A Data Analytics and Machine Learning Study on Site Screening of CO2 Geological Storage in Depleted Oil and Gas Reservoirs in the Gulf of Mexico

Abstract

Abstract Carbon capture, utilization and sequestration (CCUS) is playing an important role in mitigating greenhouse gas emissions and reversing global warming. Finding suitable sequestration sites is a key stage for CCUS, however, numerous potential sites do not hold sufficient data for evaluation based on conventional criteria. This study presents a feasible and efficient composite-scoring system using data analytics method to provide feeding data for machine learning studies to assess the potential depleted oil/gas reservoirs for carbon storage with limited data available. BOEM database in Gulf of Mexico (GoM) is used to apply the scoring system as a case study. The paper first presents a description of the reservoir properties and geological properties in GoM; then, the paper introduces the analytics method that maximize values of the available data and generate the final composite score for machine learning models; finally, the paper applies unsupervised machine learning method to cluster the reservoirs based on the location and composite scores. The findings of this study present that the scoring system quantifies and generalizes different types of geological and petroleum data to be feasibly applied in machine learning studies. Besides, with the BOEM database, the study revealed the best scored reservoirs located along the Louisiana coastline with large capacity, good injectivity potential and great performance in safety and economy. The result is consistent with the previous multi-criteria screening studies.