Degradation of a rock bed thermal energy storage system

Large-scale energy storage is inevitable for the increasing integration of renewable energy and decarbonization of the electricity sector. The potential degradation of a rock bed thermal energy storage system is investigated systematically from both material-as well as system-level perspectives. The performance changes of a 1 MWh th rock bed pilot plant which has been operated up to 675 ° C for 249 cycles (3458 h) is evaluated. We have assessed the potential chemical, structural and thermo-physical changes of the storage material by comparing the properties of rocks as received and rocks retrieved during a post-operational inspection using optical and scanning electron microscopy, energy-dispersive x-ray spectroscopy, dilatometry, densitometry, a vibrating-sample magnetometer, x-ray diffraction and differential scanning calorimetry. However, system-level changes are identified as the main reason for the decreased storage performance even though a 13% decrease in the rock’s heat capacity due to cycling is measured. This is the first study showing the long-term behavior of a rock bed thermal energy storage and no significant changes of the storage material were found. Therefore, the paper demonstrates that thermal storage based on rocks offers promising calendar as well as cycle lifetime but storage material and design should be selected holistically in order to minimize the avoidable performance losses stemming from system-level aspects such as rock rearrangement and thermal ratcheting.


Introduction
With visible impacts on weather and climate extremes, almost 75% of the emissions that have pushed global average temperatures 1.1 °C higher since the pre-industrial age stem from the energy sector [1].In order to limit the long-term increase in average global temperatures to 1.5 °C, a reduction of global carbon dioxide emissions to net zero by 2050 is needed [2], notwithstanding a growing global population [3] with rising incomes pushing up the demand for energy services [4].Associated key measures are energy demand reduction, carbon capture and storage as well as decarbonization of the energy supply [5].Therefore, the pathway calls for scaling up solar and wind rapidly this decade, reaching annual additions of 630 GW of solar photovoltaics and 390 GW of wind by 2030, which is four times the record levels set in 2020 [4].At the same time, the temporal and spatial mismatch between electricity demand and generation makes it then necessary to significantly scale up global energy storage capacities, achieving over 200 GW of capacity or close to a 10x increase compared to today alone in the U.S. by 2050 [6].
Nowadays, their ability to offer great system flexibility by integrating RES into the end-use sectors brings into sharp focus the significant potential of energy storages classified as Power-to-X (P2X) technologies or Carnot Batteries (CB).Whereas P2X technologies are predominantly understood as technologies which use electrolysis in order to store electricity in fuels such as hydrogen (or later ammonia, methane or methanol) [7], CB convert electricity to storable thermal energy that can later be converted to electricity [8,9].CB cover technologies such as pumped thermal [10] and liquid air energy storage [11], all of which are based on thermal energy storages (TES) as a key component.Unlike latent [12] and thermochemical TES [13], sensible TES stores energy by either increasing or decreasing the temperature of a storage medium [14].
Packed beds usually consist of solid storage materials used in sensible TES.Besides their use as both low- [15] and high-temperature storages [16] in CB, prominent application fields of packed beds include Concentrated Solar Power (CSP) plants [17], industrial waste heat recovery [18] and chemical reactors [19].The first industrial-scale TES based on packed beds were based on liquid HTF such as thermal oil [20] or molten salt [21].Well-known examples of gas/solid storages used packed beds in a truncated conical structure to minimize thermomechanical stresses on the walls [17] and with a storage capacity of 6.5 MWh th .Zavattoni et al. reported a 20 m 3 storage for a CSP plant in 2014 [22].The main disadvantages are related to the lower volumetric heat capacity and thermal conductivity of air compared to those of https://doi.org/10.1016/j.applthermaleng.2022 liquid HTF but there is effectively no temperature restriction as well as a promising cost-effectiveness, mainly due to the non-pressurized operation avoiding a heavy construction with complex sealing [23].Solid materials commonly used in TES are crushed rocks [17,24], silica sand [21], alumina spheres [25], ceramic pebbles [26] and perforated concrete blocks [27].Rocks are highly attractive for largescale packed beds thanks to their low cost, competitive volumetric heat capacity and wide availability [28].Their different types, namely sedimentary (e.g.sandstone [17]), igneous (e.g.basalt [29,30] or diabase [16]) or metamorphic (e.g.quartzite [15]), are investigated in several studies.Furthermore, monomineralic materials such as magnetite [31] or hematite [32], found as a primary mineral or as an alteration product, are promising in particular due to their high density.While an extensive review of natural rocks for sensible TES is provided by Allen et al. [33] or Alami et al. [34], waste material such as induction furnace slags [35], demolition waste [36], recycled ceramic [37], copper slags [38], steel slags [39] and electronic arc furnace slags [40] are also the subject of topical discussions.Additionally, a coating to modify the effective thermal properties of packed beds [41], a segmentation of packed beds [42], the thermal shock minimization [43] as well as the addition of phase change material (PCM) for discharge stabilization purposes have been proposed [44].
The suitability of rock types and specific rocks is discussed with respect to compatibility with the selected HTF and thermal cycling [34].The range of heating/cooling rate from 2 [45] to 25 °C min −1 [46] demonstrates the uncertainty surrounding real-life TES cycling, which is also affected by each rock's position inside the packed bed.Additionally, a majority of studies are limited to 110 or less cycles.Several chemical changes under heating may affect the rock degradation, with the most prominent ones being decarbonization [47], dehydration (for example during iron oxidation) [48], deserpentinization of olivine [49],  −  phase transition in quartz leading to anisotropic expansion [50], antiferromagnetic transition of magnetite at around 570 °C [31] or silanol transformation [51].Associated consequences range from an unproblematic surface color change to highly relevant observations such as powder on the surface (dust formation), micro-cracks, disintegration, mass loss and decrease of specific heat capacity.
In conclusion, thermal cycling studies in the literature are limited to laboratory setups which do not consider the (simultaneous) occurrence of real-life phenomena in TES highly relevant for a performance estimation.Factors that can affect the performance include inter alia the thermal ratcheting of rocks, rock rearrangement and debris inside the packed bed.Therefore, the goal of this work is to evaluate the performance changes of a 1 MWh th packed bed TES pilot plant operated up to 675 °C in terms of potential degradation mechanisms.This article presents the degradation of a rock bed TES systematically studied from both material-as well as system-level perspectives.We have assessed the potential chemical, structural and thermo-physical changes of the storage material by comparing the analyses of rocks as received and rocks retrieved during a post-operational inspection.Furthermore, the overall performance of the TES pilot plant is studied in relation to material properties.We draw conclusions on the selection of storage materials and lifetime performance of packed bed TES systems.

