Abstract
Summary Carbon capture and storage will be a valuable tool for continuing consumption of fossil fuels. In this study, deep saline aquifer (DSA) model with saline water production well was constructed for the development of optimization workflow. Due to the time-consuming of realizations of the numerical simulation, a surrogate reservoir model was developed based on the data extracted from the numerical simulator. pre-processing and data normalization was done and the most effective parameters were selected on the two outputs, namely CO2 breakthrough time and storage. Then, through 3 steps of validation, testing, and blind-validation, performance of artificial neural network (ANN) is confirmed. To deal with time-consuming and high-computing overhead required of numerical simulator, we couple the trained ANN to a multi-objective genetic algorithm. Instead of taking long-time, the optimization in DSA is carried out in minutes and the Pareto front is produced. Due to the lack of optimization studies in DSAs, especially the absence of our two-objectives and its coupling with a data-driven model, the importance of this study is doubled. In this framework, we used 8000 experiments chosen randomly within the range of algorithm input features and we found that all experiments were dominated by the developed Pareto front.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have