IP Library › Granted Patent US 12,380,261
Granted Patent B2
US 12,380,261 · App. 17/358,447 · Granted Aug 5, 2025

Apparatus and method for oil production forecasting

Inventors: Amos Ben-Zvi (Calgary, CA); Seyide Hunyinbo (Calgary, CA); Juliana Y. Leung (Calgary, CA); Prince Azom (Calgary, CA)
Assignee: Cenovus Energy Inc.
G06F30/27E21B43/128E21B43/24E21B47/008E21B47/06E21B49/00G06N20/20E21B2200/20G06F2111/08G06N5/01
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,380,261
App. No.
17/358,447
Granted
Aug 5, 2025
Kind
B2
Abstract

A method and apparatus for forecasting oil production from an oil well in a geological formation includes receiving a plurality of sets of predicted geological data, for each of the plurality of sets of predicted geological data, determining a probability for the predicted geological data of the formation, iteratively selecting one of the plurality of sets of predicted geological data using Monte Carlo sampling based on the determined probabilities, assigning the selected set of predicted geological data to a cluster of historical data, and for each set of historical data of the cluster generating a predicted oil production rate as a function of time utilizing a machine learning based oil model, generating, based on the predicted oil production rates, a forecasted oil production rate, determining, based on the forecasted oil production rate, a preferred operating parameter for the well, and operating based on the preferred operating parameter.

Claims (70)

1. A method of producing oil in a steam-assisted gravity drainage (SAGD) system that includes at least one injector well and at least one producer well in a geological formation, the method comprising:

i) training a machine learning based model to generate and output a predicted oil production rate as a function of time given predicted geological data for the geological formation and historical data as inputs, wherein the historical data comprises geological data and historical operational data from other geological formations geologically similar to the geological formation;

ii) receiving a plurality of sets of predicted geological data for the geological formation and sets of historical data comprising the historical operational data, and clustering the sets of historical data into clusters based on geological similarity;

iii) for each given set of predicted geological data of the plurality of sets of predicted geological data, determining a probability associated with the predicted geological data of the given set;

iv) repeating for a number of iterations:

a) selecting one set of the plurality of sets of predicted geological data using Monte Carlo sampling based on a probability distribution defined by probabilities for the plurality of sets of predicted geological data as determined in iii);

b) assigning the selected one set of predicted geological data of a) to one cluster of the clusters of the historical data based on similarity of the one set of predicted geological data to geological data in the historical data in the one cluster;

c) extracting the historical operational data and the geological data from the one cluster of historical data of b); and

d) using the historical operational data and the geological data extracted from the one cluster of historical data in c) as input to the machine learning based model trained in i), which generates and outputs a predicted oil production rate as a function of time through the at least one producer well given the historical operational data and the geological data as inputs;

v) generating a forecasted oil production rate at a particular time and an associated uncertainty from the predicted oil production rate of d) over the number of iterations of iv); and

vi) operating the at least one injector well and the at least one producer well in the geological formation using at least one operational parameter determined from the forecasted oil production rate at the particular time and the associated uncertainty of v);

wherein the operating of vi) includes injecting steam through the at least one injector well into the geological formation and producing oil extracted from the geological formation through the at least one producer well, and

wherein the at least one operational parameter of vi) is selected from the group consisting of injector well bottom hole pressure, steam injection rate, or other operational parameter related to production of oil from the geological formation.

2. The method according to claim 1 , further comprising:

generating a predicted injectant rate as a function of time for the selected set of predicted geological data by inputting into a machine learning based injectant model geological data included in the selected set of predicted geological data and the historical operational data from the historical operational data; and

generating the predicted oil production rate as a function of time by inputting the predicted injectant rate as a function of time into the machine learning based model.

3. The method according to claim 2 , further comprising:

generating, based on the predicted injectant rates as a function of time, a forecasted injectant rate as a function of time, where at a particular time the forecasted injectant rate is determined based on the predicted injectant rates at the particular time, and the forecasted injectant rate at the particular time having an uncertainty defined by a range of predicted injectant rate values at the particular time.

4. The method of claim 1 , further comprising:

receiving measured data from the well up to time t, the measured data including measured operational data and measured oil production rate data;

wherein determining the probability associated with the predicted geological data for each given set of predicted geological data of the plurality of sets of predicted geological data at iii) comprises:

inputting the measured operational data into the machine learning based model to generate a predicted measured oil production rate as a function of time up to time t; and

determining the probability associated with the predicted geological data based on a difference between the predicted measured oil production rate and the measured oil production rate.

