DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Responsive to the communication dated 04/22/2026
Claims 1-7, 9-14, and 16-22 are presented for examination
Information Disclosure Statement
The IDS dated 09/28/2022 and 01/23/2023 has been reviewed. See attached.
Drawings
The drawings dated 09/28/2022 have been reviewed. They are accepted.
Finality
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Response to Arguments- 35 USC § 112
Applicant’s arguments, see Page 8 filed 04/22/2026, with respect to the rejection of claims 6, 13, and 20 under 35 USC § 112 have been fully considered and are persuasive. The rejection of claims 6, 13, and 20 under 35 USC § 112 has been withdrawn.
Response to Arguments- 35 USC § 101
Applicant’s arguments, see Pages 9-16, filed 04/22/2026, with respect to the rejection of claims 1-7, 9-14, and 16-22 under 35 USC § 101 have been fully considered and are persuasive. The rejection of claims 1-7, 9-14, and 16-22 under 35 USC § 101 has been withdrawn.
In particular, the specific machine learning training operations and the use of these operations to improve the accuracy of a simulation successfully integrate the claims into a practical application.
Response to Arguments- 35 USC § 103
Applicant's arguments filed 04/22/2026 have been fully considered but they are not persuasive.
Applicant argues that no prior art teaches “adjusting, by a reinforcement learning (RL) agent, the input parameter to the simulator based on a difference between the actual net pressure value and a corresponding simulated net pressure value, wherein the difference comprises a mean squared error calculation and is provided to the RL agent as a reward signal;” Specifically, the applicant argues that no prior art teaches that the adjustment is based on a mean squared error difference between the actual and simulated net pressure values. Further, applicant argues that no prior art teaches providing an indication of an event at the well based on the actual and corresponding simulated net pressure values.
Examiner responds by explaining that this is taught by the previously cited references.
In particular, Bello teaches c) adjusting, ([Par 83-85] “The calibration or adjustment process includes automatically determining or adjusting the values of a set of model parameters that result in the best matching of the simulated and measured data. … For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. The model is also continuously or periodically monitored by comparing values simulated by the forward model to a pre-specified tolerance. If simulated and measured values are in agreement within the tolerance, that is, if the current deviations are below a pre-determined threshold, the model is stored for continuous use and the updating process is suspended. However, if the simulated values are outside the tolerance, the system automatically calibrates the model by using the inverse model to adjust the forward model parameters so that acceptable agreement between simulated or calculate values and measured values is reached. When satisfactory agreement is reached, the production and formation properties may be output to a user, and the resulting numerical model may be used to make deterministic and probabilistic production forecasting until the next scheduled update time. FIGS. 6 and 7 illustrate workflows for automatic calibration and updating of the thermal reservoir model using the inversion algorithm. An embodiment of a workflow 130 includes specifying initial forward model parameter values (block 131) and forwarding them to, for example, the simulation module 80. The forward model is run using the initial values to generate calculated data, i.e., predicted values for measurement data (block 132). Measurement (field observed) data is collected and processed, and an objective function value is calculated based on differences between the calculated data and the measurement data (block 133). If the objective function value is within a selected range, e.g., within some selected minimum of the objective function or error value (block 134), the model parameters are considered to be acceptable and the calibration or update process ends. If the objective function value is not within the selected range, the forward model is run again (block 135) and a new objective function value is calculated. This process is repeated using the above-described inversion until the objective function value is minimized or within the selected range.” [Par 53] “The fluid composition data is used to perform a non-linear regression analysis to adjust the SRK equation of state. The basic objectives of using the non-linear regression in the fluid property model is to determine the optimum set of multipliers such that the observed or measured PVT data best match or fit as closely as possible to the calculated data from the SRK equation of state fluid model. Key parameters on which the regression is performed include any set of three independent variables. Examples of such variables include gas-oil ratio (GOR), oil density or API gravity, oil viscosity and bubble point pressure. The GOR, API gravity and bubble point pressure variables are chosen to minimize the objective functions with a value between +1%.”[Fig. 4, 6, and 7] Show aspects of calibration process, including an initial simulation/model run before gathering actual data, comparing the simulated data to measured data, adjusting simulation/model parameters based on this comparison, and rerunning the simulation/model, repeating the comparing, adjusting, and rerunning until the simulation/model output sufficiently matches the measured data.) ([Par 84] “For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. The model is also continuously or periodically monitored by comparing values simulated by the forward model to a pre-specified tolerance. If simulated and measured values are in agreement within the tolerance, that is, if the current deviations are below a pre-determined threshold, the model is stored for continuous use and the updating process is suspended. However, if the simulated values are outside the tolerance, the system automatically calibrates the model by using the inverse model to adjust the forward model parameters so that acceptable agreement between simulated or calculate values and measured values is reached. When satisfactory agreement is reached, the production and formation properties may be output to a user” [Par 71] “The inverse model is used to generate real time estimates of parameters such as production rates (also referred to as well rates) and formation properties specific to a given measurement domain by numerically reproducing available measurements and comparing them to actual measurements. The inverse model employs an inversion algorithm that attempts to minimize an objective function that is dependent on the difference between actual measurement values and values estimated using the forward model.”)
Craig makes obvious adjusting, by ([Col 30 line 27-48] “In some implementations, pressure history matching may also be used to refine a probability distribution for fracture parameters. In some implementations, in addition to comparing fracture pattern models to microseismic event data, formation pressures observed during an injection treatment are compared to formation pressures simulated using the fracture pattern model. For example, a fracture pattern models (e.g., "matches" or "mismatches") may be selected based on a correlation (or lack thereof) between observed formation pressure and simulated formation pressure. The observed formation pressure may be recorded during an injection treatment, and the fracture pattern model may be used to calculate a model formation pressure. Selecting fracture property values that minimize the difference between the observed formation pressure and the model formation pressure may lead to an improved distribution of fracture property values. For example, comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used. A pressure matching technique may present graphical comparisons to a user (e.g., Cartesian, log-log, and/or other plots of observed pressure and model pressure versus time. A pressure matching technique may include an automated technique that calculates differences between observed and model formation pressures over time.”) ([Col 30 line 27-48] “formation pressures observed during an injection treatment are compared to formation pressures simulated using the fracture pattern model. For example, a fracture pattern models (e.g., "matches" or "mismatches") may be selected based on a correlation (or lack thereof) between observed formation pressure and simulated formation pressure. … comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used.”)
Li makes obvious adjusting, by a reinforcement learning (RL) agent, ([Abstract] “. In this study, we develop deep neural networks within the reinforcement learning framework to achieve automated history matching that will reduce engineers’ efforts, human bias, automatically and intelligently explore the parameter space, and remove the need of large set of labeled training data. To that end, a fast-marching-based reservoir simulator is encapsulated as an environment for the proposed reinforcement learning. The deep neural-network-based learning agent interacts with the reservoir simulator within reinforcement learning framework to achieve the automated history matching” [Page 4 Col 1 Par 14] “In our study, the RL-based agent learns to fit the historical production data in optimal number of steps by iteratively adjusting reservoir model parameters. The agent interacts with the reservoir simulator (environment) by adjusting the reservoir parameters (action) based on the existing data misfit (state)”) the model, based on a difference wherein the difference comprises a mean squared error calculation and is provided to the RL agent as a reward signal; ([Page 4 Col 2 Par 2] “To utilize the balanced exploration and exploitation characteristic of reinforcement learning, we define the history matching process as a sequential decision-making problem. The process contains a number of episodes and each episode contains a number of iterations. Each episode represents a game to obtain good fit with production history and each iteration represents the action taken by the agent at each step of the game to obtain the optimal fit. In each episode, the environment will start with the initial guess of the reservoir parameters, and the learning agent will take an action for each iteration to increase or decrease the reservoir parameters. Each action changes the parameters of the reservoir and the environment will run a simulation to simulate the pro duction history with new reservoir parameters. The misfit between the predicted and true production history will be calculated by the objective function, which is Mean Squared Error (MSE) in our study. The reward is negative MSE, which will increase when misfit decreases.”) …
Clearly, Li teaches the use of mean squared error as a reward function for the RL agent.
