Fabrication, Modeling, and Testing of a Prototype Thermal Energy Storage Containment

A suite of technologies across energy sectors needs to be developed and deployed to meet decarbonization targets. In the electricity sector, this suite includes renewable energy resources (e.g., wind and solar), energy storage, transmission enhancement, demand response, and dispatchable, carbon-free generation assets [1]. In the buildings sector, heat pumps and electric appliances can remove direct emissions. In the industrial sector, hydrogen, direct electrification of processes (paired with the decarbonization of the electricity grid), and alternative carbon-free heat sources can replace fossil fuel heat sources [2]. Particle-based thermal energy storage (TES) is one proposed technology that can contribute to the decarbonization of all three of these sectors—see Fig. 1.

Particle-based TES systems can store thermal energy using sensible [3,4] or thermochemical [5,6] methods. Particle-based TES systems show promise in being a cost-competitive option in these sectors due to the low material cost of the storage medium and leveraging established thermal power technologies [7]; these systems could have durations of up to 100 h (i.e., deliver rated discharge power for 100 h) at utility scales (⁠$>100$ MW$e$⁠) [4]. Long-duration energy storage systems such as particle-based TES are seen as a key technology for achieving cost-effective complete decarbonization of the electricity sector [8,9]. Storage durations, defined as the amount of time a device can discharge at rated power, of 50 h and greater are seen as necessary to overcome lulls in variable renewable energy resources all year long [8–10]. Therefore, particle-based TES systems have also been proposed to supply 24/7/365, carbon-free heat for buildings and industrial processes when charged from resources such as solar photovoltaics, wind, and/or concentrated solar thermal resources. Additionally, solid particles have a large operational temperature range making them adaptable to the wide range of temperatures required for integration with industrial processes and different thermal power cycles for grid-scale electricity generation [7]. Particle-based TES systems are already being used to supply district heat in Finland [11].

However, particle-based TES systems are still largely in the laboratory- and pilot-scale development stages. Therefore, there are several key needs including component development, testing, and modeling. Work by these authors has focused on material testing and prototype design, testing, and analysis of various components in particle-based TES systems [7,12–14] and system integration analysis [15]. The storage vessel for the particles is a key component in any particle-based TES system regardless of application. Underground storage methods have been proposed as well as the traditional silo [11,13]. Therefore, in this study, the design, fabrication, and testing of a prototype-scale particle storage silo is presented. Then, the prototype-scale silo is used to validate a transient finite element analysis (FEA) model. The validated model can be used to simulate the performance of a proposed commercial-scale particle storage silo for TES applications.

Finite element analysis methods are ubiquitous in literature for analyzing thermal performance [16–18]. Gage et al. [19] used FEA methods to analyze and optimize the thermal and mechanical performances of a commercial-scale tank for molten chloride salts thermal energy storage systems, another leading thermal energy storage technology. El-Leathy et al. [20] used FEA methods to examine the thermal performance of tank design concepts for particle-based TES applications. Among other findings, the study showed the slow transient response of the layers of insulation highlighting the importance of transient modeling and operating characteristics when analyzing these components computationally [20].

The remainder of the article is organized as follows: Section 2 describes the experimental and computational methods used. Section 3 presents the results of the comparison between the experimental and FEA results for the prototype-scale test and the results of the commercial-scale model. Section 4 discusses the implications of this work in context with other literature and the limitations of the methodology. Section 5 concludes the paper with key takeaways and areas of future work.

A containment silo prototype was built to test the effectiveness of candidate insulation materials in contact with hot particles. Figure 2 shows the entire test station which includes three stages of heaters, two actuating slide gates, and the TES silo prototype. The heaters are used to raise the particles to the desired test temperature, after which point, they are deposited into the TES silo prototype which is the focus of this work. The TES silo prototype is initially at ambient temperatures.

The TES silo prototype consists of a refractory lined 110-gal steel Skolnik drum, the contents of which are shown in Fig. 3. The inner-most refractory crucible has a 5-gal particle capacity, 1-inch thick sidewalls, and consists of Allied Mineral Product’s Petromax ^®700LT (1000 kg/m³). This relatively dense product was selected to resist particle abrasion and help distribute loading in commercial operations. The second 2-inch thick refractory layer is composed of Petromax ^®550 (950 kg/m³), and the third layer consists of lower density INSL2025 ^®(400 kg/m³) shells where the voids are packed with mineral wool to reduce the influence of internal natural convection.

The lids on the two inner-most layers each include a small, concentrically oriented hole that allows hot particles to fall into the inner crucible cavity. Once the entire hot particle volume is deposited, a guiding alumina tube is removed, and the final wedge-shaped block is gently pushed into place to seal the silo.

The FEA methodology is well-established for use in the thermal analysis of various geometries among other types of analyses. The base methodology was identical for both the prototype-scale and commercial-scale simulations; the unique computational geometries and case details are described for each simulation in this section.

