Perpustakaan Digital ITB

The inherent immobility of heavy oil under reservoir conditions is the primary challenge in the extraction of heavy oil. Thermal processes are the preferred method for reducing viscosity and augmenting overall recovery, as heavy oil exhibits temperature-dependent viscosity. This research concentrates on a sophisticated steam injection technique known as Steam Assisted Gravity Drainage (SAGD). The purpose of this study is to improve the efficacy of SAGD by examining the heterogeneity of parameters that affect it. This study used a reservoir located in the Celtic field as a benchmark for constructing simulations in CMG Builder and CMG STARS. Multiple factors were chosen to assess sensitivity and determine the recovery factor. Subsequently, these factors were amalgamated into 800 experiments, which were utilized to construct a proxy model using CMG CMOST. This proxy will be capable of predicting the desired model. The proxy model is built with polynomial regression and neural networks method to seek the value of recovery factor. Polynomial regression resulting R2 training of 0.882 and R2 verification of 0.899, while neural networks resulting R2 training of 0.972 and R2 verification 0.864. Then, the proxy model between these two methods is determined for optimization. Optimization is done by changing the adjustable parameters. The results come out as RF proxy and will be verified by simulations. By comparing the results from proxy calculations and simulations, it is known that its similarity is suitable, marked with R2 of 0.993. This paper conducts a thorough examination of the crucial factors that impact the effectiveness of SAGD and gives practical knowledge on how to optimize thermal recovery techniques for heavy oil reservoirs by including the correlation between RF with CSOR and Np.