Further, McLeod makes obvious detecting an event based on pressure values; ([Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described. By accurately detecting tip screen-out with a bottom hole pressure gauge mounted to a perforating gun, the optimal amount of proppant can be supplied to a fracture while avoiding the risks associated with wellbore screen-out.”)
Applicant argues that the teachings of Li are unapplicable because Li performs history matching and the claimed invention allegedly only works in real time.
Examiner responds by explaining that the broadest reasonable interpretation (BRI) of the claim language does not actually require that the operation is currently happening, let alone that the data is updated in real time. As “near-real time” has an indefinite scope, due to the fact that what is close enough to be considered “near” real time is a subjective opinion that may differ from person to person, the language of “d) performing, by the simulator, an updated simulation in real time or near-real time” includes simulations that are updated at any time scale, be that every 10 ms or every 10,000 years. With such unlimited scope as to what is considered “near real time,” data collected at any time, including “historical” data, such as disclosed in Li, reads on these limitations.
Further, it should be noted that, while historical data is used, Li also discusses the use of “current data” ([Page 4 Col 2 Par 1] “he interaction between the agent and the environment is implemented using states, actions, and rewards. The learning agent must take an action on the environment based on the current state and will receive a corresponding reward and new state. The agent then estimates the value function, which depicts the goodness of the current state in terms of cumulative future reward. The ultimate goal is to learn a policy that maximizes expected future rewards.”)
Additionally, while the features of Li are not specifically directed to the detection of a tip screen-out event, one of ordinary skill in the art would have recognized that the RL agent-based reservoir simulation optimization system disclosed therein would be directly applicable to and improve the operations of the other applied references.
Further, as to arguments about McLeod, although McLeod does teach its screen-out detection within the context of physical measurements of a well rather than through the use of a simulation, it does this detection by calculating if the change in pressure is of a certain magnitude and rate ([Par 10] “ Preferably the sensor is a pressure sensor used for measuring pressure. When pressure is measured, in response to measured pressure a sudden buildup in pressure in the wellbore at the location of the perforating gun system during the operation wherein a proppant is being pumped into a formation adjacent to the wellbore may be detected; and in response to the detection of a sudden buildup in pressure in the wellbore, a flushing operation may be commenced in the wellbore, thereby removing excess proppant from the wellbore and preventing the wellbore from filling with excess proppant. Also, the sudden buildup of pressure that causes the flushing operation may be such that, when the pressure measurement is plotted against time on a Nolte-Smith Plot, the slope of the pressure measurement exceeds one (1.0).” [Par 33] “During hydraulic fracturing, the pressure in the wellbore is closely monitored. The pressure is typically plotted on, but not limited to, a Nolte-Smith plot 200, shown in FIG. 4, which plots the logarithm of net pressure 210 versus the logarithm of time 220. Formation characteristics and fluid friction combine to limit the effective length of a given fracture. The ideal Nolte-Smith plot 200 reflects the pressure in the wellbore 102 at the fracture treatment zone 126. Here, an increase in net pressure with a slope of less than 1.0 230 indicates that the fracture has a confined height and unrestricted propagation. A slope at or near 0.0 (zero) 240 can indicate restricted height growth with reduced propagation of the fracture, or, if a critical net pressure has been reached, it can indicate the opening of natural fissures in the productive formation 122 which cause greater leak-off of fracturing fluid. A negative slope 240 indicates unrestricted height growth. A slope of 1.0 260 indicates that propagation of the fracture has ceased near the tip of the fracture, a condition known as tip screen-out. A slope of greater than 1.0 270 indicates that the fracture is no longer accepting proppant 142.”) As this main detection calculation is based on pressure change measurement, the simulation of pressure being a prominent feature of the reservoir simulation systems of Bello, Craig, and Li, one of ordinary skill in the art would have recognized that this calculation would work identically within the context of these simulations as it does in the context of the physical, measured data of McLeod.
Applicant argues that no prior art teaches wherein the indication of the tip screen-out event is provided only after the iterative adjusting of step e) has caused the difference between the actual net pressure value and the corresponding simulated net pressure value to fall below the first threshold amount, such that the slope of actual net pressure values and the slope of simulated net pressure values are compared only when the simulator is sufficiently calibrated.
Examiner responds by explaining that this feature is taught by the previously cited references.
In particular, Bello teaches wherein the indication ([Par 83-85] “The calibration or adjustment process includes automatically determining or adjusting the values of a set of model parameters that result in the best matching of the simulated and measured data. … For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. The model is also continuously or periodically monitored by comparing values simulated by the forward model to a pre-specified tolerance. If simulated and measured values are in agreement within the tolerance, that is, if the current deviations are below a pre-determined threshold, the model is stored for continuous use and the updating process is suspended. However, if the simulated values are outside the tolerance, the system automatically calibrates the model by using the inverse model to adjust the forward model parameters so that acceptable agreement between simulated or calculate values and measured values is reached. When satisfactory agreement is reached, the production and formation properties may be output to a user, and the resulting numerical model may be used to make deterministic and probabilistic production forecasting until the next scheduled update time. FIGS. 6 and 7 illustrate workflows for automatic calibration and updating of the thermal reservoir model using the inversion algorithm. An embodiment of a workflow 130 includes specifying initial forward model parameter values (block 131) and forwarding them to, for example, the simulation module 80. The forward model is run using the initial values to generate calculated data, i.e., predicted values for measurement data (block 132). Measurement (field observed) data is collected and processed, and an objective function value is calculated based on differences between the calculated data and the measurement data (block 133). If the objective function value is within a selected range, e.g., within some selected minimum of the objective function or error value (block 134), the model parameters are considered to be acceptable and the calibration or update process ends. If the objective function value is not within the selected range, the forward model is run again (block 135) and a new objective function value is calculated. This process is repeated using the above-described inversion until the objective function value is minimized or within the selected range.” [Par 53] “The fluid composition data is used to perform a non-linear regression analysis to adjust the SRK equation of state. The basic objectives of using the non-linear regression in the fluid property model is to determine the optimum set of multipliers such that the observed or measured PVT data best match or fit as closely as possible to the calculated data from the SRK equation of state fluid model. Key parameters on which the regression is performed include any set of three independent variables. Examples of such variables include gas-oil ratio (GOR), oil density or API gravity, oil viscosity and bubble point pressure. The GOR, API gravity and bubble point pressure variables are chosen to minimize the objective functions with a value between +1%.”[Fig. 4, 6, and 7] Show aspects of calibration process, including an initial simulation/model run before gathering actual data, comparing the simulated data to measured data, adjusting simulation/model parameters based on this comparison, and rerunning the simulation/model, repeating the comparing, adjusting, and rerunning until the simulation/model output sufficiently matches the measured data.” [Examiner’s note: [Par 84] specifies that final data output is only provided “{w}hen satisfactory agreement is reached” between simulated and calculated values in relation to the associated threshold])
Note that the term “sufficiently calibrated” renders the claim indefinite; what level of calibration is considered “sufficient” can vary between person to a person; a first person might consider a first simulation accurate enough while a second person may decide that more refinement would be needed before they are satisfied that calibration is sufficient. With this in mind, any level of calibration is interpreted as reading on “sufficiently calibrated.”
