Data analytics to investigate the cohort of injection wells with earthquakes in Oklahoma

Abstract

The number of recorded earthquakes in Oklahoma has substantially increased during the last few decades, a trend that coincides with the increases in the injected volume in underground injection control (UIC) wells. Several studies have suggested the existence of spatial and temporal links between earthquakes and injection wells. However, creating a spatial connection between the earthquakes and UIC wells requires making a prior assumption about the radius of induced seismicity. In this study, we use intrinsic features of the UIC wells to find the cohort of wells with associated earthquakes, based on the level of activity and proximity of the wells to the events. For this purpose, a hybrid genetic algorithm–K-means (GA-K-means) algorithm was applied over UIC wells, and the geographical representation of the clustered wells was co-visualized with earthquake data to determine wells with induced seismic activities. The analysis was performed every year since 2002, and the most critical attributes to distinguish the behavior of wells were identified. The analysis showed a distinct change in cluster identifiers before the year 2010, which is believed to be the beginning of increased seismic activities, compared to later dates. Our approach was able to group the earthquake-associated wells from the rest of the data, and centroid analysis of these wells helped us identify the critical pressure and cumulative volume range that result in induced seismicity. These findings can be used as guidelines for designing safer injection sites for sustainable energy production in Oklahoma.