Using Data-Driven Tubing Head Pressure Proxy for Integrated Production Network Coupled with Subsurface Models: A Way to Improve Computational Speed While Keeping Real Field Limits and Prediction Accuracy

Abstract The coupling of surface and sub-surface modeling approach has been deployed in managing short- and long-term forecasts in the industry over the decades. Computation time for the fully coupling model poses a real challenge for long term forecast. The compromise solution was proposed for long term reservoir management studies, where surface networks are removed from the fully coupling model. The key challenge to this approach is the assignment of the wellhead pressure constraints, which are retrieved from surface network in fully coupled model. Multiple assumptions of constant wellhead pressures were tested with random success of replication to fully coupled model results. The field level forecasts were reasonably reproduced; however, the reproduction was not achieved at gathering stations which are several kilometers away from each other. This poses a potential divergence of regional depletion rate of reservoirs from fully coupled model forecasts, which could impact reservoir management studies. The advancement of native functionality of reservoir simulator allows parameters to be calculated dynamically. Thus, "Tubing Head Pressure (THP) Proxy" was proposed to dynamically calculate values of wellhead pressure constraints over forecasting periods, where the pressure values calculated from subsurface models are close to solutions derived from fully coupled model. The production forecasts with coupled subsurface models and THP proxy show excellent representation of fully coupled model at a significantly shorter turnaround time. The usage of THP proxy resulted in 4.5-fold decrease in computational run time compared to fully coupled model with the 0.3% error in cumulative oil produced at the end of concession production forecast date. For end of field life forecasts, the run time optimization achieved nearly 3.3 times reduction compared to surface and subsurface coupled model with the reasonable match for total volumes produced for the given period. The robustness of the solution over the wide ranges of reservoir management options and subsurface model ranges (P10, P50, P90) was observed. The possibility of evaluating surface system debottlenecking options using THP proxy was also explored and tested. This paper describes the workflow which is aided by data science/analytics that generates THP proxy functions and integrate them into coupled subsurface modeling. This method helps enterprises to greatly improve the speed of long-term forecast deliverables from business planning & reserves booking processes, and governmental regulation documents that support the field development plan while maintaining the actual field constraints and prediction accuracy.