Energy storage in the geological subsurface: dimensioning, risk analysis and spatial planning: the ANGUS+ project

The current status of the German laws concerning the use of the deeper subsurface is a first-come first-serve approach, in which the first applicant for a specific location is granted exclusive prospection and, in a second step, utilization rights. This implies that a geological formation, which is well suited for subsurface energy storage, may be used for other purposes or may be blocked by other types of use in intermediate formations, which inhibit access to the intended storage formation. Right now, no mechanism is in place which allows to reserve certain locations in the subsurface for specific types of use. Due to the limited transmission capacities of the power grids, geotechnical storage sites will have to be found close to large producers of renewable electric power. Therefore, a planning of the use of the geological subsurface is required to ensure that storage capacities remain available where needed, and are not blocked by other types of use. This would also require a prioritization of common public interests over private company interests (Bovet 2014).

The competition for subsurface space occurs in two distinct forms: Firstly, in the form of a direct competition for an individual storage formation, where different use options apply but the formation can accommodate only one. Secondly, in the form of an indirect blocking, where a certain subsurface space may be unavailable for energy storage due to the induced effects of other types of use (e.g., pressure rises, temperature increases or induced brine movement). Access to an intended storage formation may also be blocked indirectly through operations in shallower formations which, e.g., prohibit drilling. In the case of direct competition, a mechanism is sought, which allows to prioritize individual types of use. Here, tools are required to assess the benefits of the individual storage options in terms of achievable extraction or injection rates and capacity. In the case of indirect competition, a quantification of the mutual effects of individual storage operations on each other and on other types of use in the subsurface is required. In both cases, a delineation of the space required for an individual storage option is needed to be able to reserve and assign subsurface locations for this specific type of use. Three categories of used subsurface space have been identified and developed in the ANGUS+ project in the context of geotechnical energy storage: firstly, the “operational space” (Fig. 2), i.e., the space directly used by the storage operation, which comprises the technical installations and the space taken up by the injected gas or heat. Secondly, the “affected space” (Fig. 2), i.e., the surrounding space where effects of the storage operations are induced, such as zones of elevated pressure, induced brine movement, elevated temperatures, etc. And, thirdly, the “monitoring space” (Fig. 2) that needs to be free of other types of use due to monitoring requirements of the energy storage. Here, for instance, pressure monitoring of a porous medium gas storage site prohibits other types of use close by which also influence the formation pressure and thus disturb the monitoring signal. Typically, volumes affected by induced effects will be larger than volumes occupied by the direct use, and the volumes reserved for monitoring will typically again exceed the affected space. The monitoring space may extend to the land surface, if, for example, geophysical monitoring methods are used. A similar, less physically based approach was postulated by Kahnt et al. (2015), who suggested the definition of a project space as the volume taken up by the actual subsurface operation of concern and a used space around the project space as a buffer zone where induced effects exceed a certain threshold. All concepts of space in the subsurface, however, require the definition of boundaries. Thus, the definition of, for example, threshold values for pressure increases and temperature changes is required in order to delimit the operational, affected and used space. This will always imply the use of modeling tools, as delimiting these spaces by measurements only is not feasible in the geological subsurface. The process-based simulation of the governing processes on a site-specific basis and based on site-specific observation data can be a means to delimit these subsurface spaces.

An enhanced knowledge of the induced effects and monitoring requirements of subsurface energy storage and its interactions with other types of subsurface use (e.g., waste disposal, gas storage vs. heat storage, water supply and distribution) needs to be made available to authorities and policy makers, and requires consolidation in a subsurface spatial planning scheme in order to serve as beneficial information to planning procedures. Informed decisions can prevent interferences and can avoid wasting of subsurface space by implementing less efficient types of subsurface use. In view of the demands on a future sustainable subsurface spatial planning scheme comprising the concept of operational, affected and monitoring spaces of prospective storage sites, the overall objectives of the ANGUS+ project are to develop

