WO2014039036A1 - Procédés d'optimisation de réalisation de puits de forage et de fracturation hydraulique et systèmes associés - Google Patents

Procédés d'optimisation de réalisation de puits de forage et de fracturation hydraulique et systèmes associés Download PDF

Info

Publication number
WO2014039036A1
WO2014039036A1 PCT/US2012/053795 US2012053795W WO2014039036A1 WO 2014039036 A1 WO2014039036 A1 WO 2014039036A1 US 2012053795 W US2012053795 W US 2012053795W WO 2014039036 A1 WO2014039036 A1 WO 2014039036A1
Authority
WO
WIPO (PCT)
Prior art keywords
wellbore
data
parameters
instances
data regarding
Prior art date
Application number
PCT/US2012/053795
Other languages
English (en)
Inventor
Robert SHELLEY
Nijat GULIYEV
Amir NEJAD
Original Assignee
Stratagen, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Stratagen, Inc. filed Critical Stratagen, Inc.
Priority to PCT/US2012/053795 priority Critical patent/WO2014039036A1/fr
Publication of WO2014039036A1 publication Critical patent/WO2014039036A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • E—FIXED CONSTRUCTIONS
    • E21—EARTH OR ROCK DRILLING; MINING
    • E21B—EARTH OR ROCK DRILLING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR A SLURRY OF MINERALS FROM WELLS
    • E21B43/00—Methods or apparatus for obtaining oil, gas, water, soluble or meltable materials or a slurry of minerals from wells
    • E—FIXED CONSTRUCTIONS
    • E21—EARTH OR ROCK DRILLING; MINING
    • E21B—EARTH OR ROCK DRILLING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR A SLURRY OF MINERALS FROM WELLS
    • E21B43/00—Methods or apparatus for obtaining oil, gas, water, soluble or meltable materials or a slurry of minerals from wells
    • E21B43/25—Methods for stimulating production
    • E21B43/26—Methods for stimulating production by forming crevices or fractures
    • E—FIXED CONSTRUCTIONS
    • E21—EARTH OR ROCK DRILLING; MINING
    • E21B—EARTH OR ROCK DRILLING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR A SLURRY OF MINERALS FROM WELLS
    • E21B47/00—Survey of boreholes or wells

