One-Liner Modeling carbon storage operations as a POMDP to show how different monitoring strategies can influence decision quality. Evalutae Novelty Applying POMDP to the task of carbon capture monitor planning. Notable Methods POMDP formulation Solver: POMCPOW Reward: trapped, free, and exited Co2 Action: injector placement Observation: CO2 saturation Belief: the permeability of the rock POMDP Solution: particle filter tree. Experimental design validated by simulations of CO2 sperad through injectors Fourier Network Simulation The actual fluid dynamics is really really hard to solve. As such, we do the evaluation over a lot of scenarios and then train a neural network to act as surrogate. Key Figs New Concepts Notes