Using Artificial Neural Network to Evaluate the Performance of Cyclic Steam Stimulation on Recovering Crude Oil

Junastri, Ade and Hidayat, Fiki (2022) Using Artificial Neural Network to Evaluate the Performance of Cyclic Steam Stimulation on Recovering Crude Oil. In: The 2nd International Conference on Oil and Gas Industry Technologies and Applications 2022 (ICOG-ITA2022), 14-15 September 2022, Kota Kinabalu, Sabah, Malaysia.

Text ICOG-ITA2022 Ade J.pdf Download (1MB)

Abstract

This study applies a data-driven technique to evaluate the performance of Cyclic Steam Stimulation (CSS) based on oil recovery and steam oil ratio (SOR). CMG-STARS is used to build the conceptual 3D model of CSS. Approximately 500 data was generated from CMG CMOST with six (6) input parameters affecting the performance of both oil recovery and steam oil ratio: injection volume, steam quality, injection rate, soaking time, injection temperature, and injection pressure. Meanwhile, Python programming language was used to code the artificial neural network (ANN) algorithm code to evaluate the overall model performance. We used the trial-and-error method to choose the optimum nodes and hidden layer. The ratio of 0.8:0.2 is used to train and test the model. The optimal and stable hidden layer nodes are obtained at node 14 with R2 train 0.9976 and R2 test 0.9882. Moreover, other statistical performances from the RMSE and MAPE indicated good model prediction. We also found that the injection volume, rate, and steam quality affect the recovery and SOR more than others.

Actions (login required)

View Item