5. The method of claim 4 , wherein generating the predicted oil production rate, for each set of historical geological data of the assigned cluster to which the selected set of predicted geological data belongs, comprises:

generating the predicted oil production rate as a function of time for times greater than time t by inputting into the machine learning based model geological data included in the selected set of predicted geological data and historical operational data from the set of historical data and utilizing the predicted measured oil production rate as the predicted oil production rate for times up to time t.

6. The method of claim 5 , further comprising:

determining that the measured oil production rate for a particular time is outside of the uncertainty of the forecasted oil production rate; and

in response to determining that the measured oil production rate for the particular time is outside of the uncertainty of the forecasted oil production rate, transmitting an alert.

7. The method of claim 5 , further comprising:

inputting measured operational data that includes measured injectant rate as a function of time up to time t into a machine learning based injectant model to generate a predicted injectant rate as a function of time up to time t;

wherein generating the predicted oil production rate, for each set of historical geological data of the assigned cluster to which the selected set of predicted geological data belongs, comprises:

generating a predicted injectant rate as a function of time for the selected set of predicted geological data for time greater than time t by inputting into the machine learning based injectant model geological data included in the selected set of predicted geological data and historical operational data from the historical data; and

generating the predicted oil production rate as a function of time by inputting the predicted injectant rate as a function of time for times up to time t and times greater than time t into the machine learning based model.

8. The method of claim 7 , further comprising:

generating, based on the predicted injectant rates as a function of time, a forecasted injectant rate as a function of time, where at a particular time the forecasted injectant rate is determined based on the predicted injectant rates at the particular time, and the forecasted injectant rate at the particular time having an uncertainty defined by a range of predicted injectant rate values at the particular time;

determining that the measured injectant rate as a function of time for a particular time is outside of the uncertainty of the forecasted injectant rate for the particular time; and

in response to determining that the measured injectant rate as a function of time for a particular time is outside of the uncertainty of the forecasted injectant rate, transmitting an alert.

9. The method of claim 7 , further comprising:

generating, based on the predicted injectant rates as a function of time, a forecasted injectant rate as a function of time, where at a particular time the forecasted injectant rate is determined based on the predicted injectant rates at the particular time, and the forecasted injectant rate at the particular time having an uncertainty defined by a range of predicted injectant rate values at the particular time;

wherein the at least one operational parameter of vi) is determined by performing an optimization scheme utilizing the forecasted injectant rate as a function of time and the forecasted oil production rate as a function of time.

10. The method of claim 9 , wherein:

the at least one operational parameter of vi) is injector well bottom hole pressure; and

the optimization scheme includes an economic objective function determination based on the cost of steam injected into the at least one injector well in accordance with the forecasted injectant rate as a function of time and the value of the oil produced from the at least one producer well in accordance with the forecasted oil production rate.

11. The method of claim 1 , wherein, prior to receiving any measured data from the SAGD system, determining, for each given set of predicted geological data of the plurality of sets of predicted geological data, a probability based on a ratio of the number of sets of historical data that is included within a type associated with the given set of predicted geological data to the total number of sets of historical data.

12. The method of claim 1 , wherein the machine learning based model utilizes random forest machine learning.

13. The method of claim 1 , wherein each given set of the plurality of sets of predicted geological data of the formation includes at least one of:

rich pay thickness;

non-rich pay thickness;

bottom water thickness;

rich vertical permeability;

non-rich vertical permeability;

rich porosity;

non-rich porosity;

bottom water porosity;

rich oil saturation;

non-rich oil saturation;

bottom water saturation; and

permeability of bottom water zone.

14. The method of claim 1 , wherein the operational data extracted from the sets of historical data includes at least one of:

production start data;

electric submersible pump speed; and

injector bottom hole pressure.

15. The method of claim 1 , further comprising:

updating the forecasted oil production rate at a particular time and associated uncertainty of v) using actual data received from the SAGD system.

16. The method of claim 1 , wherein the plurality of sets of predicted geological data of ii) includes one or more of P10 to P95 data sets.

17. The method of claim 1 , wherein the at least one operational parameter of vi) includes a dynamic variable that is input to the machine learning based model that generates and outputs the predicted oil production rate.

18. The method of claim 1 , wherein the at least one operational parameter of vi) is determined by an optimizing routine that maximizes an objective function related to production of oil from the geological formation.

19. The method of claim 18 , wherein the objective function is an economic objective function that relates to value of oil produced from the geological formation and costs to produce oil from the geological formation.