Definitions

  • the present disclosure relates to methods, devices, and systems associated with wellbores, including wellbores utilized to extract natural gas, oil, other hydrocarbons, and/or other natural resources.
  • the methods, devices, and systems of the present disclosure are utilized to optimize wellbore completions.
  • Hydraulic fracturing technology has been developed to boost production from unconventional reservoirs by creating high permeability channels in the reservoir rock. This greatly improves the ability of fluids to flow toward the well and subsequently be transferred to the surface.
  • the design parameters necessary for successful hydraulic fracturing are greatly affected by reservoir quality, completion scheme, wellbore properties, and other factors, many of which are not fully understood. This leads to only a few successful parameter combinations among numerous possible combinations.
  • FIG. 1 is a diagrammatic, partial cross-sectional side view of a field having a vertical well and a horizontal well in accordance with aspects of the present disclosure.
  • FIG. 2 is a diagrammatic top view of a field having a plurality of completed and uncompleted wells in accordance with aspects of the present disclosure.
  • FIG. 3 is a diagrammatic flow chart of a method of optimizing well completion according to an embodiment of the present disclosure.
  • FIG. 4 is a diagram illustrating aspects of a data gathering step of the method illustrated in the flow chart of Fig. 3 according to an embodiment of the present disclosure.
  • FIG. 5 is a diagram illustrating aspects of a data screening step of the method illustrated in the flow chart of Fig. 3 according to an embodiment of the present disclosure.
  • FIG. 6 is a diagram illustrating aspects of a data processing step of the method illustrated in the flow chart of Fig. 3 according to an embodiment of the present disclosure.
  • FIG. 7 is a collection of scatter plots associated with a bivariate analysis according to an embodiment of the present disclosure.
  • FIG. 8 is a scatter plot matrix according to an embodiment of the present disclosure.
  • FIG. 9 is a dendrogram according to an embodiment of the present disclosure.
  • FIG. 10 is a self-organizing map according to an embodiment of the present disclosure.
  • FIG. 11 is a diagram illustrating aspects of a data modeling step of the method illustrated in the flow chart of Fig. 3 according to an embodiment of the present disclosure.
  • FIG. 12 is a diagram illustrating aspects of a model selection step of the method illustrated in the flow chart of Fig. 3 according to an embodiment of the present disclosure.
  • FIG. 13 is a diagram illustrating aspects of a model application step of the method illustrated in the flow chart of Fig. 3 according to an embodiment of the present disclosure.
  • FIG. 14 is a graph illustrating a manner of evaluating a formation using hydrocarbon ratios according to an embodiment of the present disclosure.
  • FIG. 15 is a graph illustrating estimated well production based on a predictive model according to an embodiment of the present disclosure.
  • FIG. 16 is a chart illustrating correlation values of various well completion and reservoir parameters to well production according to an embodiment of the present disclosure.
  • FIG. 17 is a graph illustrating well production values for a well relative to a plurality of well completion options for the well according to an embodiment of the present disclosure.
  • FIG. 18 is a graph illustrating economic value of a well relative to a plurality of well completion options for the well according to an embodiment of the present disclosure.
  • FIG. 19 is a graph illustrating well production values for a well relative to a plurality of well completion options for the well according to another embodiment of the present disclosure.
  • FIG. 20 is a graph illustrating economic value of a well relative to a plurality of well completion options for the well according to another embodiment of the present disclosure.
  • FIG. 21 is a graph illustrating actual oil recovery from a well relative to a predicted oil recovery based on a predictive model according to an embodiment of the present disclosure.
  • a field 100 having a vertical well 102 and a horizontal well 104 shown therein is a field 100 having a vertical well 102 and a horizontal well 104.
  • the field 100 is representative of any type of field where natural resources are obtained.
  • the field 100 is an oil field, natural gas field, and/or other hydrocarbon field. Due to the current state of drilling, the following description will primarily focus on the optimization of horizontal well completions. However, it is understood that the concepts described herein are equally applicable to vertical wells. To that end, the field and reservoir parameters associated with horizontal wells have identical and/or analogous parameters that are associated with vertical wells. Likewise, the various well completion parameters discussed herein for horizontal wells also have identical and/or analogous parameters for vertical wells.
  • a map of a field 110 shown therein is a map of a field 110 according to an embodiment of the present disclosure.
  • the map 110 shows the location of a plurality of completed wells 112 (each marked with an "x", some numeric labels have been omitted for clarity) and the location of uncompleted wells 114 (each marked with an "o").
  • data regarding the completed wells 112 is utilized to generate one or more predictive models. At least one of the predictive models is subsequently utilized to estimate well production for the uncompleted wells 114.
  • a predictive model identified as the "best" predictive model for the field is utilized to estimate well production for a plurality of different completion parameters in order to identify a combination of completion parameters that optimizes well production.
  • evaluation of the predictive models to identify the "best" predictive model for a particular field can take numerous factors into consideration.
  • more than one "best” predictive model may be identified for a particular field.
  • the results from each of the "best” predictive models may be taken into consideration in identifying the optimized well completion parameters.
  • the larger the number of completed wells 112 utilized to generate the predictive model the more accurate the model will be.
  • the methods and systems of the present disclosure are suitable for use with large or small data sets (e.g. , ten or fewer wells for a field).
  • Fig. 3 shown therein is a flowchart of an exemplary method 120 of optimizing wellbore completion according to an embodiment of the present disclosure.
  • the method 120 will be described with respect to various steps or stages below. This is simply to provide clarity to the concepts of the method 120 and does not require a particular order or separation of the steps/stages. Rather, it is understood that aspects of the various steps and/or stages described below with respect to any of the embodiments of the present disclosure may be combined into a single step/stage from multiple steps/stages, separated into multiple steps/stages from a single step/stage, reordered, and/or combinations thereof.
  • any combination of components from any one component, to any two components, to any three components, to all of the components
  • any combination of components from any one component, to any two components, to any three components, . . . , to all of the components
  • the methods and systems described herein aim to optimize hydraulic fracture and wellbore completion designs through the use of data typically obtained during well drilling and/or operation (e.g. , mud logs) and artificial intelligence (AI) without the need for well-specific wireline or open hole logging data.
  • the process includes multiple stages including data gathering, screening, processing, modeling, and forecasting.
  • the data used is comprised of data such as drilling, completion, fracturing, seismic, and production data such that separate logging is not necessary.
  • mud logs are a major component of some implementations of the process because they are often readily available and carry reservoir quality properties that can be correlated to completion and fracture designs.
  • ANN Artificial Neural Networks
  • GA Genetic Algorithms
  • ANN Artificial Neural Networks
  • GA Genetic Algorithms
  • Forecasting is done with trained and cross-validated neural network models.
  • Further detailed analysis e.g. , sensitivity analysis
  • the final outcome of the methods and systems described herein enable the optimization of fracture designs for a well with respect to the cost of completing the well and the resulting hydrocarbon production.
  • the present disclosure provides techniques that lead to both cost savings and increased hydrocarbon production.
  • neural networks are the most commonly used artificial intelligence method. Neural networks have numerous advantages such as the ability to find nonlinear patterns, handle missing data, and handle large number of unknowns (e.g. , outputs). But to be most effective neural networks require a sufficient amount of data to train the networks (i.e. , learn and understand the complexity of the given datasets). This requires rigorous attempts of data gathering, screening, and pre-processing. Accordingly, as shown in Fig. 3, the method begins at step 122 with data gathering. Data related to the wells of a particular field/reservoir, both completed and uncompleted, are gathered from all the readily available sources.
  • the initial data gathering step 122 is intended to capture all data that may be useful in evaluating the completed wells that can be used to predict or estimate the impact of various completion design parameters for a particular uncompleted well.
  • the data can come from a wide variety of information sources, such as drilling, geology, completion, stimulation, and production.
  • the well completion optimization is primarily and/or entirely based on mud log data. Mud logs are typically available and contain rock properties suitable for use in predictive modeling in accordance with embodiments of the present disclosure.
  • Data gathering is an important part of the method 120 as many of the subsequent steps and analysis depend upon the data gathered in step 122. To this end, data is extracted from all available sources in some instances. Referring now to Fig. 4, examples of the various types of data that are commonly utilized will be described, however, no limitation is intended thereby. Rather, it is understood that the methods and systems of the present disclosure can utilize essentially any type of information related to a field/reservoir and/or wells thereof that can be quantified in some manner. Accordingly, one of ordinary skill in the art will recognize that extension of the methods and systems of the present disclosure to types of data not explicitly described within the present disclosure is still within the scope of the present disclosure.
  • the data may come in various types of file formats, including databases, spreadsheets, .pdf files, text files, ASCII files (e.g. , LAS files designed for well logs), xml files, SGY files (e.g. , special ASCII files designed for seismic data), and/or combinations thereof.
  • file formats include both common file formats and proprietary file formats.
  • data obtained from any type of format may be utilized within the methods and systems of the present disclosure.
