The goal of this project is to predict the efficiency of CO2 trapping with #machinelearning. To evaluate the efficiency of CO2 trapping performance, we need to estimate the Residual Trapping Index (RTI) and Solubility Trapping Index (STI).

Using the flowchart on the picture above, we extract petrophysical properties from multiple real field data to get the input and output. Then, we found that the #randomforest algorithm gives us the robust model.

So what do we find? The residual gas saturation and post-injection play an important role when developing the model. This is also indicated by the high correlation between these features with the targets.

We both really enjoy the learning process. With fantastic assistance from Mr. Hung, our model achieved 0.99 R2 scores both for train and test data with the minimum error.

Skills: Python, Numpy, Pandas, scikit-learn, Keras, Scipy