Craig makes obvious ([Col 30 line 27-48] “In some implementations, pressure history matching may also be used to refine a probability distribution for fracture parameters. In some implementations, in addition to comparing fracture pattern models to microseismic event data, formation pressures observed during an injection treatment are compared to formation pressures simulated using the fracture pattern model. For example, a fracture pattern models (e.g., "matches" or "mismatches") may be selected based on a correlation (or lack thereof) between observed formation pressure and simulated formation pressure. The observed formation pressure may be recorded during an injection treatment, and the fracture pattern model may be used to calculate a model formation pressure. Selecting fracture property values that minimize the difference between the observed formation pressure and the model formation pressure may lead to an improved distribution of fracture property values. For example, comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used. A pressure matching technique may present graphical comparisons to a user (e.g., Cartesian, log-log, and/or other plots of observed pressure and model pressure versus time. A pressure matching technique may include an automated technique that calculates differences between observed and model formation pressures over time.”)
Li makes obvious([Fig. 9] Shows a comparison of the deviation between measured and simulated pressure and the deviation between the derivatives of the measured and simulated pressures)
PNG
media_image1.png
578
1557
media_image1.png
Greyscale
McLeod makes obvious wherein the indication of the tip screen-out event is provided; ([Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described. By accurately detecting tip screen-out with a bottom hole pressure gauge mounted to a perforating gun, the optimal amount of proppant can be supplied to a fracture while avoiding the risks associated with wellbore screen-out.”)
Applicant argues that no prior art teaches wherein the slope of actual net pressure values deviating from the slope of simulated net pressure values by more than the second threshold amount comprises an increase in the slope of actual net pressure values relative to the slope of simulated net pressure values.
Examiner responds by explaining that this feature is taught by the previously cited references.
In particular, Bello teaches ([Par 70] “At block 84, pressure and temperature changes are calculated. At block 85, estimated parameters including production and/or injection rates are compared to pre-selected facility constraints (e.g., oil, gas and water processing capacity), and if the estimated parameters exceed such constraints, they are re-calculated.”)
Li makes obvious wherein the slope of actual net pressure values deviating from the slope of simulated net pressure values comprises an increase in the slope of actual net pressure values relative to the slope of simulated net pressure values. ([Fig. 13] Shows a situation where there is an in the slope of actual net pressure values relative to the simulated net pressure values for a portion of the time period)
PNG
media_image2.png
519
1361
media_image2.png
Greyscale
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 21-22 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The term “sufficiently calibrated” in claim 21 is a relative term which renders the claim indefinite. The term “sufficiently calibrated” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. What level of calibration is considered “sufficient” can vary between person to a person; a first person might consider a first simulation accurate enough while a second person may decide that more refinement would be needed before they are satisfied that calibration is sufficient. With this in mind, for the purposes of this examination, any level of calibration is interpreted as reading on “sufficiently calibrated.”
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
(1) Claims 1-7, 9-14, and 16-22 are rejected under 35 U.S.C. 103 as being unpatentable over Bello (US 20160356125 A1) in view of Craig (US 8437962 B2) in further view of Reinforcement learning based automated history matching for improved hydrocarbon production forecast (Hereinafter Li) as well as McLeod (US 20090218094 A1)
Claim 1. Bello makes obvious A method for a ([Abstract] “A method of online real-time estimation of production performance properties includes receiving real-time field data taken by downhole sensors, and estimating formation properties and production performance properties by applying the field data to a two or three-dimensional numerical transient thermal multiphase reservoir flow model, and automatically calibrating the model.” a) performing, by a simulator, an initial simulation based on geological data of the well and an input parameter, wherein the initial simulation provides simulated net pressure values as a function of time for the well; ([Par 84-85] “For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. … An embodiment of a workflow 130 includes specifying initial forward model parameter values (block 131) and forwarding them to, for example, the simulation module 80. The forward model is run using the initial values to generate calculated data, i.e., predicted values for measurement data (block 132).” [Par 80] “Static data 104 and dynamic or transient (e.g., real time) data 106 are forwarded via an input module 108 to a forward simulation module 110. Simulation output 112 (e.g., flow rates, production estimates, production predictions) is output to a graphical user interface (GUI) module 114 for presentation to a user.” [Par 76] “At block 92, simulation input is prepared, which includes the initial parameters, and forward simulations are run using the forward model at block 93. At block 94, forward simulation output is received, such as fluid flow rates, multi-phase flow rates and production allocation. At block 95, additional measurements are taken during production (at a next time step immediately following the current time step), such as downhole pressure, temperature and flow rates. At block 96, an inversion model is run to calculate values of expected measurements, e.g., pressure, temperature and/or flow rates, compare the expected measurements to the additional measurements and iteratively adjust parameters of the forward model.” [Par 60] “Initial conditions used to run the forward model include, e.g., pressure (P), temperature (T), water-oil ratio (WOR), gas-oil ratio (GOR), and concentrations of any chemical solutes as a function of borehole radius (r) and depth (z) at an initial time “t”. Permeability and porosity are also specified as functions of position. Random and stochastic distributions of properties are possible, as well as uniform distributions in each producing layer. The boundary conditions required by the model are P, T, WOR, GOR and concentrations of any solutes in the borehole, and P, T, WOR, GOR and concentrations of solutes at the outer side of the domain (shut-in values).” [Par 68] “At block 81, the forward model is initialized by inputting static data such as well trajectory and injection fluid properties, and inputting dynamic data such as temperature and pressure measurements (e.g., PDG and/or PLT measurements) taken downhole at a specific time (t=0). In the absence of certain design parameters from provided datasets, certain assumptions and starting values may be assumed, such as static geothermal gradients of reservoirs, thickness of the producing zones and others.” [Fig. 4] Shows the calibration process, including an initial simulation run before gathering actual data) b) receiving an indication of an actual ([Par 83-85] “The calibration or adjustment process includes automatically determining or adjusting the values of a set of model parameters that result in the best matching of the simulated and measured data. … For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. The model is also continuously or periodically monitored by comparing values simulated by the forward model to a pre-specified tolerance. If simulated and measured values are in agreement within the tolerance, that is, if the current deviations are below a pre-determined threshold, the model is stored for continuous use and the updating process is suspended. However, if the simulated values are outside the tolerance, the system automatically calibrates the model by using the inverse model to adjust the forward model parameters so that acceptable agreement between simulated or calculate values and measured values is reached. When satisfactory agreement is reached, the production and formation properties may be output to a user, and the resulting numerical model may be used to make deterministic and probabilistic production forecasting until the next scheduled update time. FIGS. 6 and 7 illustrate workflows for automatic calibration and updating of the thermal reservoir model using the inversion algorithm. An embodiment of a workflow 130 includes specifying initial forward model parameter values (block 131) and forwarding them to, for example, the simulation module 80. The forward model is run using the initial values to generate calculated data, i.e., predicted values for measurement data (block 132). Measurement (field observed) data is collected and processed, and an objective function value is calculated based on differences between the calculated data and the measurement data (block 133). If the objective function value is within a selected range, e.g., within some selected minimum of the objective function or error value (block 134), the model parameters are considered to be acceptable and the calibration or update process ends. If the objective function value is not within the selected range, the forward model is run again (block 135) and a new objective function value is calculated. This process is repeated using the above-described inversion until the objective function value is minimized or within the selected range.” [Par 53] “The fluid composition data is used to perform a non-linear regression analysis to adjust the SRK equation of state. The basic objectives of using the non-linear regression in the fluid property model is to determine the optimum set of multipliers such that the observed or measured PVT data best match or fit as closely as possible to the calculated data from the SRK equation of state fluid model. Key parameters on which the regression is performed include any set of three independent variables. Examples of such variables include gas-oil ratio (GOR), oil density or API gravity, oil viscosity and bubble point pressure. The GOR, API gravity and bubble point pressure variables are chosen to minimize the objective functions with a value between +1%.”[Fig. 4, 6, and 7] Show aspects of calibration process, including an initial simulation/model run before gathering actual data, comparing the simulated data to measured data, adjusting simulation/model parameters based on this comparison, and rerunning the simulation/model, repeating the comparing, adjusting, and rerunning until the simulation/model output sufficiently matches the measured data.) f) providing an indication of([Par 84] “For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. The model is also continuously or periodically monitored by comparing values simulated by the forward model to a pre-specified tolerance. If simulated and measured values are in agreement within the tolerance, that is, if the current deviations are below a pre-determined threshold, the model is stored for continuous use and the updating process is suspended. However, if the simulated values are outside the tolerance, the system automatically calibrates the model by using the inverse model to adjust the forward model parameters so that acceptable agreement between simulated or calculate values and measured values is reached. When satisfactory agreement is reached, the production and formation properties may be output to a user” [Par 71] “The inverse model is used to generate real time estimates of parameters such as production rates (also referred to as well rates) and formation properties specific to a given measurement domain by numerically reproducing available measurements and comparing them to actual measurements. The inverse model employs an inversion algorithm that attempts to minimize an objective function that is dependent on the difference between actual measurement values and values estimated using the forward model.”)