  • Those of ordinary skill in the art will recognize that some file conversion and/or other processes are implemented in some instances to allow for the proper processing of the data from the various file formats within the context of the present disclosure. Accordingly, the details of such conversions and processing will not be described in detail herein.
  • the data gathering step 122 includes gathering or obtaining geological information 134, such as estimated fracture geometry using micro-seismic and tiltmeter measurements, seismic data, and/or other geological information.
  • geological information 134 such as estimated fracture geometry using micro-seismic and tiltmeter measurements, seismic data, and/or other geological information.
  • saturation, structure, and/or other rock properties are correlated to the geological information 134.
  • the geological information 134 is derived from a plurality of data sources having various file formats.
  • the geological information 134 is in SGY format (i. e. , an ASCII file format designed for seismic data).
  • the geological information 134 is in other file formats.
  • the data gathering step 122 includes gathering or obtaining petrophysics information 136, such as gamma ray, caliper, resistivity, neutron, density, sonic, magnetic resonance imaging and detailed lithology logs, and/or other petrophysics information.
  • the petrophysics information 136 is obtained from open hole wireline logs. In that regard, the logs are typically measured relative to a depth of the well.
  • reservoir, rock, and/or fluid properties are correlated to the petrophysics information 136.
  • the petrophysics information 136 is derived from a plurality of data sources having various file formats.
  • the petrophysics information 136 is in ASCII (LAS files designed for well logs). In other instances, the petrophysics information 136 is in other file formats.
  • the data gathering step 122 includes gathering or obtaining drilling information 138, such as rate of penetration, average mud weight, total gas, bottomhole temperature, vertical depth, azimuth and well inclination (for slanted and horizontal wells), lateral orientation (for horizontal wells), and/or other information regarding drilling.
  • drilling information 138 is obtained from mud logs, which may include data such as total gas, penetration rates, chemo- stratigraphy, gas chromatograph measurements of alkanes having one or more carbon atoms versus depth (i.e.
  • mud log data is typically obtained during drilling of a well and, therefore, is commonly available. Accordingly, in some implementations of the present disclosure mud lug data is the primary source of information utilized for deriving predictive models and optimizing well completion. In particular, in some instances, the gas ratios (e.g.
  • Ci, C 2 , C3, C 4 , and C5 ratios including ratios of C1/C2 (methane/ethane), C1/C3 (methane/propane), Ci/C 4 (methane/butane) and C1/C5 (methane/pentane) are utilized as a primary factor in both deriving the predictive models and evaluating the predictive models to identify a "best" predictive model.
  • Some exemplary gas ratio evaluations will be discussed below in the context of Fig. 14.
  • Data from drilling information 138, including mud logs, is used for selection of an optimized pay zone in some instances, which can, in turn, be used to select appropriate completion design parameters in accordance with the present disclosure.
  • the drilling information 138 is derived from a plurality of data sources having various file formats.
  • the drilling information 138 is in Microsoft Excel spreadsheet format, ASCII format (LAS files designed for well logs), .pdf format, .hlg format, .mlg format, .hl6 format, and/or other standard or customized formats.
  • the drilling information 138 is in other file formats.
  • the data gathering step 122 includes gathering or obtaining completion information 140, such as perforation clusters, phasing, perforation interval depth and thickness, cluster per fracture stages, tubing size, completion time/date, and/or other information regarding completion.
  • completion information 140 such as perforation clusters, phasing, perforation interval depth and thickness, cluster per fracture stages, tubing size, completion time/date, and/or other information regarding completion.
  • hydrocarbon production is correlated to the completion information 140.
  • the completion information is derived from a plurality of data sources having various file formats.
  • the completion information 140 is in Microsoft Excel spreadsheet format and/or .pdf format. In other instances, the completion information 140 is in other file formats.
  • the data gathering step 122 includes gathering or obtaining stimulation information 142, such as treatment volumes and rates, fluid types, proppant size and mass, acid volume, fracture stages, initial and final instantaneous shut-in pressure (ISIP), breakdown pressure, pressure trend, closure pressure, conductivity, fracture gradient, and/or other information regarding stimulation.
  • stimulation information 142 such as treatment volumes and rates, fluid types, proppant size and mass, acid volume, fracture stages, initial and final instantaneous shut-in pressure (ISIP), breakdown pressure, pressure trend, closure pressure, conductivity, fracture gradient, and/or other information regarding stimulation.
  • the stimulation design parameters from completed wells are defined and described by the stimulation information 142.
  • the stimulation information 142 is utilized in a predictive model to estimate the effects of various stimulation options (e.g. , hydraulic fracturing parameters) on well production.
  • well production is correlated to stimulation parameters from the stimulation information in some instances.
  • the stimulation information 142 is in Microsoft Excel spreadsheet format and/or .pdf format. In other instances, the stimulation information 142 is
  • the data gathering step 122 includes gathering or obtaining well, field, and/or reservoir information 144, such as lease and API names, surface coordinates, Kelly bushing elevation, surface elevation, different types of maps (geology map, isocore thickness, isopach, etc.), target depth, pay zone thickness, pay top depth, porosity, permeability, saturation, pore pressure, total organic carbon (TOC) measurements, and/or other information regarding the well, field, and/or reservoir.
  • rock and/or fluid properties can be correlated to the well, field, and/or reservoir information.
  • the well, field, and/or reservoir information is derived from a plurality of data sources having various file formats.
  • the well, field, and/or reservoir information is in Microsoft Excel spreadsheet format. In other instances, the well, field, and/or reservoir information is in other file formats.
  • the data gathering step 122 includes gathering or obtaining production information 146, such as oil, water, and gas production rates, cumulative productions, water cut, production decline parameters, estimated ultimate recovery, and/or other information regarding production.
  • production rates are a result of the response of the reservoir to the drilling, completion, and stimulation designs.
  • production information 146 facilitates the evaluation of the effectiveness of various parameters associated with well completion, including any hydraulic fracturing design parameters.
  • one or more of the outputs of the predictive models of the present disclosure include estimated production values, such as total production, peak production, average production, and/or other production parameters over a set amount of time (e.g. , per day, per week, per month, per quarter, per year, or otherwise).
  • the production information 146 is derived from a plurality of data sources having various file formats.
  • the production information 146 is in Microsoft Excel spreadsheet format. In other instances, the production information 146 is in other file formats.
  • the data is filtered, reformatted, and/or pre-processed to be in useable form for subsequent steps of the method 120.
  • many data types such as well logs and digital mud logs, provide high frequency detailed measurements that cannot be easily used in the data modeling directly. Therefore, in some instances representative values of such data types are found for the entire fractured region. In some implementations, this is done by analyzing the wellbore directional survey, extracting perforation schemes and locating the hydraulic fracture intervals. Once the fracture intervals are identified, average representative values of well logs and digital mud logs are calculated.
  • Production data is modified or calibrated to a common scale and/or reference point in some instances to accommodate for differences in production period and/or measurement noise, which are common issues with the raw production data.
  • cumulative data is preferred over the production rates because the magnitude of noise is reduced.
  • a constant time window is selected to handle differences in days on production, which can be dependent on the maturity of the field (e.g. , a six-month interval is utilized in some instances for more mature wells/fields, while a thirty- day -interval is utilized in some instances for newer fields).
  • the outcome of the data gathering process is a dataset that includes a list of wells with their respective attributes (data types). Often the dataset includes data from multiple wells with more than 50 attributes for each well. There is no minimum number of wells or attributes, but the larger the number of wells and attributes the more accurate the resulting predictive model(s) will be.
  • the method 120 continues at step 124 with data screening.
  • the data screening step 124 serves to manipulate the available data to make the best use of the available data.
  • Fig. 5 examples of the types of data screening utilized in the context of the present disclosure will be described, however, no limitation is intended thereby. Rather, it is understood that the methods and systems of the present disclosure can utilize essentially any type of data screening and/or filtering, especially those suitable for addressing incomplete and/or inaccurate datasets. Accordingly, one of ordinary skill in the art will recognize that extension of the methods and systems of the present disclosure to types of data screening and/or filtering not explicitly described within the present disclosure is still within the scope of the present disclosure.
  • the data gathered in step 122 is typically raw data that needs to be examined and manipulated for consistency before proceeding to the data modeling stage.
  • the dataset often includes both bad data and missing data that must be addressed.
  • the data screening 124 includes a step 148 for addressing missing data.
  • wells with missing attributes can be deleted entirely. This option is best suited for datasets with a large number of wells such that the deletion of a single well does not significantly affect the overall dataset.
  • a well having many missing data entries compared to other wells in the dataset is particularly suitable for deletion or removal.
  • a second option is to set to the missing data field to a constant value.
  • a missing data entry can be set to a constant value.
  • the specific value can be selected based on the type of data entry that is missing. In some instances, the same value is used for all wells in the dataset missing that type of data entry. However, in other instances the value of a specific missing data entry may be estimated by making calculations based on other known and present data entries.
