Development of a Real-Time Reservoir Simulation and Digital Twin for Enhanced Oil Recovery under Uncertainty

Project Abstract

This study presents a novel framework for a real-time reservoir simulation coupled with a digital twin to optimize enhanced oil recovery (EOR) processes under reservoir uncertainty. The core objective is to integrate high-fidelity physics-based simulation models with data-driven analytics and real-time telemetry to enable rapid decision-making for field development and production optimization. The methodology leverages multi-physics reservoir modeling, data assimilation, uncertainty quantification, and machine learning to create a dynamic digital representation of an oil field that evolves with ongoing measurements, seismic updates, production history, and laboratory data. A modular workflow is proposed, where geological models, flow simulation, and surface network models are synchronized within a cloud-enabled platform, ensuring scalable computation and secure data exchange across stakeholders. Key components include (i) a compositional, multi-phase reservoir simulator capable of modeling complex fluid behaviors under EOR schemes such as polymer, surfactant, CO2, and thermal methods, (ii) a digital twin core that maintains a live replica of reservoir state variables, well controls, and equipment health, (iii) an advanced data assimilation system using ensemble Kalman filters, particle filters, and variational techniques to fuse noisy production data, downhole measurements, and seismic attributes, thereby reducing forecast uncertainty, (iv) uncertainty quantification workflows that propagate geological, petrophysical, and operational uncertainties to production forecasts, (v) a decision-support layer that generates near real-time optimization of injection strategies, well placement, and surface facility operations under risk constraints, and (vi) a verification and validation protocol employing history matching, cross-validation, and synthetic tests to ensure reliability under varying field scenarios. The research investigates the performance of the real-time reservoir simulation and digital twin in steering EOR under uncertainty, focusing on robustness, computational efficiency, and adaptability to data streams of varying quality. It evaluates the impact of updated geological models and sensor data on forecast accuracy and production optimization, and it examines how control strategies respond to uncertainties in reservoir properties, fluid contacts, and injection efficacy. The project also explores methods for reducing computational burden, including surrogate modeling, reduced-order representations, and parallelized algorithms, to achieve near-real-time turnaround for decision cycles. Expected outcomes include a validated framework that delivers accurate short- to medium-term forecasts, transparent quantification of uncertainty, and optimized operational plans that improve ultimate recovery while mitigating risk. The study contributes to the field by providing a replicable methodology for integrating real-time data with physics-based models in petroleum systems, enabling smarter field management, accelerated decision-making, and resilient performance in the face of incomplete information and dynamic reservoir behavior.

Project Overview

What This Project Is About

---

The project explores how to simulate oil reservoir behavior in real time and use a digital twin, a virtual replica of the reservoir, to improve oil recovery while accounting for uncertainty in data and conditions.

---

---

The Problem It Addresses

---

---

---

Objectives of the Project

---

1. Understand basic reservoir concepts and why real-time data matters.
2. Learn how to build a simplified simulation model of a reservoir.
3. Introduce the idea of a digital twin and how it mirrors the real system.
4. Demonstrate how uncertainty is represented and managed in models.
5. Develop a workflow to update simulations with new data.

---

---

What You Will Do Step by Step

---

1) Review key reservoir concepts and data types. 2) Create a simple model of a reservoir’s flow. 3) Build a digital twin framework and link it to real data streams. 4) Incorporate uncertainty using basic probabilistic methods. 5) Run case studies to show updates over time. 6) Compare decisions made with and without real-time updates. 7) Document lessons and limitations. 8) Prepare a final report and presentation.

---

---

Expected Outcome