ProBHP is an AI-powered system for accurate flowing bottom hole pressure (FBHP) estimation and analysis in oil wells. It utilizes advanced machine learning and Explainable AI techniques to optimize production, monitor fluid movements, and assess reservoir performance. ProBHP provides cost-effective solutions for FBHP estimation in single-phase and multiphase flow scenarios, overcoming the limitations of conventional methods. With superior accuracy and real-time monitoring, ProBHP enhances wellbore management and production optimization.
The estimation of flowing bottom hole pressure (FBHP) in oil wells is crucial for monitoring fluid movements, optimizing production, quantifying reservoir performance, and understanding wellbore behavior. However, conventional methods for estimating FBHP are expensive and unreliable, especially when dealing with multiphase flow. Existing empirical correlations and mechanistic models developed in laboratory settings often yield inaccurate results when applied in the field. Additionally, obtaining FBHP using pre-installed permanent gauges in smart wells and well-testing analysis is costly and time-consuming, requiring constant calibration and maintenance. Therefore, there is a clear need for a more efficient and accurate approach.
To address the FBHP estimation problem, a dataset consisting of 206 data points was collected from various sources, including Govier and Fogarasi (1975) and Asheim (1986). These sources conducted BHP surveys by deploying down-hole pressure gauges just above the perforations to record the FBHP. The collected data serves as a valuable resource for developing and validating improved FBHP estimation methods.
In this project, I have developed a model capable of accurately predicting FBHP using a Feedforward Neural Network (FFNN). To optimize the FFNN and obtain the optimal hyperparameters, I employed Bayesian optimization, which proved to be highly efficient in implementation. This approach allowed me to fine-tune the network and achieve superior performance in FBHP estimation.
The developed model has been successfully deployed and is capable of performing both online and batch predictions. This means that it can provide real-time FBHP estimates during ongoing operations, as well as handle large datasets for retrospective analysis. The deployment of the model enables continuous monitoring of FBHP and enhances decision-making in oil and gas well operations.
| Notebook | Colab Link |
|---|---|
| Linear Regression & Decision Tree |  |
| FeedForwad Neural Network-Bayesian Optimization | |
To enhance the interpretability of the FBHP prediction model, I incorporated Explainable AI techniques using SHAP (SHapley Additive exPlanations) values. SHAP values provide insight into the contribution of each feature to the model's predictions. By understanding the feature importance, I gain valuable insights into why the model is making specific predictions.
The utilization of SHAP values allows me to interpret the results of the FBHP model accurately. It provides a transparent and understandable framework to comprehend the factors influencing FBHP and facilitates better decision-making in well operations.
This app allows you to input relevant parameters and receive accurate FBHP predictions in real-time or for batch analysis, depending on your requirements. The integration of Explainable AI ensures that you can also access feature importance information, gaining valuable insights into the model's decision-making process and enhancing your understanding of FBHP dynamics in oil wells.
Follow these steps to install and run the project locally:
Set up a virtual environment (optional but recommended):
python -m venv env
env\Scripts\activate.batgit clone https://github.com/ashrafalaghbari/ProBHP.git
cd <project-directory>
pip install -r requirements.txt
streamlit run app.pyAccess the web application by opening the following URL in your web browser:
http://localhost:8501Follow the instructions on the web application to use the project.
-
Prediction of pressure in different two-phase flow conditions: Machine learning applications
-
A new flowing bottom hole pressure prediction model using M5 prime decision tree approach
If you have any questions or encounter any issues running this project, please feel free to open an issue. I'll be happy to help!