Data-Driven Decision-Making Strategy for Thermal Well Completion

Various wellbore completion strategies have been developed for thermal wells in Western Canada. The idea in this paper is estimating the improvement of oil production and steam injection if flow control devices (FCDs) will be installed for the next wells to be drilled, or if FCDs were installed at a particular well-pad that has not yet been completed with any FCDs. The approach is based on labeled real data for 68 well-pads from seven major thermal projects in Western Canada. Three phases make up the paper's methodology. The first phase compares wells with and without FCDs to evaluate the performance of the FCDs based on normalized oil production and cumulative steam oil ratio (cSOR). The second phase involves clustering well-pads using an unsupervised incremental-dynamic algorithm. An estimation of FCD contribution to enhancing oil production and cSOR is also performed for test well-pads based on their most similar cluster. In the third phase, cross-validation is employed to ensure that the estimation is trustworthy, and that the procedure is generalizable. To evaluate the performance of FCDs, a reliable comparison was made using normalized oil production and cSOR. Based on our analysis from October 2002 to March 2022, successful FCD deployment resulted 42% more normalized oil and a 37% reduction in cSOR. Among these, liner deployed (LD) FCDs increased oil production by 44% while decreasing cSOR by 58%. Although tubing deployed (TD) FCDs are installed in problematic wells, they produced 40% more oil while decreasing cSOR by 21% in successful cases. Successful inflow control devices (ICDs) increased oil production by 40% while lowering cSOR by 45%. Successful outflow control devices (OCDs) increased oil production by 82% while reducing cSOR by 22%. The clustering algorithm separates the database into four clusters that will be utilized in the estimating phase. In the estimation phase, ten well-pads (15% of the database) are presumed to be new well-pads to be drilled (test data). Based on the estimation results, the root mean square errors (RMSEs) for FCDs contribution to enhancing oil production and cSOR for the test well-pads are 12%. Cross-validation was also performed to assess the approach's predictability for new data, to verify that our technique is generalizable. The findings indicate that FCDs might result in lower capital expenditures (CapEx) and greenhouse gas (GHG) emissions intensity for SAGD well-pad developments, allowing them to reduce emissions. The conclusions of this research will aid production engineers in their knowledge of relative production performance. The findings may be used to examine paradigm shifts in the development of heavy oil deposits as technology advances while keeping economic constraints in mind.

[1]  J. Leung,et al.  A hybrid GBPSO algorithm for permeability estimation using particle size distribution and porosity , 2022, Journal of Petroleum Science and Engineering.

[2]  E. Artun,et al.  Machine-Learning Approach for Forecasting Steam-Assisted Gravity-Drainage Performance in the Presence of Noncondensable Gases , 2022, ACS omega.

[3]  Bo Yang,et al.  Data-driven model for predicting production periods in the SAGD process , 2022, Petroleum.

[4]  J. Leung,et al.  Data-Driven Well Pad Development Performance Review: Focus on the Role of Liner Design on Well Performance , 2021, Day 1 Fri, November 26, 2021.

[5]  Shangqi Liu,et al.  Data-Driven Proxy Model for Forecasting of Cumulative Oil Production during the Steam-Assisted Gravity Drainage Process , 2021, ACS omega.

[6]  Farzad Bashtani,et al.  Analysis of the Performance of Various Well Liner Completion Strategies in Surmont , 2021 .

[7]  Georgios Kollias,et al.  Reinforcement Learning For Field Development Policy Optimization , 2020 .

[8]  Vahidoddin Fattahpour,et al.  Altered mineral segmentation in thin sections using an incremental-dynamic clustering algorithm , 2020, Eng. Appl. Artif. Intell..

[9]  M. Irani,et al.  Modeling the Conformance Improvement Using Flow Control Devices in Infill Wells Adjacent to SAGD Well Pairs: No Flashing , 2020 .

[10]  F. Torabi,et al.  An Improved Mathematical Model for Accurate Prediction of the Heavy Oil Production Rate during the SAGD Process , 2020, Processes.

[11]  M. Irani On Subcool Control in SAGD Producers—Part II: Localized-Hot-Spots Effects and Optimization of Flow-Control Devices , 2019, SPE Journal.

[12]  F. Torabi,et al.  Predicting the performance of steam assisted gravity drainage (SAGD) method utilizing artificial neural network (ANN) , 2019 .

[13]  M. Irani On Subcool Control in the SAGD Producers. Part II: Localized Hot Spots Effects and Optimization of Flow-Control-Devices , 2018, Day 3 Thu, November 29, 2018.

[14]  Xiaohu Dong,et al.  An empirical correlation to predict the SAGD recovery performance , 2015, J. Comput. Methods Sci. Eng..

[15]  Lin Shi,et al.  Oil production rate predictions for steam assisted gravity drainage based on high-pressure experiments , 2013 .

[16]  Sally M Benson,et al.  Carbon capture and storage. , 2007, JAMA.

[17]  Radford M. Neal Pattern Recognition and Machine Learning , 2007, Technometrics.

[18]  Saul B. Suslick,et al.  Quantifying the value of technological, environmental and financial gain in decision models for offshore oil exploration , 2001 .

[19]  Du Ta,et al.  Effects of Reservoir Parameters and Operational Design on the Prediction of SAGD Performance in Athabasca Oilsands , 2012 .

[20]  L. B. Cunha,et al.  Prediction of SAGD Performance Using Response Surface Correlations Developed by Experimental Design Techniques , 2006 .