Intelligent MPD Set-Point Control for Narrow Windows: A Reinforcement Learning Framework for Automated Choke Control in Deepwater HPHT Wells

Deep water high-pressure, high-temperature (HPHT) drilling in managed pressure drilling (MPD) offers great challenges because of very thin changes between pore pressure and fracture gradient, frequently below equivalent mud weight of 0.3 ppg. Traditional rule-based MPD tuning algorithms fail to hold optimal setpoints in real-time resulting in influx/loss events and inefficient rate of penetration (ROP).

This paper introduces a new reinforcement learning (RL) model of intelligent MPD set-point control, which is more specifically backpressure and equivalent circulating density (ECD) optimization in narrow-margin deepwater wells that characteristic of the Gulf of Guinea area, such as Ghana and Nigeria. A Deep Deterministic Policy Gradient (DDPG) model was optimized using past MPD operational data in West African deepwater campaigns which included a multi-objective reward function balancing between influx risk, loss risk, and ROP optimization. The paradigm shows that the pressure control accuracy (23% improvement over rule-based approaches) and mean absolute pressure deviation (42 psi to 32 psi) are lower in comparison to rule-based approaches.

Moreover, the model had an average increase of 15 percent in the average ROP and was able to keep wellbore stable despite the thin drilling window. The intelligent control system detected and reacted to simulated kick situations 18 seconds earlier than the traditional automated MPD systems, which is a giant development in real-time well control capacity. The above results imply that the operational benefits of RLbased MPD control are enormous in strenuous deepwater HPHT drilling operations, and could be applicable to the whole range of the Gulf of Guinea deepwater drilling ventures.

Managed pressure drilling, reinforcement learning, HPHT, narrow margin, deepwater drilling, DDPG, choke control, Gulf of Guinea

1. Arnø, M., Godhavn, J.M., Aamo, O.M. (2020). Deep reinforcement learning applied to managed pressure drilling. SPE Bergen One Day Seminar, Bergen, Norway. [Google Scholar] [Crossref]

2. Ding, Y., Chen, Z., Zhang, K., et al. (2023). A reinforcement learning method for optimal control of oil wells. Heliyon, 9(7), e17751. [Google Scholar] [Crossref]

3. Ekechukwu, G., Uzochukwu, E., Ibekwe, K. (2024). Explainable machine-learning-based prediction of equivalent circulating density. Scientific Reports, 14, 17620. [Google Scholar] [Crossref]

4. Elliott, D., Montilva, J., Francis, P., et al. (2011). Managed pressure drilling erases the lines. Oilfield Review, 23(1), 14-23. [Google Scholar] [Crossref]

5. Gamal, H., Elkatatny, S., Abdulraheem, A. (2021). Machine learning models for equivalent circulating density prediction. ACS Omega, 6(40), 26267-26276. [Google Scholar] [Crossref]

6. Gao, X., Liu, Y., Wang, H., et al. (2024). Equivalent circulation density prediction using random forest. TPE, MS.ID.000541. [Google Scholar] [Crossref]

7. Hauge, E., Aamo, O.M., Godhavn, J.M. (2013). Automatic kick detection and handling in managed pressure drilling. Ph.D. Thesis, Norwegian University of Science and Technology. [Google Scholar] [Crossref]

8. Huang, X., Luu, H., Shang, S., et al. (2024). Deep reinforcement learning for automatic drilling optimization using an integrated reward function. SPE/IADC Drilling Conference and Exhibition, Galveston, Texas. [Google Scholar] [Crossref]

9. Keshavarz, S., Elahifar, B., Gholami, A. (2024). Deep reinforcement learning algorithm for wellbore cleaning across drilling operation. Fourth EAGE Digitalization Conference & Exhibition. [Google Scholar] [Crossref]

10. Keshavarz, S., Elahifar, B., Gholami, A. (2025). Deep reinforcement learning for automated decision support in oil well operations. Energy Reports, 13, 5967. [Google Scholar] [Crossref]

11. Najjarpour, M., Jalalifar, H., Soleimani, B. (2022). Managed pressure drilling technology, mechanical specific energy and bit management for ROP management. Journal of Petroleum Science and Engineering, 209, 109834. [Google Scholar] [Crossref]

12. Park, J., Price, C., Pixton, D., et al. (2020). Model predictive control and estimation of managed pressure drilling using a real-time high fidelity flow model. ISA Transactions, 97, 76-89. [Google Scholar] [Crossref]

13. Scoular, T., Hathaway, K., Essam, W., et al. (2012). BP case study: MPD application supports HPHT exploration. Drilling Contractor, July/August 2012. [Google Scholar] [Crossref]

14. Squintani, E., Bonin, R., Borello, F., et al. (2018). Deepwater HPHT drilling through ultra narrow PPFG window: A case study by ENI. Abu Dhabi International Petroleum Exhibition & Conference. [Google Scholar] [Crossref]

15. Xiong, M., Wang, Y., Liu, X., et al. (2024). A rate of penetration (ROP) prediction method based on BiLSTM-SA-IDBO. Scientific Reports, 14, 24812. [Google Scholar] [Crossref]

16. Zhou, J., Gravdal, J.E., Strand, P., et al. (2016). Automated kick control procedure for an influx in managed pressure drilling operations. Modeling, Identification and Control, 37(1), 31-40. [Google Scholar] [Crossref]

• Drone-Based Phenotyping and its Utilization in Crop Improvement: A Review
• An IoT-Enabled Smart Healthcare Monitoring System Using Machine Learning for Early Health Risk Prediction
• Strategic Integration of Artificial Intelligence for Achieving Operational Excellence in Container Freight Stations Using Evidence from Chennai, India
• Adoption of OTT platforms: Analyzing User Behavior through the UTAUT2 Model
• Customers’ Buying Behaviour in Relation to Online Food Delivery (Evidence from National Capital Region (NCR) of India)