pore pressure research gap presentation.pptx

research gap on pore pressure prediction

1. Contents Introduction Literature Review Research Gap Research Objective Methodology Timeline  References

2. Introduction Pore pressure refers to the pressure applied by fluids within the pores of a reservoir rock or sediment. It is a critical parameter in the exploration and production of hydrocarbons. Pore pressure is a crucial parameter in well design, well stability analysis, and mud program design, impacting the overall safety and efficiency of drilling operations. In a drilling operation pore pressure is regarded as a safe pressure only if the hydrostatic pressure of the drilling fluid in the wellbore falls between the formation pressure and formation fracture pressure

3. Introduction Overpressure (Abnormal High Pressure) Overpressure occurs when the pore pressure exceeds the normal hydrostatic pressure. This condition can arise due to various geological and mechanical processes, such as rapid sedimentation, fluid expansion, clay compaction, and hydrocarbon generation. Overpressure is a critical factor in drilling operations, as it can lead to well control issues, including kicks and blowouts. Underpressure (Abnormal Low Pressure) This occurs when the pore pressure is lower than the hydrostatic pressure. Underpressure can result from fluid withdrawal (such as oil or gas production), reservoir depletion, or structural changes in the formation. Underpressure can lead to issues like wellbore instability and formation collapse.

4. Introduction

5. Literature Review Author Year Topic Conclusion Goutami Das, Saumen Maiti 2023 A machine learning approach for the prediction of pore pressure using well log data of Hikurangi Tuaheni Zone of IODP Expedition 372, New Zealand The study uses well log data such as sonic travel time, density, gamma-ray, and porosity as inputs to the DTR model to predict PP.  The DTR model is built and compared to the empirical Eaton’s method for PP prediction. The DTR model achieves an R2 of 0.897 on the test dataset. The DTR model is able to identify an anomalous high-pressure zone below 145 mbsf. Hao Yu , G uoxiong Chen , Hanming Gu 2020 A machine learning methodology for multivariate pore-pressure prediction The paper introduces a machine learning-based method for predicting pore pressure using a nonparametric multivariate model that connects petrophysical properties (sonic velocity, porosity, and shale volume) to effective stress. Four different machine learning methods multilayer perceptron neural network (MLP), support vector machine (SVM), random forest (RF), and gradient boost machine (GBM) were applied to predict pore pressure for two offshore wells. Random forest outperforms the other algorithms in terms of goodness-of-fit, generalizability, and prediction accuracy.

6. Literature Review Author Year Topic Conclusion Guodao Zhang, Shadfar Davoodi , Shahab S. Band , Hamzeh Ghorbani, Amir Mosavi, Massoud Moslehpour 2022 A robust approach to pore pressure prediction applying petrophysical log data aided by machine learning techniques The paper aims to develop four intelligent predictive models - Random Forest (RF), Support Vector Regression (SVR), Artificial Neural Network (ANN), and Decision Tree (DT) - to predict pore pressure (PP) using petrophysical data. A feature selection analysis was conducted to identify the most influential 9 input variables for the PP prediction models, including corrected gamma ray, neutron porosity, photoelectric index, deep resistivity, shear-wave velocity, Laterolog shallow, spectral gamma-ray, and bulk density. Among the four models, the DT algorithm was found to outperform the others, achieving the highest prediction accuracy with an R2 of 0.9985 and RMSE of 14.460 psi Huayang Li, Jingen Deng, Baohong Dong, Bojia Li, Jinlong Guo 2023 A Comprehensive Prediction Method for Pore Pressure in Abnormally High-Pressure Blocks Based on Machine Learning The paper presents a comprehensive approach to predicting pore pressure in an abnormally high-pressure reservoir block using machine learning techniques. Conventional pore pressure prediction methods using empirical formulas have limitations, especially in complex geological settings with sparse data or recently established blocks. The paper constructs an intelligent pore pressure prediction model using four machine learning algorithms: KNN, Extra Trees, Random Forest, and LightGBM , LightGBM model outperforms the other three models, with prediction errors within ±10%, making it suitable for on-site pore pressure prediction.

