Determination of location and type of reservoir fluids based on downhole pressure gradient identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for determining pressure gradients in wellbores are susceptible to outliers, local optimality, and fail to scale well with discontinuous fluid barriers, leading to ambiguity in identifying reservoir fluids and their types.
Innovation Solution
A meta-heuristic process, such as simulated annealing, is used to partition the depth range into fluid depth ranges, combined with a constrained multi-gradient fitting method to construct fluid gradients, which inherently handles outliers and adheres to physical constraints, providing robust and scalable solutions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional methods are used to determine pressure gradients, then the process is simple, but the interpretation is ambiguous and not robust against measurement errors
Solution Approach 1:
The patent replaces conventional mechanical/mathematical gradient calculation methods with a machine learning-based system. The ML model is trained on synthetic pressure data generated from reservoir simulation, enabling it to robustly interpret pressure gradients while accounting for measurement errors and outliers without relying on traditional mathematical approaches.
Solution Approach 2:
The patent performs preliminary training of the machine learning model using synthetic pressure data generated from reservoir simulation before actual field application. This pre-training phase creates a robust model that can handle measurement errors and outliers, eliminating the need for manual data cleaning and preparation during actual gradient determination.
2Measurement precision
If manual outlier removal is performed, then measurement precision improves, but time consumption increases
Solution Approach 1:
The machine learning model performs self-service by automatically identifying and handling outliers during the gradient interpretation process. The model was trained on synthetic data that includes various error conditions, enabling it to autonomously distinguish between valid measurements and outliers without requiring manual intervention or time-consuming data cleaning steps.
Solution Approach 2:
The patent replaces manual outlier removal procedures with an automated machine learning-based filtering system. The ML model inherently handles outliers during prediction, substituting the manual inspection and removal process with an automated computational approach that maintains precision while eliminating time loss.
3Loss of information
If comprehensive pressure sampling is conducted, then reservoir characterization improves, but measurement cost and complexity increase
Solution Approach 1:
The patent applies partial action by using the machine learning model to determine whether additional pressure measurements are necessary. The model can provide reliable reservoir characterization with minimal sampling by identifying the most informative measurement locations, avoiding the need for comprehensive extensive sampling while maintaining characterization completeness.
Solution Approach 2:
The machine learning model provides feedback on the quality and sufficiency of pressure measurements for reservoir characterization. By analyzing the pressure gradient data and uncertainty, the model can indicate whether additional measurements are needed or if the current data set is sufficient, enabling optimized sampling strategies that reduce complexity while maintaining information completeness.
Data Source
AI summary
A method comprises receiving a measurement of a pressure in a subsurface formation at a number of depths in a wellbore formed in the subsurface formation across a sampling depth range of the subsurface formation to generate a number of pressure-depth measurement pairs. The method comprises partitioning the sampling depth range into a number of fluid depth ranges, wherein each of the number of fluid depth ranges comprises a range where a type of reservoir fluid is present in the subsurface formation. The method comprises determining a fluid gradient for the type of the reservoir fluid for each of the number of fluid depth ranges.