Bello does not explicitly teach A method for a frac packing completion operation of a well, wherein the simulation provides simulated net pressure; an actual net pressure value measured during the frac packing operation; adjusting, by a reinforcement learning (RL) agent, the model, based on a difference between the actual net pressure value and a corresponding simulated net pressure value, wherein the difference comprises a mean squared error calculation and is provided to the RL agent as a reward signal; simulated net pressure corresponding to a simulation that generates simulated net pressure values; detecting an event based on the actual net pressure value and the simulated net pressure value.
Craig makes obvious ([Col 30 line 27-48] “In some implementations, pressure history matching may also be used to refine a probability distribution for fracture parameters. In some implementations, in addition to comparing fracture pattern models to microseismic event data, formation pressures observed during an injection treatment are compared to formation pressures simulated using the fracture pattern model. For example, a fracture pattern models (e.g., "matches" or "mismatches") may be selected based on a correlation (or lack thereof) between observed formation pressure and simulated formation pressure. The observed formation pressure may be recorded during an injection treatment, and the fracture pattern model may be used to calculate a model formation pressure. Selecting fracture property values that minimize the difference between the observed formation pressure and the model formation pressure may lead to an improved distribution of fracture property values. For example, comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used. A pressure matching technique may present graphical comparisons to a user (e.g., Cartesian, log-log, and/or other plots of observed pressure and model pressure versus time. A pressure matching technique may include an automated technique that calculates differences between observed and model formation pressures over time.”) ([Col 30 line 27-48] “formation pressures observed during an injection treatment are compared to formation pressures simulated using the fracture pattern model. For example, a fracture pattern models (e.g., "matches" or "mismatches") may be selected based on a correlation (or lack thereof) between observed formation pressure and simulated formation pressure. … comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used.”)
Craig is analogous art because it is within the field of well and reservoir simulation. It would have been obvious to one of ordinary skill in the art to combine it with Bello before the effective filing date. One of ordinary skill in the art would have been motivated to make this combination in order to more accurately model fractures, particularly when dealing with highly complex fracture geometries with unknown conditions. As noted by Craig, the particular features of underground reservoirs are typically unknown. ([Col 1 Line 27-37] “In many underground petroleum reservoirs, properties of the discrete rock blocks and characteristics of discontinuities are known with some uncertainty. For example, the exact pattern of fractures, faults, fissures, and other features, existing in the reservoir are typically not known with certainty, and probability distributions for the discontinuities can be generated based on data from analog fields, outcrop mapping, open hole logging, microseismic data, and/or other information. The uncertainty may result from imprecise or incomplete knowledge of the rock properties, inhomogeneity of the rock properties, and/or other sources of uncertainty”) Logically, this lack of data can result in the inaccuracy or incompleteness of typical models attempting to use this data. To this end, Craig presents a method for simulating using a probabilistic model that can account for this uncertainty ([Col 8 Line 37- Col 9 Line 14] “Uncertainty in the properties of the rock blocks and characteristics of the discontinuities can be accounted for in numerical simulations of the fracture network by defining a probabilistic earth model. The probabilistic earth model, which includes probability distributions that describe ranges of values for each input variable (and a probability for each value), can be used to populate geometric models that serve as an inputs for probabilistic simulations of complex fracture growth. A probabilistic earth model can describe, among other things, discontinuities in a subterranean region. For example, the discontinuities can include discontinuities at any orientation, including lateral discontinuities that create rock blocks in a single layer, vertical discontinuities that create a multilayer system of reservoir rocks, fracture sets having a primary orientation, fracture sets having a secondary orientation, and/or others. … In some implementations, using a probabilistic earth model to populate a geometric model for complex fracture simulation can be used to achieve one or more advantages. For example, a probabilistic earth model may allow for both lateral discontinuities and vertical discontinuities to be included in the geometric model. The lateral discontinuities may represent, for example, lateral and vertical changes in lithology as well as fracture discontinuities, fissures, and faults. A probabilistic earth model may allow complex rock geometries (e.g., lenticular rock geometries, etc.) to be included in the geometric model used for complex fracture simulation.”) Overall, one of ordinary skill in the art would have recognized that combining Craig with Bello would result in a system capable of more complete, accurate simulations, even if certain information is unknown.
The combination of Bello and Craig does not explicitly teach A method for a frac packing completion operation of a well, a pressure value measured during the frac packing operation; adjusting, by a reinforcement learning (RL) agent, the model, based on a difference wherein the difference comprises a mean squared error calculation and is provided to the RL agent as a reward signal; detecting an event based on pressure values;
Li makes obvious ([Abstract] “. In this study, we develop deep neural networks within the reinforcement learning framework to achieve automated history matching that will reduce engineers’ efforts, human bias, automatically and intelligently explore the parameter space, and remove the need of large set of labeled training data. To that end, a fast-marching-based reservoir simulator is encapsulated as an environment for the proposed reinforcement learning. The deep neural-network-based learning agent interacts with the reservoir simulator within reinforcement learning framework to achieve the automated history matching” [Page 4 Col 1 Par 14] “In our study, the RL-based agent learns to fit the historical production data in optimal number of steps by iteratively adjusting reservoir model parameters. The agent interacts with the reservoir simulator (environment) by adjusting the reservoir parameters (action) based on the existing data misfit (state)”) based on a difference wherein the difference comprises a mean squared error calculation and is provided to the RL agent as a reward signal; ([Page 4 Col 2 Par 2] “To utilize the balanced exploration and exploitation characteristic of reinforcement learning, we define the history matching process as a sequential decision-making problem. The process contains a number of episodes and each episode contains a number of iterations. Each episode represents a game to obtain good fit with production history and each iteration represents the action taken by the agent at each step of the game to obtain the optimal fit. In each episode, the environment will start with the initial guess of the reservoir parameters, and the learning agent will take an action for each iteration to increase or decrease the reservoir parameters. Each action changes the parameters of the reservoir and the environment will run a simulation to simulate the pro duction history with new reservoir parameters. The misfit between the predicted and true production history will be calculated by the objective function, which is Mean Squared Error (MSE) in our study. The reward is negative MSE, which will increase when misfit decreases.”)