The coordination of subsurface use options, environmental protection and surface infrastructure is crucial for an economically and ecologically feasible exploration of the subsurface as a spatial asset. A sustainable approach to subsurface use therefore requires spatial planning for minimized competition and maximized synergies (ARL 2012; Gerling 2010; Bovet 2014). Bauer et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-013-2883-0)] qualitatively demonstrated the expectable mutual effects of different subsurface applications on each other and identified a need for site-specific process quantifications. With heat storage, gas storage and 3D spatial information as cross-cutting themes of the ANGUS+ project (Fig. 3), this aim is pursued by using numerical simulations of synthetic, however, realistically defined energy storage scenarios. The literature research, laboratory experiments and numerical modeling contribute to the parameterizations preceding the realistic scenario definition of geological formations and induced processes. Enhanced numerical tools enable the computation of coupled thermal (T), hydraulic (H), mechanical (M) and chemical (C) processes (THMC-processes) induced by geotechnical energy storage operations for realistically parameterized sites on the field scale. This simulation of virtual storage site operations allows a generic approach to storage environments, as the quantification of occupied subsurface space can be used in the permitting process for all storage types considered. The identification of spatial extents and magnitudes of induced processes subsequently also allows for the evaluation of monitoring schemes. Comprehensive data on geological parameters, energy infrastructure and protective areas in Schleswig–Holstein are compiled in a newly developed, web-based 3D spatial information system.

For the assessment of heat storage capacities and associated implications, scenarios of daily and seasonal storage are designed to represent typical operation modes for residential or office buildings. These scenarios are investigated with respect to the coupled thermal, hydraulic, mechanical and geochemical processes induced by the imposed temperature changes. Competition for subsurface space can also occur with unintentional types of subsurface use. As typical examples, interactions with preexisting contaminations in the formations affected by heat storage, and accidental leakages of heat transfer fluid from borehole heat exchangers (BHE) are investigated experimentally and numerically. The definition of realistic virtual scenarios for the assessment of heat storage operations requires geostructural, hydraulic, hydrogeochemical and mineralogical parameterizations of the subsurface formations involved. However, Dethlefsen et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5626-1)] characterized the availability of these data as at least partly affected by scarcity, value uncertainty and process knowledge uncertainty. Within the same study, consequently, a categorization of process knowledge and concepts for the quantification of prediction errors were developed: Process knowledge is categorized here based on the uncertainties of process parametrization and their resulting impacts on simulation results. Prediction errors are quantified by sensitivity analysis with regard to statistical data uncertainty and by experimental studies with regard to scenario uncertainty. This work thus also shows where further characterization efforts yield the largest effect on reducing uncertainty. In a deficit analysis of the data availability for the parameterization of the shallow subsurface in Schleswig–Holstein with respect to energy storage, Dethlefsen et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6343-5)] found that information on geostructures and hydraulic properties is typically available on the regional scale, but detailed information exists only patchy and locally. Near-surface groundwater is found to be well characterized concerning the main constituents, but data on trace elements and dissolved gas concentrations are sparse, as well as sediment properties and mineral compositions. Often, results are available for typical ambient or laboratory temperatures, but not for higher temperatures as expected in heat storage applications. Also, the temperature dependencies of individual geochemical or mechanical processes are often poorly investigated. Near-surface geophysical exploration, sediment characterization and geostatistical description of shallow subsurface parameters in cooperation with federal authorities and federal states are identified as requirements for reliable prognoses of the effects on drinking water aquifers [Dethlefsen et al. 2015; Dethlefsen et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6343-5)].

As an effect of larger temperature changes due to thermal energy storage, gas phases can form in shallow aquifers. A gas phase is bypassed by the groundwater flow and hinders the heat transport in ATES or BTES sites due to a reduced heat conductivity of the gas phase compared to the groundwater. These alterations reduce the system performance and induce changes in groundwater flow and chemistry. Experimental work using Pleistocene sands as typical aquifer sediments from Northern Germany and atmospherically equilibrated groundwater shows that gas phases may develop and accumulate, when temperatures are increased to 40 °C or up to 70 °C [Lüders et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6181-5)]. The same study shows with the help of geochemical equilibrium calculations that the gas phase formation in a water saturated aquifer is limited to the upper 12 m at temperatures up to 70 °C, assuming atmospheric equilibrium, due to the rising pressure, but extending somewhat deeper if H₂S, CO₂ or CH₄ are present as dissolved gases in higher concentrations.

