Energy storage is an essential component of future energy systems that use large amounts of variable renewable resources. Apart from pumped-storage hydro-power, large scale energy storage is mainly provided by underground energy storage systems. Underground energy storage can be mechanical, chemical or thermal energy storage. Mechanical storage includes compressed air energy storage (CAES), chemical storage means, as an example, hydrogen and synthetic natural gas storage (SNG), and thermal storage can be underground thermal energy storage (UTES).To improve understanding of the complex and coupled processes and enable planning and risk assessment of subsurface energy storage, efficient, consistent and adequate numerical models for multi-phase flow, transport and energy processes are required. Simulating underground energy storage requires simulations on a large domain over the whole time of plant operation and beyond, including local features such as fault zones and a representation of the transient saline front. In addition, often a large number of simulation runs need to be conducted to quantify parameter uncertainty (e.g. Monte Carlo simulation). Efficient models are needed as well to integrate measurements during simulation to improve predictions (e.g. by means of data assimilation). Within acceptable computational time this cannot be achieved by three-dimensional multi-phase multi-component models due to limited computational resources. Therefore, a reduction of model complexity and thus computing effort is required.Numerous simplified models that require less computational resources have been developed. However, these models might be less accurate and based on assumptions that might not hold in the whole domain during the whole simulation time. Therefore, we aim to combine the individual benefits of these simplified models and more complex and thus more accurate models by coupling both model types in one domain. This will be a step towards developing an integrated grid-adaptive multi-scale multi-physics modeling approach that will facilitate simulations of large-scale problems efficiently without losing the relevant accuracy.
05/2014 - 04/2017