Scribd Logo
Upload

Welcome to Scribd!

  • 50 views

    Paper Review: Machine-Learning-Based Petrophysical Property Modeling

    Uploaded by

    Thinh On
    This paper aims to develop a machine learning model to enable real-time reservoir management by updating porosity models in 3D static models. The model integrates seismic data with petrophysical properties. The authors collect and preprocess data, select relevant features, and build and compare multiple models to accurately model porosity using well data. The results are then validated against conventional geostatistical methods. However, the paper lacks organization and details about the models, features, hyperparameters and validation metrics, limiting its value for further research.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd

    Paper Review: Machine-Learning-Based Petrophysical Property Modeling

    Uploaded by

    Thinh On
    50 views1 page
    This paper aims to develop a machine learning model to enable real-time reservoir management by updating porosity models in 3D static models. The model integrates seismic data with petrophysical properties. The authors collect and preprocess data, select relevant features, and build and compare multiple models to accurately model porosity using well data. The results are then validated against conventional geostatistical methods. However, the paper lacks organization and details about the models, features, hyperparameters and validation metrics, limiting its value for further research.

    Original Description:

    AI, ML, paper review

    Original Title

    Paper_Review

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This paper aims to develop a machine learning model to enable real-time reservoir management by updating porosity models in 3D static models. The model integrates seismic data with petrophysical properties. The authors collect and preprocess data, select relevant features, and build and compare multiple models to accurately model porosity using well data. The results are then validated against conventional geostatistical methods. However, the paper lacks organization and details about the models, features, hyperparameters and validation metrics, limiting its value for further research.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    50 views1 page

    Paper Review: Machine-Learning-Based Petrophysical Property Modeling

    Uploaded by

    Thinh On
    This paper aims to develop a machine learning model to enable real-time reservoir management by updating porosity models in 3D static models. The model integrates seismic data with petrophysical properties. The authors collect and preprocess data, select relevant features, and build and compare multiple models to accurately model porosity using well data. The results are then validated against conventional geostatistical methods. However, the paper lacks organization and details about the models, features, hyperparameters and validation metrics, limiting its value for further research.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 1

    Paper Review: Machine-Learning-Based Petrophysical

    Property Modeling
    1. Paper summary:
    This paper aims to provide a method to overcome the challenges for developing real-time
    reservoir management, which is very difficult to implement because updating the reservoir
    requires plenty of time to be completed. More specifically, the model introduced in this paper
    enable a real-time or near-real-time decision making which helps update the porosity in the 3D
    static reservoir models. This machine learning model integrates 3D spatial availability of seismic
    data with petrophysical properties.
    The machine learning model in this paper complies with the normal process of building a
    machine learning model. The data is collect and QA/QC to remove the outliers in log data. The
    quality of seismic data is also evaluated. After evaluating the dataset, the authors select relevant
    features with input from geostatistician, which will be filtered again when combining statistics
    with science.
    The main part is model building, using algorithms for modeling porosity using the data at well
    locations. Then several models will be compared based on accuracy and computational
    efficiency.
    The last step is to validate the model. The results of the models will be compared to the results
    from the conventional geostatistical method.
    2. Review and Conclusion:
    The idea given in this paper is an innovative one, it enables real-time decision making, which is
    very important in the industry. The methodology is also standard and comprehensible, complies
    to the standard process of building a machine learning model. However, this paper is not well-
    organized. Firstly, the authors did not mention all the references that they used in their paper.
    Secondly, it would be more professional if the authors use the flowchart to describe the model A
    and model B.
    The results are relatively good for both model A and model B, and the authors also explained
    why their models cannot capture the sharp variations in the actual petrophysical log. However,
    the authors did not show the RMSE and correlation value for the model B. They could not
    explain the disparity between the predicted and actual porosity.
    The authors said that there are several hyper-parameters that need fine-tuning, but no detail is
    given about the hyper-parameters, what they are, and how they affect the model. The selected
    features are not listed neither, so the readers cannot have a concept what input the machine
    learning models capture. If these details are given, this paper could be an idea-prolific one for
    further researches.
    Although there are some relatively good results, this paper should not be published on the journal
    due to several above-mentioned limits.

    You might also like