Li is analogous art because it is within the field of well and reservoir simulation. It would have been obvious to one of ordinary skill in the art to combine it with Bello and Craig before the effective filing date. One of ordinary skill in the art would have been motivated to make this combination in order to make the calibration process more efficient. As noted by Li, a typical calibration or “history matching” process in the field of well and reservoir simulation can take immense computational resources, making them cumbersome to operate ([Page 2 Col 1 Par 3] “During the history matching process, many numerical simulations are required to reliably evaluate the misfit between the predictions of the numerical models and the historical reservoir production data. Running many numerical simulations is computationally expensive and cannot be afforded under limited computational resources”) To this end, Li presents a system using Multistencils Fast Marching and Reinforcement learning to make the calibration process more efficient while preserving high accuracy levels ([Page 4 Col 2 Par 3] “Conventional reservoir simulation calculates the reservoir pressure and flow rate by solving many nonlinear partial equations generally using the finite-difference method. With the advance of computer technology, computation resources are becoming cheaper and more accessible. However, it still takes a lot of time and computational resources to run large-scale reservoir simulations. In history matching, many reservoir simulations are needed for the various realizations of the reservoir model. The fast-marching method was developed to decrease the reservoir simulation time for purposes of reservoir simulation. The fast-marching method belongs to the streamlined method and is good at tracking the pressure front location [29]. Multistencils Fast Marching (MFM) was initially proposed for computer vision problems [30] to improve the accuracy of the fast-marching method. MFM was used to study the signatures of natural fractures in pressure transient responses [28,31]. In the proposed study, the RL-based agent needs to call a reservoir simulator multiple times to quantify the new state for a given action. To limit the computational cost, we implemented MFM simulator as a low-cost reservoir simulator.” [Fig. 17] [Page 16 Col 2 Par 1] “Reinforcement learning algorithms have shown great potential in solving various nonlinear problems with complex decision structures. In this paper, reinforcement learning successfully accomplished automated history matching by accurately estimating the reservoir permeabilities.”) Overall, one of ordinary skill in the art would have recognized that combining Li with Bello and Craig would result in a model calibration process that is significantly more efficient with little reduction of accuracy.
The combination of Bello, Craig, and Li does not explicitly teach A method for a frac packing completion operation of a well, a pressure value measured during the frac packing operation; detecting an event based on pressure values;
McLeod makes obvious A method for a frac packing completion operation of a well, ([Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described.” [Par 28] “The proppant 144 is any suitable proppant may be used, provided that it is compatible with the formation, the slurry, and the desired results. Such proppants (gravels) can be natural or synthetic, coated, or contain chemicals; more than one can be used sequentially or in mixtures of different sizes or different materials. Proppants and gravels in the same or different wells or treatments can be the same material and/or the same size as one another and the term "proppant" is intended to include gravel in this discussion.”[Examiner’s note: frac packing is a technique in which gravel is used as a proppant during hydraulic fracturing]) a pressure value measured during the frac packing operation; ([Par 10] “The sensor packages used in accordance with the invention may comprise one or more of a pressure sensor, temperature sensor, pH sensor, or any combination thereof, while the parameters measured are at least one or more of pressure, temperature, or pH. Of course, any other suitable sensor or sensed parameter may be used as well. Preferably the sensor is a pressure sensor used for measuring pressure. When pressure is measured, in response to measured pressure a sudden buildup in pressure in the wellbore at the location of the perforating gun system during the operation wherein a proppant is being pumped into a formation adjacent to the wellbore may be detected;” [Par 28] “The proppant 144 is any suitable proppant may be used, provided that it is compatible with the formation, the slurry, and the desired results. Such proppants (gravels) can be natural or synthetic, coated, or contain chemicals; more than one can be used sequentially or in mixtures of different sizes or different materials. Proppants and gravels in the same or different wells or treatments can be the same material and/or the same size as one another and the term "proppant" is intended to include gravel in this discussion.”[Examiner’s note: frac packing is a technique in which gravel is used as a proppant during hydraulic fracturing]) detecting an event based on pressure values; ([Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described. By accurately detecting tip screen-out with a bottom hole pressure gauge mounted to a perforating gun, the optimal amount of proppant can be supplied to a fracture while avoiding the risks associated with wellbore screen-out.” [Par 10] “ Preferably the sensor is a pressure sensor used for measuring pressure. When pressure is measured, in response to measured pressure a sudden buildup in pressure in the wellbore at the location of the perforating gun system during the operation wherein a proppant is being pumped into a formation adjacent to the wellbore may be detected; and in response to the detection of a sudden buildup in pressure in the wellbore, a flushing operation may be commenced in the wellbore, thereby removing excess proppant from the wellbore and preventing the wellbore from filling with excess proppant. Also, the sudden buildup of pressure that causes the flushing operation may be such that, when the pressure measurement is plotted against time on a Nolte-Smith Plot, the slope of the pressure measurement exceeds one (1.0).” [Par 33] “During hydraulic fracturing, the pressure in the wellbore is closely monitored. The pressure is typically plotted on, but not limited to, a Nolte-Smith plot 200, shown in FIG. 4, which plots the logarithm of net pressure 210 versus the logarithm of time 220. Formation characteristics and fluid friction combine to limit the effective length of a given fracture. The ideal Nolte-Smith plot 200 reflects the pressure in the wellbore 102 at the fracture treatment zone 126. Here, an increase in net pressure with a slope of less than 1.0 230 indicates that the fracture has a confined height and unrestricted propagation. A slope at or near 0.0 (zero) 240 can indicate restricted height growth with reduced propagation of the fracture, or, if a critical net pressure has been reached, it can indicate the opening of natural fissures in the productive formation 122 which cause greater leak-off of fracturing fluid. A negative slope 240 indicates unrestricted height growth. A slope of 1.0 260 indicates that propagation of the fracture has ceased near the tip of the fracture, a condition known as tip screen-out. A slope of greater than 1.0 270 indicates that the fracture is no longer accepting proppant 142.”)
McLeod is analogous art because it is within the field of well and reservoir operations. It would have been obvious to one of ordinary skill in the art to combine it with Bello, Craig, and Li before the effective filing date. One of ordinary skill in the art would have been motivated to make this combination in order to more reliably produce optimal fracturing. As noted by McLeod, detection of a tip screenout is extremely important, as failing to properly address them can lead to wellbore screenouts and damage they cause can be expensive to repair; yet traditional methods are incapable of accurately detecting tip screenouts and the conditions that lead to them ([Par 4-6] “Key to a successful fracturing operation is the accurate monitoring of the bottom hole pressure in the wellbore, and determining when to stop pumping fracturing fluid and initiate flush of the wellbore. Early initiation of the flush results in less than optimal fracturing of the hydrocarbon bearing formation and a less productive well. However, surface pressure measurements are prone to result in just such early initiation of the flush. This is because the pressure at the surface does not accurately reflect the conditions at the bottom of the wellbore. In particular, surface measurements include additional effects such as the friction of the flowing slurry along the length of the wellbore or the constantly changing hydrostatic pressure of the proppant laden fracturing fluid. Modeling these effects is typically not accurate enough to determine precisely when to initiate the flush based upon the surface pressure. On the other hand, if the flush is initiated too late, the pumping of additional slurry leads to wellbore screen-out, where the proppant backs up into, and fills the wellbore. Wellbore screen-out is undesirable because the proppant restricts the free flow of hydrocarbons in the wellbore and, in the extreme, can trap downhole assemblies in the wellbore. If the wellbore screen-out is significant enough, the entire process of perforation and fracturing must be stopped while wellbore repair is performed. During repair, the overpressure is released, permitting ball sealers, put in place after previous fracture treatments, to fall out, and precluding further fracturing after the repair is completed, without the placement of additional wellbore plugs. Therefore, repair of a wellbore after a wellbore screen-out is expensive and time consuming. From the foregoing it will be apparent that there remains a need to measure bottom hole pressure during fracturing operations to accurately detect tip screen-out and prevent wellbore screen-out.”) To this end, McLeod presents a method for the accurate detection of screenouts ([Par 24] “Disclosed herein is a method of measuring bottom hole pressure during perforation/hydraulic fracturing (perf/frac) operations, and using the bottom hole pressure profile to determine when to stop pumping proppant laden fracturing fluid and initiate the flush of the wellbore. In some aspects, the invention relates to real time pressure measurement in a wellbore during fracturing operations to better detect screen-out.”) Overall, one of ordinary skill in the art would have recognized that being able to accurately detect tip screenout by combining McLeod with Bello, Craig, and Li would result in a significant reduction in potential damage to a well, ultimately resulting in a more durable well system that can function longer and therefore produce more over its longer lifetime.
