[Paper Review] Machine Learning and the Internet of Things Enable Steam Flood Optimization for Improved Oil Production

Mi Yan, Jonathan C. MacDonald+4 more|arXiv (Cornell University)|Aug 29, 2019

Oil and Gas Production Techniques26 references4 citations

TL;DR

This paper proposes a machine learning and IoT-driven workflow for optimizing steam flood operations in heavy oil recovery, using real-time time-series data to forecast production and recommend optimal steam allocation. The system achieves a 3% increase in oil production through cloud-based model training and real-time data processing, demonstrating a scalable solution for industrial applications in energy and predictive maintenance.

ABSTRACT

Recently developed machine learning techniques, in association with the Internet of Things (IoT) allow for the implementation of a method of increasing oil production from heavy-oil wells. Steam flood injection, a widely used enhanced oil recovery technique, uses thermal and gravitational potential to mobilize and dilute heavy oil in situ to increase oil production. In contrast to traditional steam flood simulations based on principles of classic physics, we introduce here an approach using cutting-edge machine learning techniques that have the potential to provide a better way to describe the performance of steam flood. We propose a workflow to address a category of time-series data that can be analyzed with supervised machine learning algorithms and IoT. We demonstrate the effectiveness of the technique for forecasting oil production in steam flood scenarios. Moreover, we build an optimization system that recommends an optimal steam allocation plan, and show that it leads to a 3% improvement in oil production. We develop a minimum viable product on a cloud platform that can implement real-time data collection, transfer, and storage, as well as the training and implementation of a cloud-based machine learning model. This workflow also offers an applicable solution to other problems with similar time-series data structures, like predictive maintenance.

Motivation & Objective

• To address the inefficiencies of traditional physics-based steam flood simulations in heavy oil recovery.
• To develop a data-driven approach using machine learning and IoT for real-time forecasting and optimization of steam flood operations.
• To create a minimum viable product (MVP) on a cloud platform for scalable deployment of real-time data collection, storage, and model inference.
• To demonstrate the applicability of the framework beyond oil production, including predictive maintenance for similar time-series data.
• To improve oil production efficiency by recommending optimal steam allocation plans using supervised learning on time-series data.

Proposed method

• The method employs supervised machine learning on time-series data collected from IoT-enabled sensors in steam flood operations.
• A cloud-based platform is developed to handle real-time data ingestion, storage, and model training for the machine learning pipeline.
• The system uses historical operational data to train models that forecast oil production under varying steam injection conditions.
• An optimization module recommends optimal steam allocation plans by leveraging trained models to maximize oil output.
• The workflow integrates IoT data streams with machine learning models to enable dynamic, data-driven decision-making in real time.
• The framework is designed as a minimum viable product (MVP) for deployment on public cloud infrastructure, ensuring scalability and real-time performance.

Experimental results

Research questions

• RQ1 Can machine learning models trained on IoT-generated time-series data improve the accuracy of oil production forecasts in steam flood operations?
• RQ2 Can a data-driven approach outperform traditional physics-based simulation in optimizing steam flood performance?
• RQ3 To what extent can real-time data processing and cloud-based ML models enhance oil production through dynamic steam allocation?
• RQ4 How scalable and transferable is the proposed AIoT framework to other industrial applications with similar time-series data?
• RQ5 What is the measurable impact of optimized steam allocation on overall oil production yield?

Key findings

• The proposed machine learning and IoT-based system achieved a 3% improvement in oil production through optimized steam allocation.
• The framework successfully demonstrated real-time data collection, storage, and model inference using a cloud-based MVP.
• The system effectively forecasts oil production using supervised learning on time-series operational data from steam flood operations.
• The approach offers a viable alternative to classic physics-based simulations, with faster and more adaptive performance tuning.
• The workflow is transferable to other domains with similar time-series data, such as predictive maintenance in industrial systems.
• The integration of IoT and machine learning enables dynamic, data-driven decision-making that enhances recovery efficiency.