Packed bed TES pilot plant
Table 1 summarizes the key characteristics of the packed bed TES pilot plant investigated in this work.

Design concept
Fig. 1 shows the packed bed TES system with a predominantly vertical flow orientation, a further development of a previously built cuboid TES [16].The combination of a conical frustum and hemispherical shaped housing gives a droplet-like shape.The system has been intensively tested in single as well as consecutive cycles [52].Being classified as an unpressurized gas/solid packed-bed storage, the system uses atmospheric air as the HTF and rocks as the storage medium.One novel element is that the heaters, valves, and inlet and outlet pipes are located on top of the storage to avoid additional excavation, simplify maintenance, and allow the rock bed to be installed partially below ground level.
Air enters and exits the rock bed in the vertical direction during charge and discharge by means of separate fans for the charge and discharge processes.The flow scheme illustrated in Fig. 3, being a mix of radial but predominantly vertical air flow direction, uses natural stratification to its advantage.A pipe leading to the bottom of the rock bed makes it possible to reverse the flow direction for charge and discharge to obtain a flat thermocline.This inner pipe acts as an outlet during charge and as inlet during discharge phases, respectively, in both cases for ambient air.
In charge mode, Fan 1 (Fig. 2) provides an air flow by overcoming the flow resistance of the system.After splitting into three pipes, electric resistance heaters increase the air temperature to a set point, up to 675 °C.Air enters an open flow chamber that aims to distribute air flow evenly down through the rock bed as it passes through the conical shaped upper part as well as the hemispherical shaped bottom part.The hemispherical shape gives a small surface area for thermal losses and avoids corner effects.The sloped walls were chosen to reduce mechanical stresses caused by thermal expansion by allowing the rocks to migrate during heating and cooling.The inner pipe is mounted at the center of the storage, giving axial symmetry.Air exits from the center tube at the storage top where the outlet air temperature is measured and is led to the chimney.This charge flow scheme, illustrated in the left side of Fig. 3, generates a steep thermocline because the hot air enters at the top and buoyant forces ensure that the hottest air will remain at the top of the bed.
In discharge mode, the flow is reversed, as shown in the right side of Fig. 3. Fan 1 is switched off and Fan 2 generates a cold air flow which flows through the inner pipe and enters the bottom of the rock bed (Fig. 2).Cold air is forced upward as it recovers heat from the rock bed and exits through the manifold into three outlet pipes that merge into one before being sent to the chimney.During this flow process, the steep thermocline allows the bed to maintain a high outlet air temperature for an extended period of time.
The theoretical thermal capacity  theo of the system before thermal cycling is calculated to be 1007 kWh th for a heater temperature  heater of 600 °C and an ambient temperature of 0 °C, by using Eq. ( 1), assuming a homogeneous temperature in the rock bed and neglecting the contribution of air.The calculation directly based on the measured rock mass  r is favored over its estimation with bed porosity , bed volume  b and rock density  r .
Where  r is the total rock mass (measured as 5394 kg), cp,r is the average heat capacity of the rock from the supplier data sheet (given as 1.12 kJ kg −1 K −1 between 50 and 600 °C) and  is the difference between heater temperature  heater and ambient temperature  amb .
In simple terms, efficiency is the ratio of energy output over the input.The charge efficiency gives an estimate of the capability of the storage unit.In the case of TES, it is the ratio of energy stored over the energy supplied by the heaters and auxiliary components such as a fan, according to the first law of thermodynamics, Eq. (2).
Where  stor is the energy stored in the rocks,  heater is electrical energy input for heaters, and  fan is the energy input for the fan.These are given as: Where ṁ is the mass flow rate of air,  r is the rock temperature,  heater is the heater set temperature, ℎ f is the specific enthalpy of air flow,  is the specific volume of air,  is the pressure drop across the whole rock bed and  is the time.
A full thermal characterization of the packed bed TES system is provided by Knobloch et al. [52].

