Extending the Lifespan of Geothermal Power Plants: How Artificial Intelligence Is Replacing Traditional Simulations

Hello, readers! Welcome to my page, I am Olvandri Nouva Pratama. I am a sixth semester student in Institut Teknologi Bandung at Bandung, Indonesia. This medium was written to fulfill a research-based learning assignment for the Physics program course titled “Data Analysis with Machine Learning.” This assignment was assigned in lieu of the final exam and is based on material covered in previous classes.

So, the project I proposed to work on is included in few points:

• Tentative Title

“Prediction of Thermal Breakthrough Probability and Spatial Optimization of Geothermal Injection Well Placement Using Machine Learning Algorithms”

• Problem to solve

A little bit about this topic, In geothermal reservoir management, the process of reinjecting cool fluid back into the subsurface is absolutely essential for maintaining reservoir pressure and extending the operational life of the power plant. However, imprecise placement of injection wells can trigger thermal breakthrough, which is the premature cooling of production wells due to the migration of injected fluid. Traditionally, modeling of thermal and fluid mechanics propagation has been performed using numerical simulations that require significant computational time and are costly. On the other hand, modern geothermal fields generate massive volumes of sensor data every second.

So, the Problem Solved, This project aims to bridge the principles of thermodynamics with large-scale data analytics. Instead of relying solely on numerical simulations, this project will build a predictive machine learning pipeline. By leveraging feature extraction from historical data (flow rates, temperature gradients, and spatial distances), XGBoost and ANN models will be trained to instantly predict high-probability zones for breakthroughs. This solution will provide recommendations for the most optimal distances and locations for new injection well drilling operations without compromising energy production capacity.

• Programming language, libraries, and platform i will use

Programming language: Python

Libraries:

1. Data Manipulation: pandas, numpy
2. Classic Machine Learning: scikit-learn (for PCA, kNN, SVM, and MLP/ANN)
3. Gradient Boosting: xgboost (for spatial probability predictions)
4. Visualization: matplotlib, seaborn (for heatmaps and evaluation metrics)

Platform to use: Visual Studio Code

• Timeline of the Work

Fig. 1 Timeline.

• Estimation number of team member

The team required to complete this project consists of 1–2 people, with the following division of tasks:

Member 1: Focus on data cleaning (preprocessing) and building basic classification models (kNN & SVM).

Member 2: Focus on complex regression modeling (ANN), spatial inference using XGBoost, and designing 2D/3D contour visualizations.

If there is 2 member to solve this project, maybe it only 2 weeks to finish this project.

• Link to Datasets

1. https://openei.org/wiki/Wells_General
2. https://gdr.openei.org/submissions/1326

• Mockup the output using AI

output code from AI

output code from AI(2)

• Link to the conversation with AI

https://gemini.google.com/share/a4101e1473cf

• References:

1. Muther, T., Syed, F. I., & Dahaghi, A. K. (2022). Machine learning applications in geothermal energy development: A comprehensive review. Geothermics, 105, 102525.
2. Sircar, A., Yadav, K., Rayavarapu, K., Bist, N., & Oza, H. (2021). Application of machine learning and artificial intelligence in oil and gas industry. Petroleum Research, 6(4), 379–391.
3. Arzola, J., & Horne, R. N. (2020). Application of machine learning to geothermal reservoir performance prediction. PROCEEDINGS, 45th Workshop on Geothermal Reservoir Engineering. Stanford University, Stanford, California.
4. Tutuncu, A. N., & Miskimins, P. T. (2019). Data-driven surrogate models for predicting thermal breakthrough in geothermal systems. Journal of Energy Resources Technology, 141(8), 082005.
5. Nwachukwu, A., Jeong, H., Pyrcz, M., & Lake, L. W. (2018). Machine learning-based optimization of well locations and controls. Computational Geosciences, 22(6), 1541–1558.
6. https://www.mdpi.com/2073-4441/15/15/2683

Thankyou for your time, If you’re interested in the projects I’m working on, feel free to reach out me by DM or my email olvandri.nouva25@gmail.com