Temperature increases may affect the mobility of potential inorganic groundwater contaminants. Trace elements of geogenic origin can potentially be released to the groundwater if shallow aquifers are exposed to temperature changes. The sorption behavior of arsenate in goethite sand under cyclic thermal loading was studied in circulating column tests and through titration of goethite surface charges. A temperature increase of 60 °C resulted in a reversible 70% increase in the arsenic concentration at the column outlet compared to the column inlet, which was not fully explained by a simple sorption transport model. Titrations showed a decrease in the pH_PZC (pH value of the point of zero charge) with increasing temperature, connected with a decrease in anion sorption sites at elevated temperatures and constant pH values. This effect resulted in a release of arsenic through the reduced sorption of arsenate anions at elevated temperatures. The experimental results shall contribute to the development of a transfer function to combine sorption isotherms at different temperatures for the description of the sorption-limited mobility as a temperature-dependent parameter (M. Ebert, pers. comm.).

In connection with thermal applications in the subsurface, changes in the microbial biocenosis are indicators for ecosystem changes and can be the cause for microbiologically induced operational disorders. Geochemical and microbiological analyses of thermally induced changes in bacterial diversity and abundance of aquifer sediments by Lienen et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (submitted)] accompanied the column experiments conducted by Lüders et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6181-5)]. The observed temperature-related biocenosis changes indicated a potential for bacterial decomposition of harmful substances like hydrogen sulfide or sulfuric acid at elevated temperatures of up to 70 °C [Lienen et al. at http://​angusplus.​de/​en/​publications/​journal-articles (submitted)]. Temperature increases to up to 70 °C induced aerobic organic matter degradation, sulfate reduction and biomass formation as the most important processes in acetate-enriched aquifer sediments. Methanogenesis was observed only within a narrow temperature range around 25 °C and coincided here with the most efficient acetate conversion, compared to lower temperatures [Westphal et al. at http://​angusplus.​de/​en/​publications/​journal-articles (submitted)].

The temperature signals induced by seasonal or shorter term heat storage with cyclic thermal loading can be used for geophysical monitoring of the heat storage site, due to the temperature dependency of propagation parameters such as seismic velocities. Oedometer tests on over-consolidated Tertiary clay as a typical sediment in Northern Germany were run to validate the representation of thermally induced alterations in the mechanical state by the velocity of elastic waves. The clay responded to the cyclic temperature changes with similar cyclic changes of the void ratio and, correspondingly, the longitudinal wave velocity. Along with the cyclic property changes, first results showed a gradual reduction of the void ratio (i.e., thermal hardening) with an increasing number of temperature cycles (V. Feeser, pers. comm.).

Simulations of thermal effects in the subsurface have been conducted for decades in the context of geothermal energy projects (Florides and Kalogirou 2007; Mielke et al. 2014; Yang et al. 2010) or thermal site remediation (Illangasekare et al. 2006). As open systems (well doublets) typically are considered for deep geothermal energy production, significant experience exists considering the availability of high-temperature data and the simulation of these systems. The representation of the technical installations in BTES systems (i.e., borehole heat exchangers in the form of single U-tubes, double U-tubes or coaxial tubes) by a numerical model, however, is challenging. Models so far use analytical solutions embedded as source terms for simplified cases, or double porosity. Computationally efficient OpenGeoSys dual-continuum models of BHE systems were applied to simulate BHE-coupled ground source heat pump systems (Zheng et al. 2016; Hein et al. 2016), where the BHE is represented as 1D domain in the surrounding 3D soil model. For high-temperature heat storage, however, a process-based detailed representation of the near borehole effects of the BHEs is required, which also allows for a representation of the coupled induced processes. Aiming for efficient but fully discretized OpenGeoSys models of multi-BHE heat storage sites representing all BHE components, Boockmeyer and Bauer [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5773-4)] developed geometrically simplified but volumetrically accurate numerical representations of BHE components. Reductions of the computational runtime by factors of up to 50 were achieved while allowing a heat balance deviation of maximum 1% compared to the geometrically accurate representation. This development now allows the fully discretized representation of real scale BTES sites in numerical models for the simulation of coupled induced processes.

