Auto-Fill: A Synergistic Approach for Surveillance and Fillbase Management of Rod Pumps Using Advanced Machine Learning Models
Bibliographic record
Abstract
Abstract Surveillance and adjustments of fillbase and fillage are crucial for optimizing pump performance and maximizing production efficiency. Proper management of fillbase not only promotes longevity by preventing unnecessary wear and tear, but also helps in optimizing fluid handling efficiency, energy efficiency, rod-load management, and overall production of rod pumps. The primary objective of this paper is to propose a comprehensive and automated solution for surveillance and optimal management of fillbase with high number of wells. Our proposed method powered by combination of machine learning models enable efficient, data-driven monitoring and fillbase tuning. The pump cards (Position-Load on the rod) generated by the wellhead controllers along with other operational parameters including pump-speed, wellhead-pressure, temperature, etc. are first processed by a LightGBM classification model to classify cards as good or bad. If the card is bad, the operators are notified to help with supervision and taking remedial actions, and if the card is good, fillbase is predicted by a XGBoost regression model and the values are adjusted automatically based on the predictions. Different machine learning models were assessed for classifying pump cards and predicting fillage and fillbase. The best models with their respective hyperparameters were then tested in the field. These models were evaluated in real-time by senior operation specialists and feedback was collected. The card classification model had an accuracy score of approximately 99% on the test data and showed similar performance during real-time monitoring. The Fillbase model had R2 values of 98% and the fillbase model correctly detected the fillbase setting in around 95% of cases, as confirmed by senior operation specialists and production engineers. The controller calculates the fillage based on the given fillbase setting and there is an average deviation of around 4% from the fillage predicted by our model. The fluid rates recorded by the controllers after adjusting the fillbase were compared to well test rates, and there was a significant decrease (15%) in mean absolute error after implementing the ‘Auto Fill’ workflow, which serves as a quantitative validation for our models. Additionally, the model can adjust the fillbase settings even in the event of sudden changes in operating conditions, reducing the need for frequent manual interventions by operators. Using data driven systems and powerful machine learning models, the proposed approach aims to automate surveillance and fillbase management, reducing the need for manual effort from operators and production engineers. This process also enhances consistency in setting fillage or fillbase and allows for automated speed control of pumps, improving production and energy efficiency. The system is specifically designed to optimize operational productivity in large oil field settings with numerous wells by streamlining surveillance and fillbase adjustment processes.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".