  • a third option is to set the missing data field to the average value for the dataset. In this regard, by setting the missing data field to the average value for the dataset, the average value(s) for the dataset are not skewed by a few or a large number of wells having missing data entries. However, such an approach changes the descriptive statistics of the attribute (e.g. , mean and variance). In some instances, the average value for the dataset is used for all wells in the dataset missing that type of data entry.
  • data screening 124 also includes data visualization 150 and/or outlier detection 152, in which various attributes are plotted against each other to find patterns, relationships, and/or outliers.
  • data visualization can often provide a qualitative understanding of the behavior of the data to a user.
  • data outliers can greatly and adversely affect the learning process in the subsequent data modeling stage and may confuse a neural network attempting to adapt to such outliers. Therefore, in some instances outliers are removed from the dataset.
  • the data visualization 150 facilitates identification of data outliers, which may be tagged for removal and/or modification by a user manually and/or automatically by the computer system in response to instructions stored in a non-transitory computer readable medium executed by the computer system.
  • the well with the outlier data can be deleted from the dataset entirely, the well with the outlier data can be ignored with respect to the outlier data field, the outlier data can be set to a constant value, or the outlier data can be set to the average value for the dataset.
  • Data visualization 150 can be done in many different ways, including through the use of various types of charts, graphs, and/or other visual representations.
  • the types of data visualization utilized are particularly suited to identifying correlations between parameters and/or outliers.
  • similar data types across multiple wells are compared and/or plotted together. For example, as shown in Fig. 7, comparing various completion data parameters may identify possible outliers.
  • Fig. 7 shows four scatter plots 170, 172, 174, and 176 that show various completion parameters for wells of a field relative to the production of those wells.
  • scatter plot 170 plots production output relative to the number of fracturing treatments
  • scatter plot 172 plots production output relative to average proppant conductivity
  • scatter plot 174 plots production output relative to wellbore location (measured in degrees from North)
  • scatter plot 176 plots production output relative to treatment volume.
  • software suites such as Microsoft Excel and TIBCO Spotfire are implemented to create such scatter plots.
  • a scatterplot matrix is utilized.
  • scatterplot matrices are particularly suited for visually representing multidimensional data.
  • a scatterplot matrix is the ordered display of p(p— 1) scatterplots.
  • An inherent advantage of a scatterplot matrix over simple bivariate analysis is the ability to visualize large amounts of data across multiple plots and scan to find relationships and outliers.
  • Fig. 8 illustrates an exemplary scatterplot matrix 180 having three variables, namely width, height, and length.
  • the scatterplot matrix 180 includes six different scatter plots 182, 184, 186, 188, 190, and 192.
  • scatter plot 182 charts height relative to width
  • scatter plot 184 charts length relative to width
  • scatter plot 186 charts width relative to height
  • scatter plot 188 charts length relative to height
  • scatter plot 190 charts width relative to length
  • scatter plot 192 charts height relative to length.
  • a user and/or a computer system using image processing programs) can identify outliers and/or potential parameter correlations that can be taken into consideration.
  • data screening 124 includes Analysis of Variance (ANOVA) 154 in addition to or in lieu of data visualization to identify outliers and/or parameter correlations.
  • ANOVA utilizes a statistical test of the mean and covariance of the data to reveal data points far from the mean (i.e. , outliers).
  • squared robust Mahalanobis distances of the data are used in some instances to calculate the distances within the data.
  • each data type is compared individually because feeding all data types in one step is not feasible. For example, completion parameters across multiple wells are compared together using the ANOVA technique.
  • categorical comparisons of data across wells are utilized in some instances.
  • the statistical analyses (including ANOVA) of the present disclosure are carried out using the open source R-programming language.
  • step 126 the method 120 continues to step 126 where the data is further processed.
  • the data screening step 124 results in a dataset where missing data and outliers have been addressed, the resulting dataset is difficult to analyze because of high dimensionality (i. e. , large number of data parameters).
  • high dimensionality i. e. , large number of data parameters.
  • the so-called curse of dimensionality problem arises from the analysis of high dimensional data and inability to describe and extract meaningful information from such data. Therefore, in some instances the number of attributes is reduced and the most important attributes are extracted for further analysis and modeling.
  • Fig. 6 shows examples of the types of data processing utilized in the context of the present disclosure to identify the most important data parameters, however, no limitation is intended thereby.
  • the methods and systems of the present disclosure can utilize essentially any type of data processing and/or filtering, especially those suitable for identifying correlations between data parameters. Accordingly, one of ordinary skill in the art will recognize that extension of the methods and systems of the present disclosure to types of data processing and/or filtering not explicitly described within the present disclosure is within the scope of the present disclosure.
  • the data processing step 126 can use one or more of bivariate analysis 156, multi-variate analysis 158, principal component analysis (PCA) 160, factor analysis (FA) 162, cluster analysis 164, self-organizing maps (SOM) 166, and/or fuzzy logic 168.
  • principal component analysis 160 and factor analysis 162 are generally understood to be types of multi-variate analysis 158
  • self-organizing maps 166 and fuzzy logic 168 are understood to be types of cluster analysis 164.
  • multivariate analysis e.g. , principal component analysis, factor analysis, etc.
  • cluster analysis e.g. , self-organizing maps, fuzzy logic, support vector machine, and dendrograms
  • combinations thereof are utilized.
  • a combination of multivariate analysis and cluster analysis is utilized.
  • Multivariate analysis is a step of data processing in which multivariate statistics of the data are studied in order to extract relationships involving more than two data parameters.
  • Principal component analysis 160 is a technique for reducing the number of variables in the dataset. Principal component analysis 160 seeks to find linear relationships within the data and, by identifying and selecting only those variables with distinct correlations, the number of variables will be reduced. To this end, principal component analysis 160 uses orthogonal transformation of the data to convert variables to the linearly uncorrected variables. Different data types are analyzed using principal component analysis 160 and, with the help of bi-plots (i.e. , visualization plot of the principal components) in some instances, variables with very high correlations are identified.
  • bi-plots i.e. , visualization plot of the principal components
  • a high correlation is considered, in some instances, to be a correlation between 0.75 and 1.0 and/or between -0.75 and -1.0.
  • factor analysis 162 is used for reducing the number of variables.
  • Factor analysis 162 is a statistical method used to describe the variability of the data with a potentially lower number of unobserved variables called factors. Then, correlations of each factor with respect to all variables are identified. Ranking variables with the highest correlation coefficients (e.g. , ⁇ 0.5 cut off is utilized in some instances, while in other instances a ⁇ 0.75 cut off is used) reveals the cluster of variables. Further, using parallel analysis (study of eigenvalues of the dataset) sheds light on the minimum number of factors necessary to describe the variability of the data. In some instances, principal component analysis and factor analysis are both necessary for reduction of the variables in the dataset.
  • engineering judgment which may be manually input by a user and/or programmed into the system, and process knowledge are utilized in some instances to further reduce the number of variables.
  • engineering judgment and/or process knowledge are utilized in some instances to eliminate variables that may show a high correlation in the dataset, but are known to have little actual impact on production and/or include variables that may show a relatively low correlation in the dataset, but are known to have impact on production.
  • particular variables can tagged to always be included, to always be excluded, and/or to only be included when correlation to one or more other parameters meets a requisite threshold.
  • cluster analysis 164 is used for reducing the number of variables.
  • multivariate analysis techniques are a great collection of tools to examine the data, they are generally weak in identifying nonlinear patterns and relationships. Therefore, more sophisticated non-linear pattern recognitions are desired.
  • cluster analysis 164 can be utilized to identify these more sophisticated non-linear patterns.
  • One example of such a cluster analysis 164 is a visualization method called a dendrogram, which is a tree diagram representing hierarchical clustering of the data.
  • Fig. 9 shows an example of a dendrogram 200 for an exemplary dataset according to an embodiment of the present disclosure.
  • the results of the cluster analysis for the dataset are illustrated by the dendrogram 200. More specifically, the dendrogram 200 lists the variables and indicates at what level of similarity any two clusters were joined.
  • the relative stem length of a tree in a dendrogram represents the similarity of variables and the smaller the stem length the higher the similarity.
  • a first type of relationship includes variables with high similarity from the same class. This type of high- correlation relationship can be used for data reduction. In this regard, since the variables have such a high degree of similarity, it is possible to omit some of them from the dataset and rely on the remaining variables. For example, the median of a variable and the average of a variable often have a very short stem such that one of them can be omitted, depending on the analysis preferences.
  • a second type of relationship includes variables with high similarity from different classes. This type of relationship is particularly useful for knowledge discovery. In this regard, variables from different classes (e.g.
  • a third type of relationship includes variables with very low similarity.
  • very low similarity e.g. 0.2 or lower