20. The method of claim 1 , wherein the at least one operational parameter of vi) is determined by an optimizing routine that involves a genetic algorithm in which solutions are represented as chromosomes which together form a population, wherein the population is initialized randomly.

21. The method of claim 1 , wherein the at least one operational parameter of vi) includes an uncertainty.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 30, 2025
From: BEN-ZVI, AMOS; HUNYINBO, SEYIDE; LEUNG, JULIANA Y.; AZOM, PRINCE
To: CENOVUS ENERGY INC.
Reel/Frame 071562/0395 →
Continuity (1)
Related Publication 20220414299A1 · Dec 29, 2022
References Cited (42)
US 20140365409A1 · Burch · 2014 [cited by examiner]
US 20160312552A1 · Early · 2016 [cited by examiner]
US 20170177992A1 · Klie · 2017 [cited by examiner]
US 20180087358A1 · Conn · 2018 [cited by examiner]
US 20180087359A1 · Conn · 2018 [cited by examiner]
US 20200284944A1 · Madasu · 2020 [cited by examiner]
US 20220205354A1 · Chaki · 2022 [cited by examiner]
Almashan et al., “Estimating Pvt Properties of Crude Oil Systems Based on a Boosted Decision Tree Regression Modelling Scheme With K-means Clustering,” Paper presented at the SPE/IATMI Asia Pacific Oil & Gas Conference … [cited by applicant]
Amirian et al., “Integrated Cluster Analysis and Artificial Neural Network Modeling for Steam-assisted Gravity Drainage Performance Prediction in Heterogeneous Reservoirs, Expert Systems With Applications,” 2014, vol. 4… [cited by applicant]
Aulia et al., “A New History Matching Sensitivity Analysis Framework With Random Forests and Plackett-burman Design,” SPE Symposium: Production Enhancement and Cost Optimisation, 2017. [cited by applicant]
Bieker et al., “Optimal Well-testing Strategy for Production Optimization: a Monte Carlo Simulation Approach,” Paper Presented at the Spe Eastern Regional Meeting, 2006. [cited by applicant]
Breiman et al., “Random Forests,” Machine Learning, 2001, vol. 45, pp. 5-32. [cited by applicant]
Butler and Stephens., “The Gravity Drainage of Steam-heated Heavy Oil to Parallel Horizontal Wells,” Journal of Canadian Petroleum Technology, 1981, vol. 20 (2), pp. 90-96. [cited by applicant]
Cenovus Energy Inc, (2018). Christiana Lake in-situ Progress Resport. https://www.aer.ca/providing-information/data-and- reports/activity-and-data/in-situ-performance-presentations. [cited by applicant]
Cenovus Energy Inc, 2019 Investors Day Presentation, Oct. 2, 2019. [cited by applicant]
Coimbra et al., “Practical Application of Pareto-based Multi-objective Optimization and Proxy Modeling for Steam Alternating Solvent Process Design,” Paper presented at the SPE Western Regional Meeting, 2019. [cited by applicant]
Dehdari et al., “Calibrating a Semi-Analytic SAGD Forecasting Model to 3D Hetetogenous Reservoir Simulations,” SPE Canada Heavy Oil Technical Conference Calgary, 2017. [cited by applicant]
Gilbert et al., “Reservoir Modeling: Integrating Various Data at Appropriate Scales,” The Leading Edge, Aug. 2004, vol. 23 (8), pp. 784-788. [cited by applicant]
Gomez and Babadagli., “High-Temperature Solvent Injection for Heavy-Oil Recovery From Oil Sands: Determination of Optimal Application Conditions Through Genetic Algorithm,” SPE Reservoir Evaluation and Engineering, May … [cited by applicant]
Gosayir et al., “Design of Solvent-assisted Sagd Processes in Heterogeneous Reservoirs Using Hybrid Optimization Techniques,” Presented at the Canadian Unconventional Resources Conference, 2011, vol. 2, pp. 1582-1593. [cited by applicant]
Guevara et al., “Optimization of Steam Injection for Heavy Oil Reservoirs Using Reinforcement Learning,” In: In SPE International Heavy Oil Conference and Exhibition. Society of Petroleum Engineers, 2018. [cited by applicant]
Kubota et al., “Machine Learning Forecasts Oil Rate in Mature Onshore Field Jointly Driven by Water and Steam Injection the Mature Field Under Study : Oil Reservoir Under Simultaneous Water and Steam Injection,” SPE Ann… [cited by applicant]