The most likely place for the application of heat storage is the urban environment, which at the same time is the most likely place for the presence of soil contaminations. To be able to assess and predict adverse or beneficial thermal effects at contaminated sites, Popp et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5743-x)] and Beyer et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5976-8)] extended the OpenGeoSys code (see Kolditz et al. 2016; Kolditz and Bauer 2004) for the non-isothermal simulation of heat storage in the shallow subsurface in the presence of dense non-aqueous phase liquids (DNAPL) as contaminants (Beyer et al. 2006). The process of temperature-dependent NAPL-dissolution was implemented and verified against a dedicated two-dimensional flow cell experiment [Popp et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5743-x)]. This extended model can be applied to simulate the induced hydraulic, chemical and microbiological effects in scenarios of seasonal heat storage in a trichloroethene (TCE)-contaminated aquifer (Popp et al. 2015).

Fully discretized and process-oriented OpenGeoSys models of BHE systems allow to investigate the influences of individual component design, BHE array design and subsurface parameters on the performance and induced effects of a BTES site [Boockmeyer and Bauer 2014; Boockmeyer and Bauer at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5773-4)]. Using the developed modeling approaches, an increase in the recovered heat to 80% was achieved for a BTES setup with a few tens of BHEs. Single BHEs, in contrast, yielded a heat recovery of only 30%, due to a lack of mutual influence among individual BHEs (Bauer et al. 2015).

First results of simplified numerical simulations of heat storage in TCE-contaminated shallow aquifers by Popp et al. (2015) showed slightly elevated TCE emissions due to increases in groundwater flow and TCE solubility, but also increases in contaminant biodegradation caused by a widening of the TCE plume. Besides interaction with existing groundwater contaminations, a BTES site may, in cases of accidental leakage, pose a threat in itself, since the circulating heat transfer fluid typically contains ethylene glycol or propylene glycol as anti-freezing compounds and benzotriazole as corrosion inhibitor. To protect groundwater resources, heat storage sites in shallow aquifers are therefore subject to a restrictive licensing policy in Germany (Haehnlein et al. 2010). The reactive transport behavior and degradation potential of the compounds mentioned above in a typical Northern German aquifer is therefore studied by both numerical simulations and flow-through column experiments. Results indicate that a compound- and site-specific regulation of the minimum distances between BHEs and water supply wells could significantly improve the compatibility of groundwater protection with an optimized use of shallow aquifers for heat storage and geothermal applications, since substantially shorter distances could be applied if degradation reactions were taken into account (C. Bendtfeld, pers. comm.).

To ensure the environmental and economic sustainability of an increasingly intensive geothermal use of the shallow subsurface especially in urban areas, Vienken et al. (2016) identified the consideration of thermal interactions among thermal subsurface uses as a planning and exploration requirement. Accordingly, Schelenz et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6331-9)] developed and tested a coordinated site investigation approach on the neighborhood scale. Results clearly show that a detailed site characterization allows an enhanced prediction of heat extraction capacities. This in turn can be used for a site-specifically optimized and cost-efficient system design and for a reliable assessment of changes in soil and ground water temperatures induced by the geothermal use. Therefore, novel site characterization approaches and techniques are needed (Vienken et al. 2015). Accordingly, Seibertz et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1016/​j.​jhydrol.​2015.​12.​013)] developed a new methodology for high-resolution vertical characterizations of the heat storage potential in the saturated zone. Fiber optic distributed temperature sensing was used to provide depth dependent measurements of temperature dissipation after thermal excitation. This approach allows the use of existing groundwater monitoring wells and thereby avoids the necessity of BHE installations for the purpose of geothermal characterization. Seibertz et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1016/​j.​jhydrol.​2015.​12.​013)] additionally showed that small changes in lithology, composition and state of compaction have a strong impact on the thermal storage potential, stressing the need for reliable high resolution in situ measurement techniques. A quantitative assessment of established and emerging methods for vertical soil water content profiling and the connected determination of total porosity in the saturated zone was conducted by Vienken et al. (2013). The yet little established direct push-based water content profiling method was identified as beneficial in terms of a high vertical resolution, time efficiency and a possible coupling with other direct push-based sensor probes.

