Adaptive Reservoir Simulation Time-Stepping for Stable Convergence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current time-step selection methods in numerical simulation of dynamic systems, such as hydrocarbon reservoirs, are inefficient due to restrictive heuristic approaches, leading to wasted computational effort and longer simulation times, as they fail to optimally balance time-step size with stability and convergence requirements.
Innovation Solution
A machine learning-based time-step selection strategy that analyzes mathematical and physical system states to predict optimally sized time-steps, using AI and ML techniques to generate predictive models that adapt and improve over time, incorporating diverse physical and mathematical indicators for robust data-driven decision-making.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If heuristic time-step selection methods are used, then implementation is simple, but simulation efficiency is poor and computational effort is wasted
Solution Approach 1:
The system implements a feedback mechanism where the time-step selector receives information about the current simulation state (including convergence status and Newton iterations) and adjusts the time-step size dynamically. This closed-loop control enables the system to learn from past performance and optimize time-step selection, resolving the contradiction by making the selection process adaptive rather than static or purely heuristic.
Solution Approach 2:
The time-step selector is designed to autonomously determine optimal time-step sizes without requiring manual intervention or complex user tuning. By incorporating machine learning algorithms that automatically adapt to the simulation characteristics, the system serves itself in optimizing performance, thereby improving efficiency while keeping the interface simple for users.
2Speed
If large time-steps are used, then simulation speed increases, but stability and convergence cannot be guaranteed
Solution Approach 1:
The time-step size is transformed from a static or manually adjusted parameter into a dynamic variable that automatically adapts to changing simulation conditions. The selector monitors convergence behavior and adjusts time-step sizes in real-time, allowing the system to use larger steps when stable and smaller steps when convergence becomes problematic, thus maintaining both speed and reliability.
Solution Approach 2:
The system changes the time-step parameter dynamically based on simulation state indicators such as convergence status and Newton iteration counts. By continuously adjusting this critical parameter based on feedback from the simulation process, the system can push for larger time-steps to increase speed while having the safety mechanism to reduce them when reliability concerns arise.
3Reliability
If small time-steps are used, then stability is maintained, but simulation time increases significantly
Solution Approach 1:
Rather than using consistently small time-steps to guarantee stability, the system dynamically adjusts time-step sizes based on actual convergence behavior. This allows the simulation to use larger time-steps during stable phases (reducing total simulation time) while automatically switching to smaller steps when stability issues arise, thus resolving the contradiction between stability and time efficiency.
4Productivity
If PID controller is used for time-step selection, then computational efficiency improves, but tuning complexity increases
Solution Approach 1:
The time-step selector incorporates self-tuning capabilities through machine learning algorithms that automatically adapt to the specific characteristics of each simulation. This eliminates the need for manual PID parameter tuning by allowing the system to learn optimal control parameters from the simulation data itself, thereby maintaining computational efficiency while removing the tuning complexity burden from users.
Data Source
AI summary
Systems and methods are provided for modeling a reservoir. An exemplary method includes: receiving a reservoir model associated with a reservoir workflow process; modifying the reservoir model associated with the reservoir workflow process using an optimum time-step strategy; extracting features from the reservoir model along with first time-step sizes; generating a first set of data for devising a training set using the first time-step sizes; determining whether the selected amount of the first set of data reaches a predetermined level; triggering a real-time training using the training set and a machine learning (ML) algorithm; generating an ML model having second time-step sizes using the training set; selecting the first step-sizes or the second step-sizes based on the confidence level; sending the selected step-sizes to a simulator for processing; receiving results from the simulator that used the selected step-sizes; and determining whether results from the simulator require updating the training set.