Claim 2. Bello teaches ([Par 84] “For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. The model is also continuously or periodically monitored by comparing values simulated by the forward model to a pre-specified tolerance. If simulated and measured values are in agreement within the tolerance, that is, if the current deviations are below a pre-determined threshold, the model is stored for continuous use and the updating process is suspended. However, if the simulated values are outside the tolerance, the system automatically calibrates the model by using the inverse model to adjust the forward model parameters so that acceptable agreement between simulated or calculate values and measured values is reached. When satisfactory agreement is reached, the production and formation properties may be output to a user”) ([Par 70] “At block 84, pressure and temperature changes are calculated. At block 85, estimated parameters including production and/or injection rates are compared to pre-selected facility constraints (e.g., oil, gas and water processing capacity), and if the estimated parameters exceed such constraints, they are re-calculated.”)
Craig makes obvious ([Col 30 line 27-48] “In some implementations, pressure history matching may also be used to refine a probability distribution for fracture parameters. In some implementations, in addition to comparing fracture pattern models to microseismic event data, formation pressures observed during an injection treatment are compared to formation pressures simulated using the fracture pattern model. For example, a fracture pattern models (e.g., "matches" or "mismatches") may be selected based on a correlation (or lack thereof) between observed formation pressure and simulated formation pressure. The observed formation pressure may be recorded during an injection treatment, and the fracture pattern model may be used to calculate a model formation pressure. Selecting fracture property values that minimize the difference between the observed formation pressure and the model formation pressure may lead to an improved distribution of fracture property values. For example, comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used. A pressure matching technique may present graphical comparisons to a user (e.g., Cartesian, log-log, and/or other plots of observed pressure and model pressure versus time. A pressure matching technique may include an automated technique that calculates differences between observed and model formation pressures over time.”)
Li makes obvious ([Fig. 9] Shows a comparison of the deviation between measured and simulated pressure and the deviation between the derivatives of the measured and simulated pressures)
PNG
media_image1.png
578
1557
media_image1.png
Greyscale
McLeod makes obvious wherein the event at the well comprises a tip screen-out event; ([Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described. By accurately detecting tip screen-out with a bottom hole pressure gauge mounted to a perforating gun, the optimal amount of proppant can be supplied to a fracture while avoiding the risks associated with wellbore screen-out.”)
Claim 3. Bello teaches providing a recommendation to an operator to adjust a pump rate, a surface pressure, or a proppant volume ([Par 36] “The system may also be configured to control wells or production systems or send control or optimization information to the production systems or users.” [Par 25] “Formation fluid from the lower production zone enters the annulus 30 of the borehole 12 through the perforations 24 and into a tubing 32 via a flow control valve 34. The flow control valve 34 (e.g., an inflow control valve (ICV) or automatic ICV (AICV)) may be a remotely control sliding sleeve valve or any other suitable valve or choke that can regulate the flow of the fluid from the annulus into the production tubing 32. An adjustable choke 36 in the tubing 32 may be used to regulate the fluid flow from the lower production zone to the surface. The formation fluid from the upper production zone enters the annulus 30 (the annulus portion above the packer 26) via perforations. The formation fluid 18 from the upper zone enters production tubing via inlets 38. An adjustable valve or choke 40 regulates the fluid flow into the tubing and may be used to adjust flow of the fluid to the surface. Each valve, choke and other such device in the well may be operated electrically, hydraulically, mechanically and/or pneumatically from the surface.” [Par 31] “In one embodiment, the control valves or other inflow control devices and sensors are disposed downhole in a “smart” or “intelligent” well configuration. Smart well technology involves measurement and reservoir flow control features that are disposed downhole. Installation of downhole active flow control devices (multi-node), inflow control valves, measurement devices (e.g., for pressure, temperature and flow rate), and/or downhole processing facilities such as hydro-cyclones in the borehole allows for active production monitoring and control. Intelligent wells facilitate control of parameters such as fluid flow and pressure, and facilitate periodically or continuously updating reservoir models during production.”[Par 26] “The system may also include an artificial lift mechanism, such as an electrical submersible pump (ESP), a gas lift system, a beam pump, a jet pump, a hydraulic pump or a progressive cavity pump, to pump fluids to the surface. For example, the system 10 includes an ESP 44 controlled by an ESP control unit.” [Par 33] “Processes that can be performed using the algorithms include estimating formation properties and production properties (e.g., fluid production rates, also referred to as well rates), performing automatic virtual well test analyses, forecasting of future production performance, automatically calibrating devices such as downhole flow meters (DFMs) by integrating production data, monitoring inflow control devices, predicting and optimizing flow control device settings to improve hydrocarbon recovery”)
Craig makes obvious a ([Col 30 Line 42- 46] “For example, comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used.”)
McLeod makes obvious teach adjusting to achieve a particular pressure gain following the event. ([Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described. By accurately detecting tip screen-out with a bottom hole pressure gauge mounted to a perforating gun, the optimal amount of proppant can be supplied to a fracture while avoiding the risks associated with wellbore screen-out.” [Par 10] “in response to measured pressure a sudden buildup in pressure in the wellbore at the location of the perforating gun system during the operation wherein a proppant is being pumped into a formation adjacent to the wellbore may be detected; and in response to the detection of a sudden buildup in pressure in the wellbore, a flushing operation may be commenced in the wellbore, thereby removing excess proppant from the wellbore and preventing the wellbore from filling with excess proppant.” [Par 35] “sensor package 136 may include a pressure sensor, pressure gauge, temperature gauge, temperature sensor, pH sensor, or any combination thereof, to measure conditions during the course of the treatment, transmit such measurement(s) to a monitoring and control computer, for real time adjustment of the treatment (i.e. fracturing treatment). As used herein, the term "real time adjustment" means measuring a downhole parameter (i.e. pressure, temperature, pH, etc.), transmitting the measurement to a monitoring system, analyzing and adjusting controllable parameters in the course of treatment, all in order to achieve treatment efficiency and reservoir optimization…”[Examiner’s note: flushing increases pressure, controlling that flushing operation in cooperation with sensors allows for that increase to be particular])
Claim 4. Bello teaches automatically adjusting a pump rate, a surface pressure, a proppant volume, or combinations thereof ([Par 25] “Formation fluid from the lower production zone enters the annulus 30 of the borehole 12 through the perforations 24 and into a tubing 32 via a flow control valve 34. The flow control valve 34 (e.g., an inflow control valve (ICV) or automatic ICV (AICV)) may be a remotely control sliding sleeve valve or any other suitable valve or choke that can regulate the flow of the fluid from the annulus into the production tubing 32. An adjustable choke 36 in the tubing 32 may be used to regulate the fluid flow from the lower production zone to the surface. The formation fluid from the upper production zone enters the annulus 30 (the annulus portion above the packer 26) via perforations. The formation fluid 18 from the upper zone enters production tubing via inlets 38. An adjustable valve or choke 40 regulates the fluid flow into the tubing and may be used to adjust flow of the fluid to the surface. Each valve, choke and other such device in the well may be operated electrically, hydraulically, mechanically and/or pneumatically from the surface.” [Par 31] “In one embodiment, the control valves or other inflow control devices and sensors are disposed downhole in a “smart” or “intelligent” well configuration. Smart well technology involves measurement and reservoir flow control features that are disposed downhole. Installation of downhole active flow control devices (multi-node), inflow control valves, measurement devices (e.g., for pressure, temperature and flow rate), and/or downhole processing facilities such as hydro-cyclones in the borehole allows for active production monitoring and control. Intelligent wells facilitate control of parameters such as fluid flow and pressure, and facilitate periodically or continuously updating reservoir models during production.” [Par 26] “The system may also include an artificial lift mechanism, such as an electrical submersible pump (ESP), a gas lift system, a beam pump, a jet pump, a hydraulic pump or a progressive cavity pump, to pump fluids to the surface. For example, the system 10 includes an ESP 44 controlled by an ESP control unit.” [Par 33] “Processes that can be performed using the algorithms include estimating formation properties and production properties (e.g., fluid production rates, also referred to as well rates), performing automatic virtual well test analyses, forecasting of future production performance, automatically calibrating devices such as downhole flow meters (DFMs) by integrating production data, monitoring inflow control devices, predicting and optimizing flow control device settings to improve hydrocarbon recovery” [Par 36] “The system may also be configured to control wells or production systems or send control or optimization information to the production systems or users.”)
