A physics-informed optimization workflow to manage injection while constraining induced seismicity: The Oklahoma case

A newly developed modelling framework is presented which specifically focusses on the central Oklahoma case and the high-volume injection of wastewater, which led to a surge of induced seismicity. However, the modelling framework is versatile enough to be applied to any anthropogenic subsurface activities and should be seen as a good practice to manage injection while minimizing induced seismicity. The objective is to account for all the available knowledge to deploy the simulation of the flow, induced stress changes and seismicity in the underground. The spatio-temporal pore pressure changes caused by high-volume injection are first determined by using the historical injection rate of the 220 wells at central Oklahoma. From these pressure fields, induced stresses at the basement depth, due to both pore pressure diffusion and poro-elastic inflation of the underground, are computed. The rate-and-state frictional response of the Oklahoma faults is then honored to derive the yearly seismicity rate. After assimilation of the observed seismicity at central Oklahoma, it is demonstrated that our predictions can well explain the historical spatio-temporal evolution of the seismicity at central Oklahoma. Finally, making use of the calibrated predictive model, a constrained optimization approach is used for an efficient screening of multiple injection scenarios. Ultimately, an optimum theoretical scenario is identified which allows the maximization of injection volumes while keeping the seismicity level below a safe cap and, more specifically, would have prevented the dramatical growth of the seismicity rate in 2015. The optimum scenario involves equalizing the injected volumes in all wells and preventing the injection of additional large volumes in the area where most of the wastewater have been already injected prior 2014.