  • all of the variables are taken into consideration and those having a low similarity are removed from the data set.
  • cluster analysis methods may also be used for knowledge discovery and pattern recognition.
  • self-organizing maps 166 which are unsupervised neural networks used to produce a low-dimensional map of the data space, may be used.
  • Fig. 10 shows an exemplary self-organizing map 210.
  • the unsupervised learning nature of the self-organizing maps helps to discover hidden patterns in the data.
  • Self-organizing maps clustering enforces neighborhood relationships on the resulting cluster centroids. Thus, clusters that are neighbors are more related to one another than clusters that are not neighbors. This greatly facilitates interpretation of hidden relationships.
  • self-organizing maps are used to visualize relationships within the input data set.
  • the production data is excluded from at least some of the self-organizing maps such that the relationships among the input data can be visualized.
  • the self-organizing mapes are used as a classification tool for the input data.
  • fuzzy clustering or logic 168 and/or support vector machines are used. Similar to self-organizing maps, the fuzzy logic 168 and support vector machines seek to extract relationships in the data. Depending on the size of the dataset a combination of clustering algorithms can be used to both reduce the size of the data set and extract information. In some instances, clustering wells based on their respective physical locations and production data can reveal possible variations in reservoir quality and/or factors that have an acute effect on production.
  • step 128 data modeling is performed, which leads to model selection at step 130 and then model application at step 132.
  • neural networks are trained and tested thousands of times with the help of genetic algorithms to generate a plurality of predictive models.
  • final model is selection is done at step 130 based on two main criteria: performance of each predictive model, and engineering validation derived from the process knowledge.
  • the final or "best" predictive model which may also be referred to as a "neural network” or “network,” can be utilized to understand the behavior of fractures and the possible improvements of their design as applied to a particular well at step 132.
  • the model application at step 132 includes knowledge discovery in which neural network sensitivity analysis reveals hidden correlations within the data.
  • FIG. 11 provides examples of the types of data modeling utilized in the context of the present disclosure, however, no limitation is intended thereby. Rather, it is understood that the methods and systems of the present disclosure can utilize other types of data modeling, especially those suitable for using correlations between data parameters of a known dataset to predict future outcomes associated with a partially-known dataset. Accordingly, one of ordinary skill in the art will recognize that extension of the methods and systems of the present disclosure to types of data modeling not explicitly described within the present disclosure is within the scope of the present disclosure.
  • the objective of the data modeling step 128 is to build a predictive tool based on the processed dataset obtained from the previous steps 122, 124, and 126.
  • a predictive model provides a tool to establish a relationship between the various parameters of the data and help understand the behavior of the data.
  • artificial neural networks 220 are used as a main modeling tool.
  • Neural networks are nonlinear pattern recognition tools. Neural networks require training to be able to understand relationship between inputs and outputs of a known dataset. In this regard, in some instances the inputs and outputs are selected and/or defined based on one or more of the available data for an uncompleted well, the available data for completed wells in a field, desired output parameters, and/or combinations thereof.
  • the inputs and desired outputs are fed to the neural network and, through the application of a learning algorithm, prediction errors are minimized to the extent possible for each neural network.
  • hundreds, thousands, tens of thousands, and/or more neural networks which may also be referred to as predictive models, are generated based on the obtained, screened, and processed dataset.
  • each neural network can be tested and/or evaluated utilizing known inputs and outputs that were omitted from the training step.
  • data associated with a particular subset of wells in the available dataset are withheld as a testing group. Accordingly, in some instances the dataset is separated into a first set of data for training and a second, separate set of data for testing.
  • between about 10% and about 50% of the number of wells in a dataset is held back for the test group, with some implementations holding back 30% for the test group.
  • the closer a particular neural network is at predicting the actual outputs from the known inputs for the testing data the better the neural network fits the dataset.
  • Part of the data modeling step 128 is determining what combination of inputs should be used to construct the neural network.
  • the input combinations are determined in two parts.
  • an initial combination of data parameters is selected in such a way as to represent geology, wellbore condition, completion, stimulation, and fluid properties.
  • thousands of neural networks are trained with different combinations of inputs.
  • the outputs of neural networks are production data (e.g. , oil, gas, and/or water production).
  • the training goal is a network error of 10 ⁇ 5 or less.
  • the error function for this purpose is average absolute error. Each network is tested on the test dataset that is not presented to networks during the training stage.
  • Fig. 15 provides a graph 260 that shows estimated well production based on a trained predictive model relative to the actual well production.
  • a commercially available neural network software program such as NeuroGenetic OptimizerTM ("NGO") developed by BioComp Systems, Inc. and/or other suitable software program, is utilized for performing aspects of method 120, including all or portions of data modeling step 128 and/or model selection step 130.
  • NGO NeuroGenetic OptimizerTM
  • the data modeling step 128 also includes application of one or more boosted regression trees 224.
  • boosted regression trees generate predictive models similar to neural networks.
  • neural networks are generally more flexible and have improved pattern recognition compared to boosted regression trees. Therefore, in some implementations the boosted regression tree 224 is used to define single output models (i. e. , models having one target output, such as total hydrocarbon production) and/or cross-validate the results of the models generated from the neural networks 220.
  • a model selection step 130 shown therein are aspects of a model selection step 130 according to embodiments of the present disclosure, however, no limitation is intended thereby. Rather, it is understood that the methods and systems of the present disclosure can utilize generally any type of model selection to select a best or final model from a plurality of available models. Accordingly, one of ordinary skill in the art will recognize that extension of the methods and systems of the present disclosure to types of model selection not explicitly described within the present disclosure is within the scope of the present disclosure.
  • the available neural networks are compared to a set of desired performance criteria.
  • the mean squared errors of the networks are compared.
  • the error with respect to particular parameters may be weighted more heavily than other parameters in the evaluation. For example, in some instances the parameters that are more directly correlated to the output(s) of the predictive model are weighted more heavily.
  • the neural networks with the least error are identified and passed to the next step of network selection. Those neural networks not meeting the performance criteria are removed from contention.
  • engineering judgment or validation 228 is also utilized as an initial step of network selection.
  • neural networks that rely on relationships or correlations that are contrary to well-established engineering principles and/or otherwise do not make logical sense from an engineering standpoint are removed from consideration. For example, in some instances fracture gradient, average total gas, formation depth and their relationships with production data can be used for screening the predictive models.
  • the information obtained in the data processing step 126 can be very useful for selecting a final neural network, which may be referred to as using process knowledge 230 (e.g. , knowledge obtained from the data screening/analysis processes).
  • process knowledge 230 e.g. , knowledge obtained from the data screening/analysis processes.
  • data from mud logs used for formation evaluation can be used to narrow down the neural networks.
  • hydrocarbon ratios of the drilling mud gas can be used to predict whether a zone will produce oil, gas and/or water.
  • FIG. 14 shows a graph 250 with exemplary plots of hydrocarbon ratios, specifically based on Ci/C 2 (methane/ethane), C1/C3 (methane/propane), C1/C4 (methane/butane) and C1/C5 (methane/pentane) ratios, for non-productive and productive wells.
  • This type of mud log data is typically available for most of the well. However, such data has not been used to optimize the design of the completion, including hydraulic fracturing parameters, of wells.
  • these hydrocarbon ratios can be utilized as a proxy for wellbore data parameters that are often unavailable, such as permeability, to facilitate completion optimization.
  • neural networks that take into account these hydrocarbon ratios are preferred and, therefore, rated higher than those neural networks that do not utilize hydrocarbon ratios.
  • the neural networks that do not utilize hydrocarbon ratios are removed from contention as the best or final neural network, either manually by a user or automatically by the computer system in response to instructions stored in a non-transitory computer readable medium executed by the computer system.
  • selection of a final neural network also includes conducting sensitivity analysis 232 on neural network predictions on the inputs. This type of analysis may reveal unrealistic, over-sensitive relationships between inputs and outputs due to the training of the neural network to match the provided dataset.
  • sensitivity analysis is performed by adding and/or subtracting a percentage (e.g. , 5-20% in some implementations and 10% in some implementations) of the range of each input while keeping the rest of the inputs unchanged. Then, the neural network is tested with each dataset and average values of outputs are calculated. By calculating the percentage of change in outputs, sensitivities are identified. To this end, an exemplary chart 270 showing the relative sensitivities of various parameters to production parameters is provided in Fig. 16.
  • the chart 270 also indicates the relative sensitivities of parameters that are controllable (e.g. , well completion and hydraulic fracturing design parameters) and those that are non- controllable, reservoir defined properties (e.g. , gas production and mud weight).
  • controllable e.g. , well completion and hydraulic fracturing design parameters