Construction and data acquisition
The rock bed has a volume of 3.2 m 3 and is filled with a total rock mass of 5394 kg.Whereas the rocks below the inner pipe (219 kg) are characterized by sizes between 16 and 22 mm to improve flow  characteristics around the flow port at the bottom of the rock bed, rocks between 8 and 11 mm investigated in this work (see Section 2.2) are used for the main part of the rock bed (5175 kg).Five different hemispherical and conical lower section, consisting of steel, superwool, rockwool, bricks and concrete, cover the rock material.These insulations not only reduce the heat losses to the ambient but also support the TES structure by balancing mechanical forces.To log experimental data, the main system components are complemented by 53 thermocouples, 8 strain gauges, 2 flow meters, 3 pressure sensors and an energy meter, as shown in Fig. 2. The rock bed houses 38 thermocouples.Most of these stem from 3 vertical sections inside the packed bed at planes 120 • from each other.Each section has a 3 ×3 matrix of thermocouples, where a set of three thermocouples are installed at a distance of 300, 700, and 1100 mm from the top of the rock bed and at three radially positions at each height.

Storage material
Using rocks as the storage material for energy storage in TES is not a new concept, however, there are significant differences in which types of rocks (see classification in Fig. 4) are suitable for this purpose.
We have based our selection of materials on previous studies and our own experiments with different rock types that are readily available (Table 2).The main criteria for the rock storage material are the modal composition and the specific heat capacity of the minerals in the rock, as well as its availability.Furthermore, we have tested the mechanical stability in laboratory experiments of the rocks listed in Table 2. Based on this we have chosen a rock that is dominated by plagioclase (40%-80%).The remainder usually contains mostly pyroxene, biotite and oxides (ilmenite).The chosen rock type in the pilot plant is an igneous, mafic rock from Sweden called diabase.A rock with even higher percentage of pyroxene and oxides could give higher heat storage capacity, due to a higher density and specific heat capacity, but was not readily available for this pilot plant.Our tests showed that magnetite would make the preferred storage material [16] from a property standpoint as has also been described by Grosu et al. [31], but it is 2-3 times more expensive and is delivered with a layer of fines that are magnetically bound to the rock surface.Therefore, diabase was chosen as the storage medium for this experiment.
The irregularly shaped diabase used in the pilot plant has an aspect ratio below 3 and a bulk volume of 0.5 to 1 cm 3 per rock.Particle sizes between 8 and 11 mm were retrieved by sieving through corresponding meshes, meaning that the rocks passed through 11 mm but not 8 mm openings.However, rocks can be longer in some direction due to their irregular shape.Since the pilot plant (see Section 2.1) inauguration in 2019, 249 cycles lasting over 3458 h were run.The operation covered charge, rest and discharge cycles.Samples used in this work were taken during a post-operation opening from an area near the top of the rock bed where maximum temperatures of 630 °C were observed.

