PROXY MODELING IN CO2 HUFF-N-PUFF INJECTION USING ARTIFICIAL NEURAL NETWORK

• Main Author: Prayoga, Yudhi
• Format: Theses
• Language: Indonesia
• Subjects: Petrologi
• Online Access: https://digilib.itb.ac.id/gdl/view/47379
• Institution: Institut Teknologi Bandung

Oilfield development processes need to be optimized in order to increase cumulative production, especially in mature field. CO2 utilization has proven to give promising results in increasing oilfield production. Stimulation using CO2 huff-n-puff injection has been utilized and developed to increase production by reducing oil viscosity and oil swelling. Evaluation of CO2 huff-n-puff injection needs to be predicted in little time. Artificial neural network model is one of the predictive models used to model a system with high dimension and non-linearity. Model is constructed using a number of experiment data that was determined using experimental design. This study uses artificial neural network to predict production profile in reservoir using CO2 huff-n-puff injection based on 22 reservoir and injection constraint parameters Model was developed by simulating 3.575 experimental data to obtain oil cumulative production, peak oil rate, time at peak oil rate, reservoir pressure, and water saturation results from simulation with a total number of 44 proxy models. Production profile is predicted by assuming a linear increment of production rate before peak production rate and an exponential decrement of production rate after peak production rate. Predicted value of observation results has a relatively low error compared to simulation results with the MSE of cumulative production model is 0.01, 2.21E-06 for peak oil rate model, and 4.15E-10 for time at peak oil rate model. Production rate profile predicted with artificial neural network model has a same trend line with production rate profile based on simulation results.