InterPore2026

Stochastic cooperative game models for CO2 storage with uncertain payoffs under pressure space competition

19 May 2026, 17:10

15m

Oral Presentation (MS19) Uncertainty-Aware Decision Support in Porous Media Applications MS19

Speaker

Per Pettersson (NORCE Norwegian Research Centre)

Description

Mitigating global warming requires a substantial growth in permanent geological CO$_2$ storage by 2050 compared to today’s scale. An increasing number of active CO$_2$ storage projects leads to increased risk due to uncertainty from potential pressure communication between different projects, as well as utilization of sites with limited data. Thus, insurance and other forms of risk sharing under uncertain geological conditions become relevant to many injection well operators.
We propose a stochastic game theoretic model for independent and competing reservoir agents (e.g., injectors) to find out whether they should collaborate in situations with uncertain geological conditions, as well as uncertainty in the possible injection volumes as a consequence of the actions of competitors. The injection operators are modeled as agents in a cooperative game with uncertainty in the amount of CO$_2$ they can safely inject. No injector knows exactly how much CO$_2$ can be injected, but may prefer to share the associated risk by collaboration with other injectors. Depending on the preferences of the agents, they can be more willing to take financial risks with the prospective of larger injection volumes, or they may prefer to make choices that avoid the risk for smaller-than-expected injection volumes. Under some conditions, they can form collaborations that are attractive for all of them, even if all of them prefer to avoid financial risk.
If the operations of the agents are truly independent of each other, there is a natural baseline scenario where they maximize their own injections without collaborating or interfering with each other. There may still be incentive to collaborate for risk sharing, but only if the outcome is assumed more likely to be more attractive than the baseline scenario. The physical uncertainty is modeled using geostatistical methods combined with numerical simulation to estimate the effect on the storage potential. Any given agent is faced with the task of choosing between an unknown outcome of the baseline scenario, and the outcome of one or more risk sharing schemes agreed upon with the other agents. A stochastic preference relation provides a means to systematicaly make such decisions.
If the agents' operations affect each other by means of, e.g., pressure communication, there may be no unique natural definition of a baseline scenario. As a remedy we suggest belief distributions that combine uncertainty in physical data with least-informative prior distributions to model a perceived baseline scenario. The belief distribution should use as much physical information as possible, but with as few as possible assumptions not directly supported by data or otherwise justified by apriori considerations.
W present numerical results for the Utsira formation in the North Sea, for cases both including and excluding pressure competition. We show that risk adverse agents benefit from collaboration in settings where there is no pressure communication or other interference between agents. It is also demonstrated that pressure communication leads to large variability in the feasible injection rates, but the resulting belief distributions are still informative and can be used to aid in decision making about collaboration.