7. Research gap Limited Data Representation: Study uses data from a single well U1517A ,need for a larger, more diverse dataset from multiple wells and different geological settings. Data Quality and Preprocessing Challenges: The quality of input data, including well logs and other geological data, can significantly impact the accuracy of machine learning models. Issues such as missing data, noise, and inconsistent measurement standards are not fully addressed. There is a need for advanced data preprocessing techniques, including data cleaning, normalization and imputation methods, to enhance the reliability of input data for machine learning models.

8. Research gap Model Validation and Comparison: Lack of comprehensive comparison with other machine learning models (e.g., Random Forest, SVM, Neural Networks). Limited Generalizability of Existing Methods: Many of the existing empirical methods are heavily reliant on specific geological formations, such as shale layers, which may not be as applicable or accurate in carbonate reservoirs. This indicates a gap in methods that are universally applicable across different geological settings. There is lack of supervised and potentially semi-supervised learning approach which can provide alternative innovative tools for pore pressure prediction.

9. Research objective The present study aims at the development of Machine learning model for pore pressure prediction using the well logs as the features and also with the pore pressure as the feature and then the comparison between different models. It aims at the development of a model which enables the ML model to give a fair prediction of Pore Pressure that can be used for different wells. Creation of a multiple ML models using well logs and pore pressure as features to forecast pore pressure. To compare between the various ML models using the statistical values to give out the best model The detection of overpressure zone with the analysis of different logs and then illustration for further analysis.

10. Methodology Data Collection: Data Sources: The study likely collected well log data, including sonic, density, and resistivity logs, along with information on geological formations. The data collection phase might also include pressure data from well tests and core samples. Data Preprocessing: This involves cleaning the data, handling missing values, normalizing, and possibly transforming the data to ensure consistency and reliability. Empirical Methods : Use empirical methods such as the Eaton method or Bowers method, which use well log data to estimate pore pressure. These methods typically rely on specific relationships between the geological data and pore pressure, tailored to particular rock types or conditions.

11. Methodology Machine Learning Techniques : Explore and Analyze Data: Perform exploratory data analysis (EDA) to understand the characteristics of the data and use statistical summaries and visualization to uncover patterns, correlations and insights. Feature Selection and Engineering: Select relevant features that will help improve the model’s performance. Split the Data: Divide the dataset into training and testing subsets (e.g., 80/20 split) or use cross-validation techniques to ensure the model is evaluated properly.

12. Methodology Choose and Train a Model: Choose appropriate machine learning algorithms (e.g., decision trees, support vector machines, neural networks). Adjust the model’s parameters to improve performance, often using techniques like grid search or random search. Evaluate the Model: The performance of the models is evaluated using metrics such as Mean Squared Error (MSE), R-squared, or other statistical measures of accuracy and reliability. Real World Application and Testing : The methodologies are tested on actual field data, possibly from different geological settings, to validate their applicability and accuracy in real-world scenarios.

13. References A machine learning approach for the prediction of pore pressure using well log data of Hikurangi Tuaheni Zone of IODP Expedition 372, New Zealand (Goutami Das, Saumen Maiti ) https://doi.org/10.1016/j.engeos.2023.100227 A machine learning methodology for multivariate pore-pressure prediction(HaoYu ,Guoxiong Chen,Hanming Gu) https://doi.org/10.1016/j.cageo.2020.104548 A robust approach to pore pressure prediction applying petrophysical log data aided by machine learning techniques ( Guodao Zhang , Shadfar Davoodi , Shahab S. Band , Hamzeh Ghorbani , Amir Mosavi , Massoud Moslehpour ) https://doi.org/10.1016/j.egyr.2022.01.012 A Comprehensive Prediction Method for Pore Pressure in Abnormally High-Pressure Blocks Based on Machine Learning (Huayang Li,Qiang Tan, Jingen Deng, Baohong Dong ,Bojia Li, Jinlong Guo,Shuiliang Zhang and Weizheng Bai) https://doi.org/10.3390/pr11092603