  • non- controllable, reservoir defined properties e.g. , gas production and mud weight
  • Fig. 13 provides examples of how the selected model is applied in the context of the present disclosure, however, no limitation is intended thereby. Rather, it is understood that the methods and systems of the present disclosure can apply the selected model to a wide array of uses for evaluating existing wells, prospective wells, and/or completed wells. Accordingly, one of ordinary skill in the art will recognize that extension of the methods and systems of the present disclosure to other uses of the predictive models not explicitly described within the present disclosure is within the scope of the present disclosure.
  • the neural network can be utilized to optimize the wellbore fracturing and completion parameters 234.
  • the optimized wellbore completion parameters identified using the selected predictive model are output in human-intelligible form such that those parameters can be implemented for completion of the well.
  • human-intelligible formats include visual displays (e.g. , including two-dimensional and/or three-dimensional images/video provided on a display and/or printed on paper or other medium), audible sounds (e.g. , analog and/or digital recordings stored in a readable format on a storage medium and/or played over speaker(s)), texture (e.g. , braille), and/or combinations thereof.
  • the final neural network can also be utilized for economic evaluations 236 associated with well completion, well drilling, well operation, and/or other aspects related to existing, completed, or prospective wells.
  • the predictive models of the present disclosure can be utilized to evaluate and optimize new wells 238 (e.g. , well placement, depth, profile, and/or completion), evaluate and optimize under-performing existing wells 240 (e.g. , determine whether improved completion design parameters can improve well production in a profitable manner), evaluate and optimize prospective well locations 242 (e.g. , desired well placement, depth, profile, and/or completion), evaluate completion design parameters 244, including hydraulic fracturing designs, for completed wells (e.g. , evaluate utilized parameters to resulting production for wells having similar profiles), and knowledge discovery 246 (e.g. , identifying hidden and/or unexpected relationships between well/field/reservoir properties, completion designs, and/or other factors with well production).
  • Fig. 17 shows a graph 280 that plots the estimated production and recovery predictions made by the predictive model using the reduced treatment volumes for various numbers of hydraulic fracturing stages.
  • the incremental best month oil cumulative production gain for each additional hydraulic fracturing treatment becomes insignificant.
  • the predictive model estimated that with 30 hydraulic fracturing treatments, a best month oil production of 27,213 BBL and estimated ultimate recovery ("EUR") of 854,550 BBL, while 40 hydraulic fracturing treatments would only obtain a slight increase to a best month oil production of 27,798 BBL and EUR of 860,967 BBL. Comparison of this behavior to the numeric reservoir modeling indicates that this production behavior is consistent with 0.05 micro-darcy range permeability reservoir rock.
  • FIG. 18 includes a graph 290 that illustrates the net present value calculated using this approach (using a net oil price of $60/BBL and a water disposal cost of $5/BBL) relative to the number of fracturing stages.
  • the economic break point for well "A” occurs at about 30 hydraulic fracturing treatments. This information was used to justify completion of well "A” using 30 hydraulic fracturing treatments, a treatment volume of approximately 3,350,000 gal placing 3,050,000 lb. of proppant.
  • the actual best month cumulative production achieved by well "A” was approximately 36,000 BBL oil, which is one of the best wells in the area.
  • Fig. 19 shows a graph 300 that plots the estimated production and recovery predictions made by the predictive model using different numbers of hydraulic fracturing treatments. As can be seen, the optimum number of treatments appears to be 45 or greater as there is no break over in production with more hydraulic fracturing treatments.
  • Fig. 20 includes a graph 310 that illustrates the net present value calculated using this approach (using a net oil price of $60/BBL and a water disposal cost of $5/BBL) relative to the number of fracturing stages. As can be seen there is no economic break over for up to 45 hydraulic fracturing treatments on this well. Unfortunately, it was too late to change the completion on well "B" using the data from the predictive model.
  • Well “B” was completed using 24 hydraulic fracturing treatments, a total proppant amount of 2,990,000 lb. consisting of 40/70 sand, 20/40 sand, and 20/40 ceramic.
  • the proppant was placed with approximately 1,420,000 gal of cross-linked fluid.
  • the best month cumulative production achieved by this well was approximately 9,595 BBL of oil, which is in close agreement with the estimate of the predictive model, which estimated 9,188 BBL of oil for 24 hydraulic fracturing treatments.
  • a graph 320 illustrating actual oil recovery from a well relative to a predicted oil recovery based on a predictive model according to an embodiment of the present disclosure.
  • a predictive model in accordance with the present disclosure was used to predict production for 10 wells that were part of a project to evaluate the effectiveness of various completion and hydraulic fracturing methods.
  • Graph 320 compares the data-driven model predicted production relative to the actual best calendar month oil production for these wells, which are located in McKenzie County, North Dakota (Truax area) and Divide/Williams Counties, North Dakota (Wildrose area). The line drawn represents a perfect correlation between the model and actual production.
  • a method of optimizing wellbore completion in accordance with the present disclosure comprises: gathering data regarding a plurality of completed wellbores; gathering data regarding an uncompleted wellbore; utilizing the gathered data regarding the plurality of completed wellbores in the field to define a plurality of predictive models, each of the plurality of predictive models providing an estimate of wellbore production based on the gathered data regarding the plurality of completed wellbores; identifying a best predictive model from the plurality of predictive models; and applying the gathered data regarding the uncompleted wellbore and a plurality of available wellbore completion parameters to the best predictive model to identify a set of wellbore completion parameters from the plurality of available wellbore completion parameters that optimizes estimated wellbore production based on the best predictive model; and completing the uncompleted wellbore based on the set of wellbore completion parameters identified using the best predictive model.
  • the uncompleted wellbore and the plurality of completed wellbores are from a single, common field. In other instances, the uncompleted wellbore and the plurality of completed wellbores are from a plurality of fields. For example, in some implementations at least one of the plurality of completed wellbores is from a first field and the uncompleted wellbore is from a second field different than the first field.
  • the gathered data regarding the plurality of completed wellbores and/or the gathered data regarding the uncompleted wellbore comprises mud log data. In some embodiments, utilizing the gathered data regarding the plurality of completed wellbores to define the plurality of predictive models comprises applying the gathered data regarding the plurality of completed wellbores to an artificial neural network.
  • applying the gathered data regarding the plurality of completed wellbores to the artificial neural network includes applying the data gathered for a first set of completed wellbores from the plurality of completed wellbores to the artificial neural network, wherein the first set of the plurality of completed wellbores is less than all of the plurality of completed wellbores.
  • identifying the best predictive model from the plurality of predictive models comprises applying the data gathered for a second set of completed wellbores from the plurality of completed wellbores to the plurality of predictive models, the second set of completed wellbores being different than the first set of completed wellbores.
  • identifying the best predictive model from the plurality of predictive models further comprises evaluating each of the plurality of predictive models relative to one or more evaluation factors.
  • the one or more evaluation factors may include factors selected from the group of factors consisting of number of data input parameters and whether the predictive model utilizes hydrocarbon ratios. In some instances, predictive models with a lower number of data input parameters and/or that utilize at least one hydrocarbon ratio are favored. In some instances, the one or more evaluation factors further includes parameter sensitivity.
  • the set of wellbore completion parameters that optimizes estimated wellbore production includes a total number of hydraulic fractures in some embodiments.
  • the set of wellbore completion parameters that optimizes estimated wellbore production further includes hydraulic fracture spacings along the uncompleted well. The hydraulic fracture spacings may be consistent or vary along the uncompleted well.
  • the set of wellbore completion parameters that optimizes estimated wellbore production further includes hydraulic fracture volumes, hydraulic fracture fluid type, hydraulic fracture proppant type, hydraulic fracture staging type, and/or other parameters regarding hydraulic fracture completion of the well.
  • utilizing the gathered data regarding the plurality of completed wellbores in the field to define the plurality of predictive models includes identifying correlating factors.
  • multivariate analysis and/or cluster analysis may be used to identify correlating factors.
  • the identification of correlating factors includes generating one or more visual representations of the gathered data selected from the group of visual representations consisting of a scatter plot, a bi-plot, a dendrogram, and a self-organizing map.
  • a computer-implemented system for optimizing wellbore completion includes non-transitory, computer readable medium having a plurality of instructions stored thereon for executing the following steps: receiving data regarding a plurality of completed wellbores in a field; receiving data regarding an uncompleted wellbore in the field; utilizing the gathered data regarding the plurality of completed wellbores in the field to define a plurality of predictive models, each of the plurality of predictive models providing an estimate of wellbore production based on the gathered data regarding the plurality of completed wellbores; identifying a best predictive model from the plurality of predictive models; applying the gathered data regarding the uncompleted wellbore and a plurality of available wellbore completion parameters to the best predictive model to identify a set of wellbore completion parameters from the plurality of available wellbore completion parameters that optimizes estimated wellbore production based on the best predictive model; and outputting the set of wellbore completion parameters identified using the best predictive model in human-intelligible form.

