← Back to Sessions

Adaptive Model Selection for the Maximum Magnitude Event During Injection

Description: 

As injection of fluids into the subsurface becomes more common, the need for a technique to assess the maximum magnitude of induced events (Mmax) becomes more pressing. There are some models in the literature (e.g., uncalibrated moment cap, calibrated moment cap, statistical formulation, residual moment, and convolutional) that relate the cumulative injection to the Mmax. These models incorporate different site characteristics and physics such as seismogenic index, Gutenberg-Richter b-value, pore pressure diffusion, and moment release. We hypothesize that the model that best describes a certain injection operation may vary during the course of the injection. In fact, the seismic response ranges from mild due to aseismic slip and up to hazardous as dynamic runaway rupture behaviors are developed. Thus, we present a dynamic calculation of the Mmax predictive model. The injection might take the form of deep saline carbon dioxide sequestration, enhanced oil recovery, hydrogen storage, or stimulation of a reservoir. The processes associated with each of these scenarios are different, and we conduct a comparison between the enhanced oil recovery operations at the Farnsworth, TX site and the stimulation of the enhanced geothermal reservoir at the Utah FORGE site in Milford, UT. The results show that no one model describes the two injection projects the best. Rather, dynamically updating the model based on a misfit calculation between observed seismicity and the tested models allows the operator to incorporate the physics and site characteristics that are most applicable for the injection. The best fitting model is used for the prediction and down the road, for risk management and mitigation.

Session: Seismology for the Energy Transition

Type: Oral

Date: 4/18/2023

Presentation Time: 05:00 PM (local time)

Presenting Author: Kevin McCormack

Student Presenter: No

Invited Presentation: 

Authors