Liao et al., “Data Mining: A Novel Strategy for Production Forecast in Tight Hydrocarbon Resource in Canada by Random Forest Analysis,” Institute of Petroleum Technology Conference, 2020. [cited by applicant]
Ma and Leung., “Design of Warm Solvent Injection Processes for Heterogeneous Heavy Oil Reservoirs: A Hybrid Workflow of Multi-Objective Optimization and Proxy Models,” Society of Petroleum Engineers—SPE Reservoir Simula… [cited by applicant]
Ma and Leung., “Integration of Deep Learning and Data Analytics for SAGD Temperature and Production Analysis,” Computational Geosciences, 2020, vol. 24, pp. 1239-1255. [cited by applicant]
Ma and Leung., “Knowledge-based Heterogeneity Characterization Framework for 3d Steam-assisted Gravity Drainage Reservoirs,” Knowledge-Based Systems, 2020, vol. 192, 19 pages. [cited by applicant]
Ma and Leung., “Practical Implementation of Knowledge-based Approaches for Steam- Assisted Gravity Drainage Production Analysis, ” Expert Systems with Applications, 2015, vol. 42 (21), pp. 7326-7343. [cited by applicant]
Ma et al., “Integration of Data-driven Modeling Techniques for Lean Zone and Shale Barrier Characterization in SAGD Reservoirs,” Journal of Petroleum Science and Engineering, 2016, vol. 176, pp. 716-734. [cited by applicant]
Mathworks, (2020). Global Optimization Toolbox™ User's Guide. https://www.mathworks.com/help/releases/R2020b/pdf_doc/gads/gads.pdf. [cited by applicant]
Mathworks, (2020). Statistics and Machine Learning Toolbox™ User's Guide. https://www.mathworks.com/help/releases/R2020b/pdf_doc/stats/stats.pdf. [cited by applicant]
Mehana et al., “Integrating Production Analysis With Monte Carlo Simulation for Estimated Ultimate Recovery EUR Prediction,” Paper Presented at the SPE Eastern Regional Meeting, Oct. 2019, pp. 15-17. [cited by applicant]
Millington et al., “Canadian Oil Sands Supply Costs and Development Projects (2015-2035),” Canadian Energy Research Institute (CERI). Study No. 152, Aug. 2015. [cited by applicant]
Mudhafar and Rao., Integrating Design of Experiments, Proxy Modeling, and Monte-Carlo Simulation for Combined Uncertainty Quantifications of Geological and Production Data in the Cyclic GAGD Process, Oct. 2016. [cited by applicant]
Muhammad et al., “Trees Vs Neurons: Comparison Between Random Forest and ANN for High-resolution Prediction of Building Energy Consumption,” Energy and Buildings, Jul. 2017, vol. 145, pp. 77-89. [cited by applicant]
National Energy Board (2006). Canada's Oil Sands: Opportunities and Challengesto 2015: An Update, An Energy Market Assessment. Calgary, Alberta; 2006. [cited by applicant]
National Energy Board, (2020). Canada's Energy Future 2017 Supplement: Oil Sands Production. https://www.cer-rec.gc.ca/en/data-analysis/canada-energy-future/2017-oilsands/index.html. [cited by applicant]
Olalotiti-Lawal and Datta-Gupta., “A Multi-objective Markov Chain Monte Carlo Approach for History Matching and Uncertainty Quantification,” Paper Presented at the SPE Annual Technical Conference and Exhibition, Journal… [cited by applicant]
Russel et al., “Artificial Intelligence—A Modern Approach,” Third Edition-Person Education, 2016. [cited by applicant]
Sibaweihi et al., “Real-time Steam Allocation Workflow Using Machine Learning for Digital Heavy Oil Reservoirs,” Journal of Petroleum Science and Engineering, Apr. 2021, vol. 199, 22 pages. [cited by applicant]
Sivanandam et al., “Introduction to Genetic Algorithms,” Springer-Verlag Berlin Heidelberg, 2008. [cited by applicant]
Zheng et al., “A Proxy Model for Predicting SAGD Production From Reservoirs Containing Shale Barriers,” Journal of Energy Resources Technology, Transactions of the ASME, 2018, vol. 140 (12), 10 pages. [cited by applicant]
Zheng et al., “An Ai-based Workflow for Estimating Shale Barrier Configurations From SAGD Production Histories,” Neural Computing and Applications, 2019, vol. 31, pp. 5273-5297. [cited by applicant]
Cited By (1)
US 12,704,064