Characterization techniques
Table 3 summarizes the sample preparation needed for the characterization techniques introduced in the following subsections.In the following, one cycle is defined as heating up and cooling down within the mentioned temperature range.When consecutive cycles were measured, cycles included a dwell time of 10 min at the highest temperature.Please note that no thermal conductivity measurements (literature values for diabase between 0 and 1000 °C range from 1 to 2.5 W m −1 K −1 [53]) were performed since a good stratification inside the packed bed [52] and no indication for a degradation over the investigated operation time were observable.

Densitometry
Bulk density measurements were based on the Archimedes principle stating that the buoyant force on a submerged object is equal to the weight of the liquid displaced by the object.By measuring the sample weight with a high-precision balance in air and distilled water, Eq. ( 6) was used to determine the bulk density  s with sample mass in air  A , sample mass in water  B , air density  B (22.5 °C) = 0.0012 g cm −3 and water density  A (22.5 °C) = 0.99768 g cm −3 .Considering the factor  = 0.99985 (recommended from manufacturer of the density  measurement kit (Mettler Toledo)), Eq. ( 7) allowed the calculation of bulk volume  s .
Results were obtained by measuring five rocks as received and five cycled rocks (Diabase-b) in two identical experimental setups and calculating mean values.Please note that bulk densities included sample porosity, meaning that skeletal densities are expected to be higher.

Dilatometry
Samples were cut to the size 5*5*15 mm.Diabase-cu's linear thermal expansion between room temperature and 750 °C was measured with a Netzsch 402 CD double/differential dilatometer in four consecutive cycles.

Scanning Electron Microscopy (SEM) and Energy-dispersive X-ray spectroscopy (EDS)
A Hitachi TM 3000 scanning electron microscope was operated in vacuum mode at 15 kV with a high-sensitive semiconductor back scattering electron detector (BSED).In order to define the semi-quantitative chemical compositions of the major minerals composing rock, EDS analysis was performed on Diabase-cr using a Bruker SDD EDS detector.

X-ray powder diffraction (XRD)
XRD was used for the structural analysis of Diabase-p.Using a Rigaku Miniflex 600, data was collected at room temperature between 10°and 90°.

Differential Scanning Calorimetry (DSC)
The heat capacity,  p , of 50-75 mg Diabase-p in argon atmosphere (50 mL s −1 ) was directly measured in four cycles between 20 and 700 °C using a NETZSCH DSC 404C with a continuous heating rate of 5 K/min.

Vibrating-sample Magnetometer (VSM)
A Lake Shore 7407 vibrating-sample magnetometer was used to measure the magnetization  of Diabase-p in magnetic fields up to ±1 T.

Microscopy
A Nikon Eclipse 600 polarization microscope with a Nikon digital camera was used to investigate thin sections of diabase (Diabase-ts).

Results
A performance change of a rock bed TES can be related to a variety of system-level (-) and material-level ( * ) degradation phenomena, whereof the most relevant ones are: -thermal or mechanical degradation of the insulation, -clogging of pipes and filters, -rearrangement of rocks, * thermal degradation of rocks, * disintegration of rocks.While Section 3.1 presents results related to system-level changes, Section 3.2 allows an investigation on the material level by presenting results of the experimental characterization.