Landscapes

  • Engineering & Computer Science (AREA)
  • Mining & Mineral Resources (AREA)
  • Geology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Fluid Mechanics (AREA)
  • Geochemistry & Mineralogy (AREA)
  • General Life Sciences & Earth Sciences (AREA)
  • Environmental & Geological Engineering (AREA)
  • Theoretical Computer Science (AREA)
  • Geophysics (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

L'invention concerne des procédés et des systèmes d'optimisation de réalisation de puits de forage et, en particulier, des procédés et des systèmes d'optimisation de paramètres de fracturation hydraulique. Dans certains modes de réalisation, un procédé permettant d'optimiser la réalisation de puits de forage inclut la collecte de données de puits de forage, le tri et le traitement des données de puits de forage collectées, utilisant les données de puits de forage triées et traitées pour définir un modèle optimisé, et utilisant le modèle optimisé pour évaluer des combinaisons de paramètres de réalisation de puits de forage disponibles. Dans certains cas, le modèle optimisé est défini à l'aide de réseaux de neurones artificiels, d'algorithmes génétiques, et/ou d'arbres de régression fortifiés. En outre, dans certains modes de réalisation les combinaisons de paramètres de réalisation de puits de forage disponibles incluent des paramètres de fracturation hydraulique, par exemple le nombre de fractures, le type fluide de fracturation, le type d'agent de soutènement, le volume de fracturation, et/ou d'autres paramètres.
PCT/US2012/053795 2012-09-05 2012-09-05 Procédés d'optimisation de réalisation de puits de forage et de fracturation hydraulique et systèmes associés WO2014039036A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/US2012/053795 WO2014039036A1 (fr) 2012-09-05 2012-09-05 Procédés d'optimisation de réalisation de puits de forage et de fracturation hydraulique et systèmes associés

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/US2012/053795 WO2014039036A1 (fr) 2012-09-05 2012-09-05 Procédés d'optimisation de réalisation de puits de forage et de fracturation hydraulique et systèmes associés

Publications (1)

Publication Number Publication Date
WO2014039036A1 true WO2014039036A1 (fr) 2014-03-13

Family

ID=50237495

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2012/053795 WO2014039036A1 (fr) 2012-09-05 2012-09-05 Procédés d'optimisation de réalisation de puits de forage et de fracturation hydraulique et systèmes associés

Country Status (1)

Country Link
WO (1) WO2014039036A1 (fr)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016079625A1 (fr) * 2014-11-18 2016-05-26 Weatherford Technology Holdings, Llc Systèmes et procédés conçus pour optimiser des opérations de fracturation de formation
WO2017078989A1 (fr) * 2015-11-05 2017-05-11 Schlumberger Technology Corporation Conception de fracturations hydrauliques
WO2018031051A1 (fr) * 2016-08-08 2018-02-15 Schlumberger Technology Corporation Génération d'un ensemble de formation à l'apprentissage machine
US9946974B2 (en) 2013-06-10 2018-04-17 Exxonmobil Upstream Research Company Determining well parameters for optimization of well performance
CN111580179A (zh) * 2019-01-30 2020-08-25 中国石油天然气股份有限公司 一种有机碳含量确定方法、装置及系统
WO2020180295A1 (fr) * 2019-03-04 2020-09-10 Landmark Graphics Corporation Optimisation de paramètres de complétion à objectifs multiples pour un puits de forage par optimisation bayésienne
US11941563B1 (en) 2022-09-23 2024-03-26 David Cook Apparatus and method for fracking optimization

