Advancing Well and Remote Compressor Performance Through Data Analytics
Suradech Kongkiatpaiboon, Sarita Laosuwan, Yuranan Tangsawanich, Chain Sopitviriyaporn, Monthep Parimontonsakul, Polake Kaivalkritiyakul, Siriwan Payaksiri
SPE/IATMI Asia Pacific Oil & Gas Conference and Exhibition (APOGCE), Jakarta, Indonesia, 2023 · SPE-215265-MS
Finding the bottleneck between a well and its unmanned remote compressor (choke, flowline or compressor settings) used to depend on engineers manually checking each case. This work turned the team's troubleshooting rules into an automated workflow: thousands of compressor simulation runs were used to train an ML model, which is matched against live plant data in the cloud to flag constraints and suggest fixes in a web app. Bottlenecks are now found faster, opportunities that used to be missed are caught, and offshore adjustments happen sooner.
Key results
- The data-science surrogate model matched the compressor simulation results with an R² above 0.99
- Several undersized or problem choke valves were replaced, and erosion-limited carbon-steel flowlines were changed to duplex stainless steel
- Using spare remote-compressor capacity let selected wells be routed to the compressors, which raised overall production
- Bottlenecks are found much faster and the operations team gave largely positive feedback (no production figure is given)
Approach
- Analysed historical shutdown and failure records and turned the usual five-step optimization loop into written rules
- Ran thousands of compressor simulation cases (suction/discharge pressure, rate, speed) into a data-lake database and trained ML models that map live operating data to simulation results
- Checked choke valves by calculating Cv with the ISA S75.01 method against a normal envelope, and flowlines against API 14E erosional limits
- Automated the pipeline in Python with PI historian data in a cloud ML deployment environment, delivering daily recommendations and e-mail alerts through an interactive web application, developed with scrum
The problem
Optimizing wells, chokes, flowlines and remote compressors was slow and manual, so many bottlenecks and production opportunities were never investigated.
Tools & methods
Summary written for this site from the published paper. For figures, data and the full method, read the paper on OnePetro.