A deep learning-based direct forecasting of CO2 plume migration

https://doi.org/10.1016/j.geoen.2022.211363Get rights and content

Highlights

  • A deep learning-based latent space mapping framework is developed to forecast CO2 plume evolution in geological reservoirs.

  • The method maps the plume into a latent space and establishes the observation-prediction relationship for direct forecast.

  • The method accurately predicts the CO2 plume migration patterns with high computational efficiency.

  • This forecasting framework avoids computationally expensive history matching and provides timely decision-making support.

Abstract

Accurate and timely forecasts of CO2 plume evolution in geological reservoirs are crucial for CO2 migration detection, leakage risk assessment, and operation decision support. Conventional forecasting usually adopts a two-step strategy, first calibrating reservoir model parameters against observations using iterative inverse modeling (or history matching) and then applying the calibrated model for predictions. This method impedes real-time forecasts due to the heavy computational demand in inverse modeling and may suffer from poor prediction accuracy because of the limited observation data. In this work, we propose a deep learning-based latent space mapping framework to forecast CO2 plume migration directly by avoiding the inverse modeling. We first use the convolutional autoencoder to map the high-dimensional complex plume extents onto low-dimensional latent space. Next, we use neural networks to learn the relationship between the observation variables and the prediction latent variables. And then for given observation data, we infer the prediction values directly. This one-step direct forecasting is computationally efficient which requires a few number of parallelizable reservoir simulations and it can provide accurate predictions with limited observations by learning the observation-prediction relationship in the reduced dimension. Therefore, our proposed method enables an in-time forecast of dynamic CO2 plume distributions. We demonstrate the effectiveness and accuracy of our method in predicting the CO2 plume migration using four metrics such as plume area, centroid movement distance, and plume spreading in the primary and secondary directions. And the spatio-temporal evolution patterns of plume migration under diverse geological complexities are also accurately quantified.

Section snippets

Introduction and motivation

Geologic carbon capture and storage technology has been considered as one of the promising technologies to mitigate the long-term CO2 concentration in the atmosphere as well as to accomplish the objective of decreasing greenhouse gas emission and limiting the global temperature exceeding to a threshold of 2° by 2100 (Alcalde et al., 2018; Viebahn et al., 2015; IEA, 2012). This technology involves collecting CO2 from large fossil fuel usage emitting sources such as power plants and subsequent

Materials and methods

In this section, we first describe the data generation. Next, we introduce two key aspects of the developed framework in effectively predicting the CO2 plume migration. The deep convolutional autoencoder is adopted to perform data parameterization by mapping the high-dimensional plume extent onto a low-dimensional latent space. Then, the non-linear relationship between observation variables and dimension-reduced prediction variables is established using a neural network. Lastly, we define four

Results

In this section, we first use the four defined metrics to evaluate the accuracy, effectiveness, and efficiency of the developed framework to forecast the high-dimensional CO2 plume distribution with streaming observation variables in three dramatically different scenarios. Then we investigate the influence of the parameterization level on the forecasting performance of the developed framework. We compare with PCA-based parameterization to evaluate whether the convolutional autoencoder can map

Discussion

To accelerate real-time decision making for carbon storage, we propose an inversion-free direct forecasting framework to reduce computational costs in traditional inversion-based model calibration (history matching). In this framework, we learn the observation-prediction relationship in their low-dimensional latent space and then infer the prediction from the observation data based on the learned ML model. We train the ML model using physics-based model simulation samples, so the observation

Conclusions and future work

Because of the high heterogeneity in the CO2 deposition environment, the CO2 plume shapes are complex and it is difficult to delineate their movement over space and time. In this work, we develop a deep learning-based framework to fast and accurately forecast the CO2 plume migration directly from the streaming observation data. The framework is tested against three dramatically different scenarios to demonstrate its robustness and effective-ness in quantifying the shape and spatio-temporal

Credit author statement

Ming Fan implemented the numerical experiments, prepared the figures and analyzed the results. Dan Lu formulated the problem, contributed to the research plan, and interpreted the results. Siyan Liu interpreted the results. All the three authors contributed to the manuscript preparation.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

Primary funding support for this work is provided by the Science-informed Machine Learning to Accelerate Real Time Decision Making for Carbon Storage (SMART-CS) Initiative, funded by the US Department of Energy (DOE), Office of Fossil Energy and Carbon Management. Additional support is from the Artificial Intelligence Initiative as part of the Laboratory Directed Research and Development Program of Oak Ridge National Lab- oratory, managed by UT-Battelle, LLC, for the US DOE under contract

References (53)

  • K. Hornik et al.

    Multilayer feedforward networks are universal approximators

    Neural Network.

    (1989)
  • E. Laloy et al.

    Inversion using a new low-dimensional representation of complex binary geological media based on a deep neural network

    Adv. Water Resour.

    (2017)
  • D. Lu et al.

    An efficient bayesian data-worth analysis using a multilevel Monte Carlo method

    Adv. Water Resour.

    (2018)
  • K. Michael et al.

    Geological storage of co2 in saline aquifers—a review of the experience from existing storage operations

    Int. J. Greenh. Gas Control

    (2010)
  • R. Pawar et al.

    Quantification of risk profiles and impacts of uncertainties as part of us doe's national risk assessment partnership (nrap)

    Energy Proc.

    (2013)
  • R.J. Pawar et al.

    Recent advances in risk assessment and risk management of geologic co2 storage

    Int. J. Greenh. Gas Control

    (2015)
  • H. Tang et al.

    A deep learning-accelerated data assimilation and forecasting workflow for commercial-scale geologic carbon storage

    Int. J. Greenh. Gas Control

    (2021)
  • P. Viebahn et al.

    Prospects of carbon capture and storage (ccs) in China's power sector–an integrated assessment

    Appl. Energy

    (2015)
  • H. Wang et al.

    Deep-learning-based workflow for boundary and small target segmen- tation in digital rock images using unet++ and Ik-ebm

    J. Petrol. Sci. Eng.

    (2022)
  • M. Abadi et al.

    Tensorflow: Large-Scale Machine Learning on Heterogeneous Systems

    (2015)
  • J. Alcalde et al.

    Estimating geological co2 storage security to deliver on climate mitigation

    Nat. Commun.

    (2018)
  • R. Aris

    On the dispersion of a solute in a fluid flowing through a tube

    Proc. R. Soc. London, A

    (1956)
  • Y. Bengio et al.

    Representation learning: a review and new perspectives

    IEEE Trans. Pattern Anal. Mach. Intell.

    (2013)
  • A. Bianco et al.

    A real field application

  • D.A. Cameron

    Optimization and Monitoring of Geological Carbon Stor- Age Operations

    (2013)
  • M. Fan et al.

    Interaction between prop- pant compaction and single-/multiphase flows in a hydraulic fracture

    SPE J.

    (2018)
  • Cited by (4)

    This manuscript has been authored by UT-Battelle LLC, under contract DE-AC05- 00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or repro-duce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public- access-plan).

    View full text