Potential gas storage reservoirs require porous formations with a tight cap rock counteracting a buoyant rise of the stored gases. In the North German Basin, these characteristics are typically present in the sandstones of the Dogger, Rhaetian, Middle Buntsandstein and Rotliegend Formations. The simulation of realistic subsurface energy storage scenarios requires detailed parameterizations of these storage sites and formations. The availability of data is a major issue, as numerous (coupled) parameters on the involved fluid and solid phases are required for the respective temperature and pressure range. Dethlefsen et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-013-2627-1)] assembled a thematic geological database with geological, hydrogeological, and geochemical parameters of reservoirs in the North German Basin at depths > 500 m, using well data from the exploration and production industry and geostatistical analyses. Although being able to determine correlation lengths for a number of variables, a lot of the required data, especially on multiphase flow parameters, is still unknown or not publicly available. The geological model and facies definitions for the North German basin could be based on previous studies by Gaupp (1991) and Hese (2012).

Rock salt caverns used for Compressed Air Energy Storage (CAES) or Power-to-Gas storage (Sterner and Stadler 2014) of hydrogen or synthetic methane are subject to cyclic changes of pressure and temperature due to gas compression and expansion during injection and production, respectively. Numerical models require an accurate representation of the coupled thermal and mechanical effects on the behavior of rock salt under cyclic loading. Khaledi et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi: 10.​1016/​j.​ijrmms.​2016.​04.​010)] used the results from experimental triaxial compression tests to calibrate an elasto-viscoplastic creep constitutive model of rock salt, accounting for temperature effects, dilatancy and damage progress. An extension of this constitutive model has been utilized by Mahmoudi et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5850-8)] to integrate the experimental investigations with the obtained results from the numerical analyses and to update the calibration procedure iteratively. In addition, the effect of uncertainties in the considered values of input parameters on the computational responses is investigated by applying global sensitivity measurements [Khaledi et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1016/​j.​cam.​2015.​03.​049)].

Numerical tools for the simulation of gas storage in porous media or salt caverns and its induced effects need to be able to represent the governing coupled thermal–hydraulic–mechanical and geochemical processes. These tools are the basis for the dimensioning of storage sizes, determining operation conditions and quantifying induced effects. Due to the large variety of possible subsurface situations, a flexible and versatile model approach is required. The ANGUS+ project provides for this requirement by developing appropriate models and by using different coupling schemes.

Pfeiffer et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6168-2)] coupled the THMC simulator OpenGeoSys (Kolditz et al. 2015) with the multiphase-multicomponent simulator ECLIPSE (Schlumberger NV 2015), to make use of the high computational efficiency of the industry simulator ECLIPSE for large-scale grids as well as the thermal and reactive mass transport capabilities of OpenGeoSys. This coupling allows for fully coupled non-isothermal multiphase flow and multicomponent reactive transport simulations, making also use of the geochemical modules described in Li et al. (2014) and Beyer et al. (2012). The coupling scheme was successfully validated using a set of dedicated benchmark simulations [Pfeiffer et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6168-2)].

As a further improvement to the OpenGeoSys framework, Böttcher et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (submitted)] developed a staggered coupling scheme for the governing thermal and mechanical processes which accounts for the thermodynamic behavior of the stored gas in the cavern. Nagel et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5429-4)] introduced Kelvin mapping for the simulation of deformation processes, as this method preserves the tensor character and provides a numerical matrix notation directly corresponding to the original tensor notation.

For the integration of large-scale geological data into numerical models, Wang and Bauer [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6138-8)] implemented a consistent mapping from corner point gridded heterogeneous geological structures and data to finite element grids. This allows the use of these grids in any THMC model based on the Finite Element Method, and facilitates the coupling between OpenGeoSys and ECLIPSE, which requires a joint computational grid. The conversion concept was shown to be able to successfully convert a reservoir scale geological model with multiple layers, pinch-outs and faults, thus accounting for the major features encountered.

A modeling scheme for the safety assessment of gas storage in salt caverns is suggested by Khaledi et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1016/​j.​ijrmms.​2016.​04.​010)]. This approach encompasses a literature-based material parameterization and numerical simulation of the excavation and the subsequent operational cyclic loading. The numerical computation of stress paths, volume convergence, damage propagation and permeability changes simulated in the numerical scenarios allows the derivation of safety limits for the cavern operation.

