Performance analysis of various machine learning algorithms for CO2 leak prediction and characterization in geo-sequestration injection wells
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The effective detection and prevention of CO2 leakage in active injection wells are paramount for safe carbon capture and storage (CCS) initiatives. This study assesses five fundamental machine learning algorithms, namely, Support Vector Regression (SVR), K-Nearest Neighbor Regression (KNNR), Decision Tree Regression (DTR), Random Forest Regression (RFR), and Artificial Neural Network (ANN), for use in developing a robust data-driven model to predict potential CO2 leakage incidents in injection wells. Leveraging wellhead and bottom-hole pressure and temperature data, the models aim to simultaneously predict the location and size of leaks. A representative dataset simulating various leak scenarios in a saline aquifer reservoir was utilized. The findings reveal crucial insights into the relationships between the variables considered and leakage characteristics. With its positive linear correlation with depth of leak, wellhead pressure could be a pivotal indicator of leak location, while the negative linear relationship with well bottom-hole pressure demonstrated the strongest association with leak size. Among the predictive models examined, the highest prediction accuracy was achieved by the KNNR model for both leak localization and sizing. This model displayed exceptional sensitivity to leak size, and was able to identify leak magnitudes representing as little as 0.0158% of the total main flow with relatively high levels of accuracy. Nonetheless, the study underscored that accurate leak sizing posed a greater challenge for the models compared to leak localization. Overall, the findings obtained can provide valuable insights into the development of efficient data-driven well-bore leak detection systems.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it