Reservoir Pressure Estimation via Eigen Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for assessing and managing well performance in hydrocarbon reservoirs are inadequate, particularly in complex reservoirs, as they lack robust and scalable techniques for estimating average reservoir pressure and well productivity index, leading to inefficiencies in reservoir management and delayed detection of performance degradation.
Innovation Solution
The implementation of machine learning-based methods, including dynamic mode decomposition (DMD), optimized DMD (optDMD), and sparse identification of nonlinear dynamics (SINDy), which utilize routine field measurements to estimate average reservoir pressure and predict pressures or flowrates, enabling automated and interpretable analysis for improved reservoir management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional manual interpretation methods are used for well productivity assessment, then interpretive analysis can be performed, but the methods do not scale for manual surveillance of large fields or wells with large data volumes
Solution Approach 1:
The patent replaces manual mechanical interpretation methods with automated machine learning algorithms. The system uses trained ML models to automatically analyze well performance data, estimate reservoir pressure, and detect productivity changes without requiring manual interpretation of well events, thereby scaling surveillance capabilities to handle large fields and data volumes efficiently
Solution Approach 2:
The system enables self-service through automated monitoring where the machine learning models continuously analyze well performance data and generate insights without human intervention. The automated detection of well performance degradation and estimation of reservoir characteristics allows the system to serve itself in monitoring and alerting, reducing the need for manual surveillance operations
2Adaptability or versatility
If automated machine learning methods are implemented for reservoir pressure estimation, then scalability to large fields is achieved, but the complexity of implementing and training ML models increases
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models using historical well performance data before deployment. The models are trained offline on representative datasets to learn patterns in reservoir pressure behavior and well responses, so that when deployed in production environments, they can automatically scale to handle large fields without requiring complex real-time training operations
Solution Approach 2:
The machine learning models are designed with universality to handle multiple surveillance tasks simultaneously. A single trained model framework can estimate reservoir pressure, detect well performance degradation, predict future productivity, and analyze various well event types across different field conditions, reducing the need for separate specialized models for each function
Data Source
AI summary
A method of determining average reservoir pressure for a subterranean formation is provided. The method includes shutting in a well in the subterranean formation; measuring a plurality of buildup pressures during a period of time wherein the well is shut in via a downhole pressure gauge; performing an eigen expansion of the plurality of buildup pressures to generate a plurality of eigenvalues; determining an estimated average reservoir pressure based, at least in part, on the plurality of eigenvalues; and producing fluids from the well based, at least in part, on the estimated average reservoir pressure.