System-level changes
Fig. 5 presents temperature profiles inside the rock bed measured in April 2019 (solid) and July 2020 (dashed).For all thermocouples, except the middle one at storage top (z=300, R2), slower temperature increases as well as lower maximum temperatures are observed in the experimental data from July 2020, even though flow rate and heater set temperature are strictly kept at 200 m 3 h −1 and 600 °C, respectively, in both experiments.Evaluated for the same energy input after approximately 15 h of charge, volumetrically averaged rock bed temperatures are estimated to be 438.7 and 430.2 °C, corresponding to a relative difference below 2%.
The results of the measurement of mechanical strain at the concrete outer shell [52] showed highest strains at the storage top but are below the recommended limits for reinforced concrete.This is also observed from the lack of any cracks in the concrete housing.After more than two years of operation, the outer layer of insulation for the storage as well as the pipes is still intact and key components like electric heater, blower and valves showed no failure to date.Additionally, neither debris nor small particles are found inside the packed bed or in the piping.
Furthermore, the rearrangement of rocks is investigated by measuring gaps between steel housing and rock bed, shown in Fig. 6.The inspection of the conical housing of the packed bed top indicates a rearrangement of rocks.The gaps at the cone side walls are measured to be 2.7 to 3.4 cm, increasing from the cone side bottom (transition to hemispherical part) to cone side top (Fig. 6(b)).A gap of 4.6 cm at the rock bed top is measured (Fig. 6(c)).These gaps all suggest that the rock bed has settled and become denser.Pressure loss differences are difficult to evaluate in terms of rock rearrangement since measurement positions are limited to positions before and after the rock bed, not inside.However, analyzing pilot plant data from April 2019 and July 2020 with the same operational parameters showed that pressure loss differences for the whole system were below 0.1 mbar, corresponding to a relative change below 1.1%.4.

Chemical and structural characterization
Samples were taken during a post-operation opening from an area where a maximum temperature of 630 °C was observed.Macroscopic as well as microscopic differences between as received and cycled rocks can be observed in Fig. 7.The rock pieces are still intact and are slightly discolored from a gray to a brownish color on the outside.Despite the presence of some hydrous phases (biotite) the rocks showed good integrity after the heating experiments.In some instances, the hydrous phase is still intact.In thin sections observed under the microscope (Diabase-ts) it can be seen that some of the pyroxene and plagioclase grains show brown staining along fractures.The coloration ranges from weak to strong.In general, the samples do not show pronounced alteration.They are still hard and the minerals have not disintegrated.
The EDS analysis in Table 4 shows the mass percent of elements of three minerals in Diabase-p.The first phase is ilmenite, the second is plagioclase and the third pyroxene.By comparing the composition of the minerals before and after heating, it can be seen that mostly Fe in ilmenite was lost (from 33.99 wt% to 19.78 wt%) and oxygen gained (from 36.01 wt% to 42.10 wt%), based on stoichiometry.The other mineral compositions did not change significantly.
The XRD analysis in Fig. 8 that was performed to see if there are any changes in the mineral assemblages shows that the structural differences between diabase as received and cycled are very small.However, compared to diabase as received, cycled diabase shows higher magnetization, see Fig. 9.The density contributes to the volumetric storage density of the TES.Table 5 presents a bulk density decrease of 1.3% from 3.02 to 2.98 g cm −3 .

Table 6
Summary of different average specific heat capacities cp of diabase.
Supplier data sheet 1.12 Diabase-p as received 0.97 Diabase-p pilot plant 0.86

Thermo-physical properties
The dilatometry up to 750 °C in Fig. 10 shows the linear expansion from diabase-cu as received and cycled diabase-cu.Whereas the diabase as received expanded by approximately 1% in investigated linear direction during the first heating and stayed at 0.1% expansion after being cooled to room temperature, the three following cycles of diabase as received are characterized by an expansion and shrinkage of 0.91%.Similar expansion and shrinkage of approximately 0.90% can be observed for all four cycles with the diabase which was previously cycled in the pilot plant.
Fig. 11 presents the specific heat capacity of diabase-p.Similar to the linear expansion in Fig. 10, behavior of the specific heat capacity of diabase-p during multiple consecutive cycles is addressed.Therefore, the specific heat capacity of diabase-p starting from the as received state is illustrated for four consecutive runs (meaning that the same sample is used for the next run), together with the specific heat capacity of cycled diabase-p from the pilot plant.For consecutive runs, the specific heat capacity of diabase-p as received decreases, starting with a relative decrease of 9.7% from run 1 to run 2 at 600 °C.The specific heat capacity of diabase-p as received approaches that of the cycled diabase-p.Again evaluated for 600 °C, the relative difference between diabase as received in run 1 and the cycled diabase is calculated to be 12.9%.Table 6 provides a comparison of the average specific heat capacities of diabase in this work.

