Probabilistic Oil Production Forecasting with Uncertainty Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current decline curve analysis (DCA) methods for oilfield production forecasting face challenges due to sporadic and low-frequency historical production data, incorrect and missing measurements, subjective data interpretation, and the lack of robust algorithms to handle uncertainty and outliers.
Innovation Solution
A method that involves receiving historical well-production data, determining curve parameters while accounting for uncertainty, using multiple models to fit the data, calculating accuracy using information criteria that accounts for uncertainty, selecting the best model, and determining the estimated ultimate recovery (EUR) for the well.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single forecasting method is selected and run for all wells to save time, then productivity is improved, but measurement precision deteriorates because the method may not fit all data
Solution Approach 1:
The patent segments the forecasting process by evaluating multiple forecasting methods (e.g., exponential decline, hyperbolic decline, arithmetic decline) for each well and selecting the best-fit method individually. This segmentation allows each well to receive customized forecasting treatment rather than applying a single method universally, thereby maintaining forecast accuracy while managing computational complexity through automated model selection.
2Device complexity
If deterministic approach is used to calculate point estimates to simplify analysis, then device complexity is reduced, but reliability deteriorates because uncertainty associated with reservoir production behavior is not accounted for
Solution Approach 1:
The patent transitions from deterministic point estimates to probabilistic parameter estimation by incorporating uncertainty quantification. Instead of calculating single deterministic values for decline curve parameters, the system uses probabilistic methods to generate parameter distributions that reflect uncertainty in reservoir production behavior. This allows the system to maintain computational tractability while significantly improving forecast reliability through uncertainty-aware predictions.
3Ease of operation
If data preprocessing and outlier removal is performed subjectively by engineers, then ease of operation is improved, but measurement precision deteriorates due to engineer bias
Solution Approach 1:
The patent implements automated data preprocessing and outlier detection systems that perform data cleaning and validation without subjective engineer intervention. The system uses statistical methods and algorithms to automatically identify and handle outliers, missing data, and data quality issues, eliminating engineer bias while maintaining ease of operation. This self-service approach ensures consistent, objective data processing across all wells.
4Productivity
If entire production history is used for DCA parameter fitting regardless of assumption validity, then productivity is improved, but measurement precision deteriorates when data do not follow DCA assumptions
Solution Approach 1:
The patent dynamically evaluates the applicability of DCA assumptions for each well's production history before performing parameter fitting. The system uses diagnostic tests and validation procedures to determine whether the data follow DCA assumptions, and only applies DCA methods when assumptions are satisfied. For wells that do not meet assumptions, the system alternatively applies appropriate forecasting methods, ensuring parameter estimation accuracy while maintaining overall forecasting efficiency through automated method selection.
Data Source
AI summary
A method includes receiving historical well-production data, and determining curve parameters based on the historical well production data. Determining the curve parameters includes accounting for uncertainty. The method also includes determining curve fits for the historical data based on the curve parameters using a plurality of models, calculating accuracy for each of the models based on a comparison of the curve fits to the well-production data, comparing the accuracy for each of the models using an information criteria that accounts for uncertainty, selecting one of the models based on the comparison, and determining an estimated ultimate recovery (EUR) for the well using at least the selected one of the models.