McLeod makes obvious teach performing operations responsive to the event. ([Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described. By accurately detecting tip screen-out with a bottom hole pressure gauge mounted to a perforating gun, the optimal amount of proppant can be supplied to a fracture while avoiding the risks associated with wellbore screen-out.” [Par 10] “in response to measured pressure a sudden buildup in pressure in the wellbore at the location of the perforating gun system during the operation wherein a proppant is being pumped into a formation adjacent to the wellbore may be detected; and in response to the detection of a sudden buildup in pressure in the wellbore, a flushing operation may be commenced in the wellbore, thereby removing excess proppant from the wellbore and preventing the wellbore from filling with excess proppant.”)
Claim 5. Bello teaches before step a), performing a well logging operation to generate the geological data for the initial simulation. ([Par 75] “FIG. 4 illustrates a workflow 90 that schematically shows how the inversion algorithm proceeds. The inversion process performed according to this algorithm is repeated for a number of successive time steps or intervals. At block 91, initial inversion parameters are loaded to the inversion algorithm. The initial inversion parameters are similar to those input to the forward model, and include static parameters and pressure, temperature and/or flow rate measurements. These initial parameters may be based on measurement data at a current time step, a previous iteration of the model, historical data and/or best guesses based on available information.”)
Claim 6. Bello teaches wherein a time associated with the corresponding simulated ([Par 84] “For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. The model is also continuously or periodically monitored by comparing values simulated by the forward model to a pre-specified tolerance. If simulated and measured values are in agreement within the tolerance, that is, if the current deviations are below a pre-determined threshold, the model is stored for continuous use and the updating process is suspended. However, if the simulated values are outside the tolerance, the system automatically calibrates the model by using the inverse model to adjust the forward model parameters so that acceptable agreement between simulated or calculate values and measured values is reached.” [Par 53] “The fluid composition data is used to perform a non-linear regression analysis to adjust the SRK equation of state. The basic objectives of using the non-linear regression in the fluid property model is to determine the optimum set of multipliers such that the observed or measured PVT data best match or fit as closely as possible to the calculated data from the SRK equation of state fluid model. Key parameters on which the regression is performed include any set of three independent variables. Examples of such variables include gas-oil ratio (GOR), oil density or API gravity, oil viscosity and bubble point pressure. The GOR, API gravity and bubble point pressure variables are chosen to minimize the objective functions with a value between +1%.” [Par 76] “At block 92, simulation input is prepared, which includes the initial parameters, and forward simulations are run using the forward model at block 93. At block 94, forward simulation output is received, such as fluid flow rates, multi-phase flow rates and production allocation. At block 95, additional measurements are taken during production (at a next time step immediately following the current time step), such as downhole pressure, temperature and flow rates. At block 96, an inversion model is run to calculate values of expected measurements, e.g., pressure, temperature and/or flow rates, compare the expected measurements to the additional measurements and iteratively adjust parameters of the forward model.”)
Craig makes obvious simulated net pressure and actual net pressure([Col 30 line 27-48] “In some implementations, pressure history matching may also be used to refine a probability distribution for fracture parameters. In some implementations, in addition to comparing fracture pattern models to microseismic event data, formation pressures observed during an injection treatment are compared to formation pressures simulated using the fracture pattern model. For example, a fracture pattern models (e.g., "matches" or "mismatches") may be selected based on a correlation (or lack thereof) between observed formation pressure and simulated formation pressure. The observed formation pressure may be recorded during an injection treatment, and the fracture pattern model may be used to calculate a model formation pressure. Selecting fracture property values that minimize the difference between the observed formation pressure and the model formation pressure may lead to an improved distribution of fracture property values. For example, comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used. A pressure matching technique may present graphical comparisons to a user (e.g., Cartesian, log-log, and/or other plots of observed pressure and model pressure versus time. A pressure matching technique may include an automated technique that calculates differences between observed and model formation pressures over time.”)
Claim 7. Bello teaches wherein the input parameter ([Par 84-85] “For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. … An embodiment of a workflow 130 includes specifying initial forward model parameter values (block 131) and forwarding them to, for example, the simulation module 80. The forward model is run using the initial values to generate calculated data, i.e., predicted values for measurement data (block 132).”)
Craig makes obvious wherein an element of a simulation comprises a modulus of the well, a toughness of the well, a stress of the well, ([Col 7 line 27-29] “the discontinuum model can simulate fracture reorientation in response to changes in the stress field or fracturing conditions;”) a leakoff coefficient of the well, or combinations thereof.
Claim 9. The elements of claim 9 are substantially the same as those of claim 1. Therefore, the elements of claim 9 are rejected due to the same reasons as outlined above for claim 1. Further, as to the elements not present in claim 1, Craig makes obvious a surface pump configured to pressurize a fluid to a downhole net pressure measurable by a sensor package; a fluid line extending between the surface pump and a wellhead positioned at an upper end of the well, wherein the fluid line is configured to flow the fluid into the well; and a monitoring system in signal communication with the sensor package and comprising a ([Col 1 Line 25-28] “During a fracture treatment, fluids are pumped under high pressure into a rock formation through a well bore to artificially fracture the formations and increase permeability and production from the formation.” [Fig. 1B] Clearly shows a fluid line between the pump truck (114) and the well, in addition to instrumentation truck (116) which measures pressure data and controls pumping [Col 11 line 3 – 54] “The packers 105 shown in FIG. 1B seal the annulus of the well bore 101 above and below the formation 122. Packers 105 may include mechanical packers, fluid inflatable packers, sand packers, and/or other types of packers. As shown in FIG. 1B, the pump trucks 114 are coupled to the working string 107 at the surface 106. The pump trucks 114 may include mobile vehicles, immobile installations, skids, hoses, tubes, fluid tanks or reservoirs, pumps, valves, and/or other suitable structures and equipment. During operation, the pump trucks 114 pump fluid 117 to the fracturing tool 119, which performs the injection treatment by injecting the fluid 117 into the formation 122. The fluid 117 may include a pad, proppants, a flush fluid, additives, and/or other materials. … As shown in FIG. 1B, the instrument trucks 116 are also provided at the surface 106. The instrument trucks 116 may include mobile vehicles, immobile installations, and/or other suitable structures. The instrument trucks 116 may include a technical command center. The example instrument trucks 116 include an injection control system that monitors and controls the injection treatment. The injection control system may control the pump trucks 114, fracturing tool 119, fluid valves, and/or other equipment used to apply the injection treatment and/or a perforating treatment. The treatment well 102 may also include surface and down-hole sensors (not shown) to measure pressure, rate, temperature and/or other parameters of treatment and/or production. The treatment well 102 may include pump controls and/or other types of controls for starting, stopping and/or otherwise controlling pumping as well as controls for selecting and/or otherwise controlling fluids pumped during the injection treatment. The injection control system in the instrument trucks 116 can communicate with the surface and/or subsurface sensor, instruments, and other equipment to monitor and control the injection treatment. The example instrument trucks 116 shown in FIG. 1B communicate with the pump truck 114, the surface and subsurface instruments, the computing subsystem 110, and/or other systems and subsystems through one or more communication links 118…”)
PNG
media_image3.png
755
606
media_image3.png
Greyscale
The combination of Bello and Craig does not explicitly teach a reinforcement learning (RL) frac packing module.