Discussion
This section discusses first the material-level changes and then the system-level changes, both reported in Section 3.
Whereas the observation of a more fine-grained mineral texture for cycled diabase in Fig. 7 can be explained with the natural variability in the grain size and not with a general change of grain size during cycling, the brown coloring that occurs along fractures in pyroxene and some plagioclase of all investigated samples is not due to natural variability.The EDS analysis from Table 4 suggests that it is likely to be related to iron-bearing and hydrous material in the small fractures that become oxidized and alters the decrepitation (break-down) of fluid inclusions in the minerals.The decrease in iron and corresponding increase in oxygen in ilmenite is in agreement with oxidation processes and possibly exsolution of hematite (although not directly observed).However, it does not noticeably weaken the rock pieces, because it is not seen at grain contacts, but within individual mineral grains.The XRD analysis in Fig. 8 confirms that no new mineral phases have formed or original mineral phases disappeared during the cycling runs.Observable small differences occur, here again, due to the fact that different samples are used (confirmed by SEM imaging of multiple samples).On the other hand, the higher magnetization of cycled diabase in the VSM analysis (Fig. 9) is found to be a general trend for multiple samples and is consistent with a stronger magnetism of ilmenite for cycled rocks.
As the dilatometry in Fig. 10 shows, the thermal cycling of diabase is characterized by an irreversible expansion of 0.1%, which occurs in the first heating cycle, and a reversible expansion up to 0.9%, when cycling up to 750 °C.For a typical cycling temperature of 600 °C, a total thermal expansion of approximately 0.8% can be expected.Since the irreversible expansion is relatively small and observed for only the first cycle, different to often reported rock behavior in the literature (for example for Basalt in [29]), there is no increase in the maximum expansion when operating the TES with multiple cycles.That leads to the conclusion that only the first cycle contains irreversible expansion whereas even the cycled rocks from the pilot plant do not show expansion behavior different from the behavior observed for the second to fourth cycle of as received diabase.However, a volume expansion for each rock clearly exists, contributing to the density decrease of 1.3 ± 0.5% (Table 5), together with the release of bound and constitution water.In principle, the reversible expansion of 0.9% may lead to micro-cracks or disintegration of the rocks, although not observed, especially when their thermal expansion coefficients of surrounding materials such as the container are not taken into account during the TES design.The observation of surface dust can also be related to the expansion, additionally to the suspicion that it is a consequence of the oxidation processes.An experimental characterization of such surface dust, localization of debris from degraded rocks (either by inspecting the rock bed bottom or by filtering the outlet air) as well as detailed models using the Discrete Element Method (DEM) [54] or measurement of rock acoustics [50] can help to further assess the challenges related to thermal ratcheting in rock beds.However, the diabase seems well suited to high-temperature TES since its small grain sizes allow heating to a higher temperature before the rock undergoes permanent strain and fracture, in line with literature reporting the unsuitability of coarse-grained rocks [46].
Similar to the thermal expansion behavior, most relevant changes in the specific heat capacity already occur during the first few heating cycles.Based on Fig. 11, we conclude that 249 cycles and 3458 h operating time lead to a value of specific heat capacity which can be referred to as the final and worst state.The 12.8% decrease in average specific heat capacity of rocks between 50 and 600 °C is in good agreement with several cycling-induced specific heat capacity reductions of mafic rocks reported in literature [16,55].The fact that the average specific heat capacity from the supplier data sheet (1.12 kJ kg −1 K −1 between 50 and 600 °C) is 15.5% higher than the measured average specific heat capacity of diabase-p as received is explained with the preheating of diabase-p as received for 6 h at 90 °C, see Table 3.
On system-level, the 2% lower volumetrically averaged temperature inside the rock bed after 15 h of charge with 200 m 3 h −1 /600 °C (Fig. 5) demonstrates that less energy is stored in July 2020 even though the total energy input was identical to the energy input in April 2019.This in good agreement with the observation that the outlet air temperature after 15 h of charge is measured as 62.6 °C in April 2019 while 54.3 °C are measured in July 2020.As a contrast to that, the ambient temperature difference of around 20 °C and in particular the decrease in average specific heat capacity of rocks during cycling (Fig. 11) would suggest that both higher average rock bed temperatures as well as faster temperature rise occur in the experimental data from July 2020.It appears the effect of lower specific heat capacity cannot be considered in isolation since many factors have an influence on final temperature profiles.These include, for example, the higher packing density and the gap between the rock bed and cone side as well as cone top stemming from the rearrangement of rocks presented in Fig. 6.Together with a potential degradation of valves, this increased high-porosity volume at the storage wall might compensate the effect of lower specific heat capacity due to flow channeling.Additionally, the total thermal expansion of up to 1% leads to the suspicion that thermocouples, together with surrounding rocks, are moved by up to 2 cm during cycling.Since pressure loss differences between experimental data from April 2019 and July 2020 are negligibly small, it is concluded that the effect of the general packing of the rock bed is partly compensated by the effect of increased porosity (up to  = 1) at the cone wall.
In order to demonstrate the importance of a thoughtful property choice for the evaluation of a packed bed TES, Fig. 12 evaluates the experimental data recorded in July 2020 in terms of two normalized evaluation parameters, based on properties from the supplier data sheet as well as on measurements for as received and cycled diabase.For this comparison, differences in density (Table 5) and specific heat capacity (Table 6) are considered.It allows for the conclusion that using uncycled properties from a supplier data sheet can lead to a 22% and 32% overestimation of charge efficiency (69.2 vs. 84.4%)and volumetric storage density (2.56 vs. 3.38 kJ/(m 3 K)), respectively.
Consequently, the knowledge of storage material properties is not only crucial for pure TES simulation tasks based on numerical models but also important when existing experimental data is evaluated quantitatively.Nevertheless, both material-as well as system-level changes contribute to the performance loss of a rock bed TES system.
Overall, the observed material-level changes were mostly limited to a decrease of the rock's heat capacity since the they did not disintegrate in any manner and showed an expansion which seems unproblematic for this rock bed TES system.However, the decreased heat capacity cannot explain the performance loss because reported system-level changes appear to dominate the final rock bed TES performance.

