes using an analytical module user interface to input the wellsite parameter data and select the range of each said parameter of the plurality of parameters to display the plurality of simulation curves generated from the neural network model in a virtual server or a virtual private cloud.

[0100] The method also includes matching simulation production data and simulation pressure data from the plurality of simulation curves generated from the neural network model in the virtual server or the virtual private cloud with the actual wellsite production data and the actual wellsite pressure data to obtain a plurality of matching simulation curves. The method also includes displaying an outcome of the plurality of matching simulation curves on a display in the analytical module. The method also includes storing the plurality of matching simulation curves and the plurality of parameters in the analytical module user interface. The method also includes creating a plurality of hydrocarbon development scenarios in the analytical module user interface for drilling operation in an area of interest.

[0101] The method also includes assigning the stored plurality of parameters to the wellsite in a hydrocarbon development scenarios of the plurality of hydrocarbon development scenarios. The method also includes displaying the plurality of simulation curves generated from the neural network model in the virtual server or the virtual private cloud using the plurality of hydrocarbon development scenarios. The method also includes calculating a probability distribution for an outcome of the plurality of simulation curves. The method also includes creating a plurality of decline curve models with an outcome of calculated probability distribution.

[0102] The method also includes matching the outcome of the plurality of probability simulation curves to the plurality of decline curve models by adjusting a plurality of decline curve parameters. The method also includes exporting the adjusted plurality of decline curve models for the current and the future producing wells into a user format for economic analysis. The method also includes re-selecting the range of each said parameter of the plurality of parameters and re-adjusting the hydrocarbon development scenarios for adjusting the probability distribution for the current and the future producing wells until achieving an optimal economic result. The method also includes using the adjusted probability distribution to perform a drilling operation to drill another wellbore at the hydrocarbon reservoir.

[0103] The range has a low variable and a high variable. The method also includes using a simulation module user interface to adjust each said range of the plurality of parameters for the simulation to obtain an outcome of the base case simulation. The method also includes using a simulation module user interface to display a plurality of hydrocarbon production and the reserve based on the adjusted range of the plurality of parameters and the outcome of the simulation. The method also includes using a simulation module user interface to display the outcome of the simulation in the plurality of simulation curves on a display in the simulation module.

[0104] The method also includes using a simulation module user interface to export and store the plurality of simulation curves into a database in a virtual server or a virtual private cloud. The plurality of key factors includes neighboring well quantities and influence, spacing differences, timing differences, and FDI factors. The tuning includes using a number of nodes, activation functions, optimizer functions, learning rates, dropout rates, and regularization.

[0105] A method of predicting an output of oil and gas production in a hydrocarbon reservoir of a current and future producing well using a neural network model including receiving, by a data collection module, a data set comprising a plurality of parameters of the hydrocarbon reservoir at a wellsite. The method also including using the data set to generate a plurality of simulation curves of the hydrocarbon reservoir, and each parameter of the plurality of parameters has a range, and the range is adjustable, and the wellsite including a wellbore penetrating a subterranean formation to extract reserves from the hydrocarbon reservoir. The method also including performing a simulation, by a simulation module, based on the range of each said parameter of the plurality of parameters, of the hydrocarbon reservoir. The method also including downloading the plurality of simulation curves into a local server to prepare training data for training the neural network model.

[0106] The method also including calculating a plurality of key factors from a neighboring well for at least two wellsites from the simulation. The method also including combining the plurality of key factors with the plurality of parameters to define as input features of the neural network model. The method also including defining the plurality of simulation curves as output features of the neural network model.

[0107] A method of predicting an output of oil and gas production in a hydrocarbon reservoir of a current and future producing well using a neural network model, including receiving, by a data collection module, a data set comprising a plurality of parameters of the hydrocarbon reservoir at a wellsite. The method also including using the data set to generate a plurality of simulation curves of the hydrocarbon reservoir, and each parameter of the plurality of parameters has a range, and the range is adjustable, and the wellsite including a wellbore penetrating a subterranean formation to extract reserves from the hydrocarbon reservoir. The method also including performing a simulation, by a simulation module, based on the range of each said parameter of the plurality of parameters, of the hydrocarbon reservoir. The method also including downloading the plurality of simulation curves into a local server to prepare training data for training the neural network model.

[0108] A computer device of predicting an output of oil and gas production in a hydrocarbon reservoir of a current and future producing well using a neural network model, including a non-transitory computer readable medium configured to store computer executable instructions. The device also includes at least one processor, wherein in response to executing the computer executable instructions, the processor is configured to receive a data set, using a graphic user interface (GUI), comprising a plurality of parameters of the hydrocarbon reservoir at a wellsite. The processor is also configured to use the data set to generate a plurality of simulation curves of the hydrocarbon reservoir on the GUI, and each parameter of the plurality of parameters has a range, and the range is adjustable using the GUI. The wellsite comprises a wellbore penetrating a subterranean formation to extract reserves from the hydrocarbon reservoir.

[0109] The processor is also configured to perform a simulation, using the GUI, based on the range of each said parameter of the plurality of parameters, of the hydrocarbon reservoir. The processor is also configured to download the plurality of simulation curves into a database to prepare training data for training the neural network model. The processor is also configured to calculate a plurality of key factors from a neighboring well for at least two wellsites from the simulation. The processor is also configured to combine the plurality of key factors with the plurality of parameters to define as input features of the neural network model. The processor is also configured to define the plurality of simulation curves as output features of the neural network model.

[0110] The foregoing description of some embodiments of the disclosure has been presented for purposes of illustration and description. It is not intended to be exhaustive or to limit the disclosure to the precise form disclosed, and modifications and variations are possible in light of the above teachings. The specifically described embodiments explain the principles and practical applications to enable one ordinarily skilled in the art to utilize various embodiments and with various modifications as are suited to the particular use contemplated. It should be understood that various changes, substitutions, and alterations can be made hereto without departing from the spirit and scope of the disclosure.