Li makes obvious a reinforcement learning (RL) ([Page 4 Col 1 Par 14] “In our study, the RL-based agent learns to fit the historical production data in optimal number of steps by iteratively adjusting reservoir model parameters. The agent interacts with the reservoir simulator (environment) by adjusting the reservoir parameters (action) based on the existing data misfit (state)”)
The combination of Bello, Craig, and Li does not explicitly teach a frac packing module.
McLeod makes obvious a frac packing module. ([Par 36] “ In one embodiment, the perforating gun sensor package 136 includes at least a pressure gauge 136A that transmits its reading through the wireline 108 to the monitoring and control computer 152. FIG. 7 is a flowchart that describes one embodiment of the present disclosure.” [Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described.” [Par 28] “The proppant 144 is any suitable proppant may be used, provided that it is compatible with the formation, the slurry, and the desired results. Such proppants (gravels) can be natural or synthetic, coated, or contain chemicals; more than one can be used sequentially or in mixtures of different sizes or different materials. Proppants and gravels in the same or different wells or treatments can be the same material and/or the same size as one another and the term "proppant" is intended to include gravel in this discussion.”[Examiner’s note: frac packing is a technique in which gravel is used as a proppant during hydraulic fracturing])
Claims 10-14 The elements of claims 10-14 are substantially the same as those of claims 2-4, and 6-7. Therefore, the elements of claims 10-14 are rejected due to the same reasons as outlined above for claims 2-4, and 6-7.
Claims 16-20 The elements of claims 16-20 are substantially the same as those of claims 1-4 and 6. Therefore, the elements of claims 16-20 are rejected due to the same reasons as outlined above for claims 1-4 and 6.
Claim 21. Bello teaches Bello teaches wherein the indication ([Par 83-85] “The calibration or adjustment process includes automatically determining or adjusting the values of a set of model parameters that result in the best matching of the simulated and measured data. … For example, the system on a pre-specified periodic basis (every five minutes, for example) determines production and formation property values using the forward model, which may be output to a user interface. Such values include one or more of multiphase flow rates, reservoir production allocation, reservoir pressure and temperature profiles, bottomhole flowing pressures and temperatures, dynamic reservoir characterization and produced water cut. The model is also continuously or periodically monitored by comparing values simulated by the forward model to a pre-specified tolerance. If simulated and measured values are in agreement within the tolerance, that is, if the current deviations are below a pre-determined threshold, the model is stored for continuous use and the updating process is suspended. However, if the simulated values are outside the tolerance, the system automatically calibrates the model by using the inverse model to adjust the forward model parameters so that acceptable agreement between simulated or calculate values and measured values is reached. When satisfactory agreement is reached, the production and formation properties may be output to a user, and the resulting numerical model may be used to make deterministic and probabilistic production forecasting until the next scheduled update time. FIGS. 6 and 7 illustrate workflows for automatic calibration and updating of the thermal reservoir model using the inversion algorithm. An embodiment of a workflow 130 includes specifying initial forward model parameter values (block 131) and forwarding them to, for example, the simulation module 80. The forward model is run using the initial values to generate calculated data, i.e., predicted values for measurement data (block 132). Measurement (field observed) data is collected and processed, and an objective function value is calculated based on differences between the calculated data and the measurement data (block 133). If the objective function value is within a selected range, e.g., within some selected minimum of the objective function or error value (block 134), the model parameters are considered to be acceptable and the calibration or update process ends. If the objective function value is not within the selected range, the forward model is run again (block 135) and a new objective function value is calculated. This process is repeated using the above-described inversion until the objective function value is minimized or within the selected range.” [Par 53] “The fluid composition data is used to perform a non-linear regression analysis to adjust the SRK equation of state. The basic objectives of using the non-linear regression in the fluid property model is to determine the optimum set of multipliers such that the observed or measured PVT data best match or fit as closely as possible to the calculated data from the SRK equation of state fluid model. Key parameters on which the regression is performed include any set of three independent variables. Examples of such variables include gas-oil ratio (GOR), oil density or API gravity, oil viscosity and bubble point pressure. The GOR, API gravity and bubble point pressure variables are chosen to minimize the objective functions with a value between +1%.”[Fig. 4, 6, and 7] Show aspects of calibration process, including an initial simulation/model run before gathering actual data, comparing the simulated data to measured data, adjusting simulation/model parameters based on this comparison, and rerunning the simulation/model, repeating the comparing, adjusting, and rerunning until the simulation/model output sufficiently matches the measured data.” [Examiner’s note: [Par 84] specifies that final data output is only provided “{w}hen satisfactory agreement is reached” between simulated and calculated values in relation to the associated threshold])
Craig makes obvious ([Col 30 line 27-48] “In some implementations, pressure history matching may also be used to refine a probability distribution for fracture parameters. In some implementations, in addition to comparing fracture pattern models to microseismic event data, formation pressures observed during an injection treatment are compared to formation pressures simulated using the fracture pattern model. For example, a fracture pattern models (e.g., "matches" or "mismatches") may be selected based on a correlation (or lack thereof) between observed formation pressure and simulated formation pressure. The observed formation pressure may be recorded during an injection treatment, and the fracture pattern model may be used to calculate a model formation pressure. Selecting fracture property values that minimize the difference between the observed formation pressure and the model formation pressure may lead to an improved distribution of fracture property values. For example, comparisons of surface pressure, bottomhole pressure, closure pressure, and/or net pressure (i.e., the difference between bottomhole pressure and closure pressure) can be used. A pressure matching technique may present graphical comparisons to a user (e.g., Cartesian, log-log, and/or other plots of observed pressure and model pressure versus time. A pressure matching technique may include an automated technique that calculates differences between observed and model formation pressures over time.”)
Li makes obvious([Fig. 9] Shows a comparison of the deviation between measured and simulated pressure and the deviation between the derivatives of the measured and simulated pressures)
PNG
media_image1.png
578
1557
media_image1.png
Greyscale
McLeod makes obvious wherein the indication of the tip screen-out event is provided; ([Abstract] “A method of determining when to stop pumping proppant during hydraulic fracturing in a wellbore is described. By accurately detecting tip screen-out with a bottom hole pressure gauge mounted to a perforating gun, the optimal amount of proppant can be supplied to a fracture while avoiding the risks associated with wellbore screen-out.”)
Claim 22. Bello teaches ([Par 70] “At block 84, pressure and temperature changes are calculated. At block 85, estimated parameters including production and/or injection rates are compared to pre-selected facility constraints (e.g., oil, gas and water processing capacity), and if the estimated parameters exceed such constraints, they are re-calculated.”)
Li makes obvious wherein the slope of actual net pressure values deviating from the slope of simulated net pressure values comprises an increase in the slope of actual net pressure values relative to the slope of simulated net pressure values. ([Fig. 13] Shows a situation where there is an in the slope of actual net pressure values relative to the simulated net pressure values for a portion of the time period)
PNG
media_image2.png
519
1361
media_image2.png
Greyscale
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Michael P Mirabito whose telephone number is (703)756-1494. The examiner can normally be reached M-F 10:30 am - 6:30 pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Emerson Puente can be reached at (571) 272-3652. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/M.P.M./ Examiner, Art Unit 2187
/JOHN E JOHANSEN/Examiner, Art Unit 2187
Prosecution Insights