Conclusion
A rock bed TES system is characterized by both material-as well as system-level changes over its lifetime.By inspecting a 1 MWh th Fig. 12.Comparison of charge efficiency  CH and volumetric storage density  p for experimental data from July 2020 based on different properties.Please note that the density of diabase as received was used for the supplier data sheet calculation since there was no bulk density given.pilot plant and characterizing as received as well as cycled mafic rocks (diabase), the following findings are derived: • Rocks with anhydrous minerals with a high heat capacity (e.g.pyroxene, olivine, oxides) are ideal for HTTES systems.• The specific heat capacity of the rocks will change, but it reaches an equilibrium state already after the first heating cycles.In the case of diabase a 13% decrease of the heat capacity was observed.The diabase density decreases by 1.3% due to both volume expansion and release of bound and constitutional water/gas.The irreversible linear expansion of 0.1% seems a lesser problem but the reversible expansion of 0.8% for typical cycling temperatures of 600 °C gives cause to use holistic approaches for packed bed TES design and storage material selection.For example matching the expansion of rocks to the expansion of the housing or considering rock rearrangement in an initial design should be taken into account.• Thermal ratcheting needs further research in order to quantify (dust collection, acoustic measurements) and solve (additives, thermal expansion matching, pretreatment of rocks) associated challenges.• Derived properties after 249 pilot plant cycles (3458 h) can be considered as the final and worst state.There is no indication that further cycling will significantly reduce density and specific heat capacity of the diabase.However, the identified performance loss appears to be dominated by system-level changes for the rock bed TES pilot plant investigated.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 2 .
Fig.2.Air flow schematic of the rock TES system[52].The storage is depicted with a cut view from above.System components are mentioned in blue, data acquisition in black.Electric supply is indicated by ''El''.

Fig. 3 .
Fig. 3. Flow schematic in the rock bed during charge(left) and discharge (right) [52].Red indicates high, orange medium and blue low temperature of air.

Fig. 4 .
Fig. 4. Igneous as the relevant rock type for this study classified by composition (left) and origin (right).

Fig. 11 .
Fig. 11.Specific heat capacity  p of Diabase-p in 4 consecutive cycles as well as after pilot plant by means of DSC.

Table 2
Selected rock together with other rocks initially considered.Values for specific heat capacity cp are based on averaging values for rocks with 50 ≤ T ≤ 600 °C.Please note that values are given for rocks without any known thermal cycling.All rocks, except for quartzite, can be considered as igneous rocks.
a also called dolerite or micrograbbo and investigated in this work.

Table 3
Overview of sample preparation.Please note that all listed preparations are done for diabase as received and cycled diabase.Polishing is done with 1 μm grit size.

Table 4
EDS analysis for as received and cycled rocks.Mass percent of elements.Note that all given values are averaged for 3 measurements and area 1, 2, 3 are identified as the three main minerals.

Table 5
Bulk density of Diabase-b by means of Archimedes principle.