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5012675A (en) * 1989-07-25 1991-05-07 Amoco Corporation Integrating multiple mappable variables for oil and gas exploration
US6832159B2 (en) * 2002-07-11 2004-12-14 Schlumberger Technology Corporation Intelligent diagnosis of environmental influence on well logs with model-based inversion
US7516793B2 (en) * 2007-01-10 2009-04-14 Halliburton Energy Service, Inc. Methods and systems for fracturing subterranean wells
US20090260880A1 (en) * 2008-04-18 2009-10-22 Thambynayagam R K Method for determining a set of net present values to influence the drilling of a wellbore and increase production
US7925482B2 (en) * 2006-10-13 2011-04-12 Object Reservoir, Inc. Method and system for modeling and predicting hydraulic fracture performance in hydrocarbon reservoirs
US8229880B2 (en) * 2008-01-11 2012-07-24 Schlumberger Technology Corporation Evaluation of acid fracturing treatments in an oilfield

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5012675A (en) * 1989-07-25 1991-05-07 Amoco Corporation Integrating multiple mappable variables for oil and gas exploration
US6832159B2 (en) * 2002-07-11 2004-12-14 Schlumberger Technology Corporation Intelligent diagnosis of environmental influence on well logs with model-based inversion
US7925482B2 (en) * 2006-10-13 2011-04-12 Object Reservoir, Inc. Method and system for modeling and predicting hydraulic fracture performance in hydrocarbon reservoirs
US7516793B2 (en) * 2007-01-10 2009-04-14 Halliburton Energy Service, Inc. Methods and systems for fracturing subterranean wells
US8229880B2 (en) * 2008-01-11 2012-07-24 Schlumberger Technology Corporation Evaluation of acid fracturing treatments in an oilfield
US20090260880A1 (en) * 2008-04-18 2009-10-22 Thambynayagam R K Method for determining a set of net present values to influence the drilling of a wellbore and increase production

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
SHELLEY ET AL.: "The use of artificial neural networks in completion stimulation and design.", COMPUTERS & GEOSCIENCES, vol. 26, 2000, pages 941 - 951, XP055185751 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10963815B2 (en) 2013-06-10 2021-03-30 Exxonmobil Upstream Research Company Determining well parameters for optimization of well performance
US9946974B2 (en) 2013-06-10 2018-04-17 Exxonmobil Upstream Research Company Determining well parameters for optimization of well performance
WO2016079625A1 (fr) * 2014-11-18 2016-05-26 Weatherford Technology Holdings, Llc Systèmes et procédés conçus pour optimiser des opérations de fracturation de formation
WO2017078989A1 (fr) * 2015-11-05 2017-05-11 Schlumberger Technology Corporation Conception de fracturations hydrauliques
WO2018031051A1 (fr) * 2016-08-08 2018-02-15 Schlumberger Technology Corporation Génération d'un ensemble de formation à l'apprentissage machine
US11377931B2 (en) 2016-08-08 2022-07-05 Schlumberger Technology Corporation Machine learning training set generation
CN111580179A (zh) * 2019-01-30 2020-08-25 中国石油天然气股份有限公司 一种有机碳含量确定方法、装置及系统
CN111580179B (zh) * 2019-01-30 2023-02-28 中国石油天然气股份有限公司 一种有机碳含量确定方法、装置及系统
WO2020180295A1 (fr) * 2019-03-04 2020-09-10 Landmark Graphics Corporation Optimisation de paramètres de complétion à objectifs multiples pour un puits de forage par optimisation bayésienne
GB2594613A (en) * 2019-03-04 2021-11-03 Landmark Graphics Corp Multi-objective completion parameters optimization for a wellbore using Bayesian optimization
GB2594613B (en) * 2019-03-04 2022-10-26 Landmark Graphics Corp Multi-objective completion parameters optimization for a wellbore using Bayesian optimization
US11493664B2 (en) 2019-03-04 2022-11-08 Landmark Graphics Corporation Multi-objective completion parameters optimization for a wellbore using Bayesian optimization
US11941563B1 (en) 2022-09-23 2024-03-26 David Cook Apparatus and method for fracking optimization

Similar Documents

Publication Publication Date Title
US9262713B2 (en) Wellbore completion and hydraulic fracturing optimization methods and associated systems
US10901105B2 (en) Method and system for regression and classification in subsurface models to support decision making for hydrocarbon operations
Luo et al. Production optimization using machine learning in Bakken shale
US20200124753A1 (en) Method and System for Evaluating Variability in Subsurface Models to Support Decision Making for Hydrocarbon Operations
WO2014039036A1 (fr) Procédés d'optimisation de réalisation de puits de forage et de fracturation hydraulique et systèmes associés
AU2011380514B2 (en) Statistical reservoir model based on detected flow events
US8229880B2 (en) Evaluation of acid fracturing treatments in an oilfield
US11748644B2 (en) System and method for oil and gas predictive analytics
AU2011283109A1 (en) Systems and methods for predicting well performance
Chaikine et al. A machine learning model for predicting multi-stage horizontal well production
NO20220089A1 (en) Probability Distribution Assessment for Classifying Subterranean Formations Using Machine Learning
Temizel et al. Production forecasting in shale reservoirs through conventional DCA and machine/deep learning methods
Sankaran et al. Data Analytics in Reservoir Engineering
Temizel et al. A thorough review of machine learning applications in oil and gas industry
Cai et al. Development of a powerful data-analysis tool using nonparametric smoothing models to identify drillsites in tight shale reservoirs with high economic potential
Temizel et al. Turning Data into Knowledge: Data-Driven Surveillance and Optimization in Mature Fields
WO2020142257A1 (fr) Procédé et système d'évaluation de variabilité dans des modèles de subsurface pour supporter la prise de décisions dans des exploitations d'hydrocarbures
US20190219716A1 (en) Classifying Well Data Using A Support Vector Machine
WO2020161518A1 (fr) Procédé de détection de passages de glissement horizontaux d'hydrocarbures
Mutalova et al. Machine learning on field data for hydraulic fracturing design optimization
Roden et al. Seismic interpretation with machine learning
Fathi et al. Simultaneous well spacing and completion optimization using an automated machine learning approach. A case study of the Marcellus Shale reservoir, northeastern United States
Willigers et al. Hot spot hunting: Optimising the staged development of shale plays
Temizel et al. Hybrid modeling in unconventional reservoirs to forecast estimated ultimate recovery
Ashayeri et al. A Stochastic Method in Investigating Basin-Wide Underlying Distribution Functions of Decline Rate Behavior for Unconventional Resources

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 12884179

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 12884179

Country of ref document: EP

Kind code of ref document: A1