Project Details

Data Analytics and End-to-End Machine Learning for Optimized Reservoir Geological Characterizations

Dates: 2022
Principal Investigator: Dr. Ardiansyah Koeshidayatullah
Description:

1. Data collection, conditioning and preprocessing from open-source subsurface datasets.

2. Apply exploratory data analysis, such as principal component analysis on the subsurface datasets to statistically select the              most important features and attributes for reservoir characterization and prediction.

3. Building supervised/unsupervised machine learning models to predict facies using a set of well log through different                        algorithms available in the public domain. These models will be generated using a limited number of wells as a pilot analysis at      the initial stage of the research. This will be then expanded to include more wells to evaluate performance of specific                        algorithms  at different scales.

4. Create synthetic geological datasets with generative adversarial networks to improve the variance and number of datasets.

5. Develop and evaluate novel deep learning models to determine the efficiency and accuracy of using automated models to              predict facies. This will be done using a set of different performance factors such as: confusion matrix, accuracy, recall, F1                score, root mean square error (RMSE) and mean absolute error (MAE).

6. Develop an end-to-end data analytics workflow for reservoir characterization and prediction

7. Deploying the proposed model to unseen, real-world datasets from different basins in Saudi Arabia and worldwide that could        provide insights to their application in an actual exploration and production processes.