The actual technical and economic significance is a precondition for the meaningful evaluation of subsurface gas storage scenarios in connection with Power-to-Gas schemes. To achieve this, load curves are required, considering the excess power production from wind or solar power plants as well as the power demand. To this end, Pfeiffer and Bauer (2015) and Pfeiffer et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5958-x)] estimated the amount of 129 × 10⁶ m³ of hydrogen gas (at surface conditions) as amount necessary to secure the electricity demand of 0.82 × 10⁶ GJ in Schleswig–Holstein during one week of a shortfall in both wind and solar power production. Injection and extraction rates were determined taking into account the efficiency of hydrogen re-electrification and assuming an array of five wells in a Rhaetian sandstone reservoir formation in a dome-shaped anticline site. They found that 0.186 × 10⁶ GJ could be supplied by the storage site during a production period of 1 week, which corresponds to about 22% of the required power. Therefore, either more storage sites or, preferably, an increased number of wells would have to be used to cover the storage demand. The anticline site considered would allow a corresponding extension (Pfeiffer and Bauer 2015). To assess the hydraulic impacts of such a large storage site, Bauer et al. (2015) quantified the pressure changes after each injection cycle compared to the initial hydrostatic pressure and the gas phase distribution by means of gas saturation in the reservoir formation around the wells. Pressure changes of more than 1 bar occurred at distances ≤ 7.5 km from a well, resulting in an area affected by pressure increases of about 88 km². The gas phase, however, was found to extend over an area of only about 4.5 km². Hydrogen storage could thus be an option for energy storage on the time scale of days.

Wang et al. (2015) derived a storage scenario from existing gas turbine technical data and daily operation cycles at the operational cavern of the Huntorf CAES power plant in Lower Saxony, Germany, and transferred it to a virtual CAES application in an idealized anticlinal structure. With an estimated production rate of 208.5 kg/s of air for 30 h, equivalent to about 10% of the total air stored in the porous formation, the observed fluctuations of the bottom-hole pressure are within the limits of minimum turbine inlet pressure and maximum safe operation pressure. This hypothetical CAES storage in a porous formation with a permeability of 1 Darcy (equal to 9.86923 × 10⁻¹³ m²) and 20 m thickness could thus provide about 320 MW for up to 41 h. This shows that under favorable reservoir conditions, porous media CAES could provide short-term storage capacities on the order of hours to days.

In a numerical study of the long-term stability of rock salt caverns as energy storage sites, Khaledi et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5970-1)] simulated the stress–strain behavior of rock salt under extreme loading conditions of an operational CAES cavern. A scenario of high-temperature operating conditions showed that an increased creep rate led to accelerated cavern closure, while a scenario of low-pressure operating conditions induced severe effects of dilatancy, damage propagation, and tensile failure in addition to a likewise increased rate of cavern closure.

Pressure increases in the deeper subsurface, induced by large-scale gas storage operations in porous structures or by other types of use, as, for example, fluid injection, may mobilize the resident formation brine. If permeable leakage pathways are present, brine may also be displaced into overlying formations. Rising saline formation water may, in the worst case, reach shallow fresh water aquifers and deteriorate drinking water resources. Studies so far looked at horizontal pressure propagation (e.g., Birkholzer et al. 2009) or vertical brine migration across one sealing formation (e.g., Zhou and Birkholzer 2011). Delfs et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6245-6)] used numerical simulations to assess brine ascent through a series of alternating reservoir and barrier formations. The reservoirs were connected by a permeable vertical pathway, representing a fault. If permeability in the vertical pathway was high, brine rose into the overlying formation, expelled the brine from there further into overlying formations and thereby drove a gradual upward displacement. In the absence of intermediary permeable reservoirs, upward brine movement was reduced or even stopped by the higher density of brine from deep saline reservoirs [Delfs et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6245-6)]. The spreading of a brine leakage once intruded into a shallow fresh water aquifer was investigated by Wiegers and Schäfer (2015) in realistically parameterized 3D simulations. At groundwater flow velocities larger than 0.5 m/d, the effect of aquifer bottom topography on the simulated brine spreading was minor compared to the effect of the groundwater flow. Numerical simulations on realistically parameterized virtual sites were also used to derive possible gas phase distributions after an accidental gas leakage in order to design adequate monitoring methods [al Hagrey et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5784-1)]. Wiegers and Schäfer (2015) worked out that the detection of such plumes is highly reliant on the monitoring setup employed. As has been shown by, for example, Bauer et al. (2006), Beyer et al. (2007) or Rein et al. (2009), monitoring setups for these cases can be adapted to heterogeneous as well as transient conditions and be used to quantify the expected measurement uncertainty.

Leakage of air from a CAES storage site into a shallow aquifer with reducing conditions will typically fuel geochemical reactions due to high oxygen availability. As one case of the possible reactions, Berta et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5985-7)] experimentally characterized pyrite oxidation in dependence of oxygen partial pressure and observed low reaction rates, explained by surface passivation of the pyrite grains. Gas leakages from methane or hydrogen gas storage sites into shallow aquifers can also impair groundwater quality. Hydrogen as a very reactive electron donor will likely induce the reduction of nitrate, sulfate and carbonate, if available (M. Berta, pers. comm.). Methane, in contrast, seems to require an established initial methane oxidizing microbial community to trigger biogeochemical reactions after a comparably long adaption phase following the gas intrusion (Berta et al. 2015a, b).

Geophysical monitoring has been shown to be a successful and promising tool for the control of subsurface gas storage operations (Dethlefsen et al. 2013). Specifically adapted seismic inversion codes, which use full waveform inversion methods (FWI) were shown to be able to resolve small structures with high resolution. To this end, an advanced FWI code was developed by Köhn et al. (2015a, 2015b). In combination with geoelectric and gravimetric methods, an integrated approach was devised, which combines the specific strengths of the individual methods to yield an improved representation of gas distributions in the subsurface, for cases of either porous medium storage or cavern storage. This integrated approach was assessed in synthetic case studies by al Hagrey et al. (2014a) and Benisch et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-014-3603-0)], and subsequently applied to numeric scenarios of gas storage operations and accidental leakages to test for the geophysical monitoring options.

The scenario simulations of hydrogen storage in porous media conducted by Pfeiffer and Bauer (2015) and Pfeiffer et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5958-x)] were accompanied by a numerical assessment of geophysical monitoring methods. Seismic FWI, electrical resistivity tomography (ERT) and gravity monitoring were tested virtually on the fully parameterized, synthetic scenario. Seismic FWI and ERT were shown to be able to identify the thin (order of meters) gas phase in the storage reservoir. P-wave velocities from seismic FWI in a cross-well geometry with a spacing of less than 500 m distance and frequencies up to 1000 Hz were found to allow the resolution of gas phase saturation in the storage reservoir. Optimized data acquisitions and constrained inversions using the highly resolved structures mapped by FWI enhanced the ERT mapping [Pfeiffer et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5958-x)].

The detectability of gas and brine leakages by the integrated geophysical monitoring approach was investigated in a sensitivity study by al Hagrey et al. (2014b). While gravity mapping and seismic FWI performed better in monitoring air leakages in the defined scenario, ERT and transient electromagnetic induction (TEM) were observed to perform most sensitive in the detection of brine intrusions and allowed for the detection of brine leakages in larger depths, compared to air leakages (al Hagrey et al. 2014b). The integrated geophysical monitoring approach using seismic FWI, ERT and gravity monitoring techniques to detect the altered geophysical properties and reduced density caused by the presence of a gas phase was also applied by al Hagrey et al. at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-5784-1) to virtual CAES leakage scenarios: The case of air leakage from a simulated CAES site into a shallow aquifer was considered in a fully parameterized numerical scenario with an accumulating gas phase resulting from a leakage rate of 1 kg/s. While seismic FWI and ERT resolved the anomalies caused by the air leakage 3 years after intrusion into the freshwater aquifer, including gravity mapping improved the detectability to 4 months after start of the leakage.

An early detection of induced cracks can help to identify possible pathways for gas leakages. A seismic monitoring strategy for cavern gas storage in complex salt structures using crack-induced microseismic events was developed by Köhn et al. [at http://​angusplus.​de/​en/​publications/​journal-articles (doi:10.​1007/​s12665-016-6032-4)]. The improved spatial resolution of the FWI method as compared to standard methods allows for a localization of these events, which can be used in cavern inspection surveys.