The material properties of four insulation materials (i.e., two refractory layers, Calcium Silicate, INSL2025^®), the particles, mineral wool, and concrete are given in Table 2. Particle properties are for a packed bed configuration (i.e., not pure silica).

The prototype-scale case used a geometry identical to that of the experimental test silo. Blocks of INSL2025^®, Petromax ^®700, and Petromax ^®550 are arranged to fill a 110-gal drum. Mineral wool domains fill the voids in the INSL2025 ^®casts. A particle domain in the interior volume was created and filled up to the top accounting for the angle of restitution of silica sand. Stagnant air filled the void in the interior region; the density of the air followed the ideal gas law. The exterior metal drum was defined as the default structural steel material in ansys mechanical. A symmetrical quarter of the geometry was simulated—see the cut plane in Fig. 5.

The prototype-scale FEA model was first initialized uniformly at ambient temperature (25 $∘$C). Then, the particle domain temperature was raised to 716.7 $∘$C over 10 min following the heating profile shown in Fig. 6. The heating profile was determined by the thermocouple (TC) buried in the particle bed (TC #1 in Fig. 7) as the particle bed was filled after being preheated in the Stage 3 batch electric heater. The model then simulated two days of cooling solely caused by heat transfer to the insulation layers and then ambient conditions through natural convection.

The commercial-scale model used the geometry shown in Fig. 8. Similar to the prototype-scale design, two layers of refractory materials compose the core of the insulation layers, and then a lighter, more thermally insulating material (e.g., Calcium Silicate or INSL2025^®) makes up the bulk of the insulation before a structural outer layer (e.g., concrete or metal drum).

The commercial-scale silo represents a 5.51 GWh_th storage silo for an operating range of 300–1200 $∘$C. This temperature delta is congruent with an air-Brayton combined cycle power plant [15]. The temperature range can change significantly depending on application (e.g., different power cycles and/or industrial process heat integration) which would impact the volumetric energy density and costs, among other considerations. The commercial-scale model simulated a theoretical operating profile (i.e., cycling between charge, store, discharge, and hold steps); the operating schedule simulated is shown in Fig. 9. The domain was initialized by finding the steady-state solution to the particles being held at 300 $∘$C. The model simulated ten operating cycles to achieve steady operating cycle results. Otherwise, the model would only simulate the thermal energy of the particles transferring to the insulation (i.e., not leaving the system).

Figure 10 shows the results of the experimental test and the prototype-scale FEA model. The one-dimensional profiles show the initial transfer of thermal energy from the warm particles into the relatively cooler insulation layers. Then, all layers begin to cool after the first 24 h have passed. The results show the FEA model is within 10% of the experimental results at most positions and time-steps. The FEA model diverges slightly from the experimental results as the FEA model predicts all layers cool slightly quicker than the experimental results showed at this point in the simulation. The FEA model did not have any contact resistance model which could decrease this error by making the FEA model more insulative. Therefore, the FEA model results could indicate slightly worse thermal performance than in practice.

This study also highlighted the importance of refinement of key simulation parameter values. The FEA model results varied depending on how the simulation was initialized, the natural convection coefficients assumed, the ambient temperature variance, etc. The results still instilled confidence in the FEA model as the quantitative values and the general temperature profile shape evolution over time predicted by the FEA both matched the experimental results well.

Figure 11(a) shows the average particle temperature through three different cycles—the first, third, and tenth cycles. The particle temperature is controlled during the charge and discharge time-steps as the rate of thermal energy is entering and leaving the particle domain is controlled and constant. Figure 11(b) shows that most of the thermal energy is still retained at the end of five days of storage with the final, steady storage efficiency being near 98%. The improvement in storage efficiency with each operating cycle is due to the increase in thermal mass of the insulation layers during the first five or so cycles. These insulation layers have a significant amount of thermal mass once up to steady operating conditions, highlighting the importance of the inclusion of thermal mass of insulation in storage silo modeling studies.

Figure 11(a) also shows the impact of the inclusion of the 14-hour hold step in the operating cycle. TES systems might have instances in their operating cycles where they must hold “cold” particles prior to recharging. During this time, the FEA model shows thermal energy is transferred from the relatively hotter insulation layers to the colder particles. In the operating cycle chosen in this study, the charging step adds a fixed amount of thermal energy. This leads to the particles starting the store step slightly above 1200 $∘$C, improving the initial state of charge of the silo. Alternatively, the system could use less thermal energy to return particles to 1200 $∘$C. However, this is not an energy recovery mechanism as it leads to increased heat transfer out of the particle domain as the insulation layers are now cooler once the silo is charged with hot particles.

This study developed a useful test station and computational tool for the further research and development of particle-based TES systems. The test station serves as a platform to examine other electric particle heater designs and storage silo materials and associated designs. FEA models are well-established methodologies, but this study sought to validate the computational method with this particular storage media and operating conditions. The combination of prototype-scale testing and component modeling tools is important in the process of increasing the TRL of an emerging energy storage technology. The FEA model could be leveraged to examine alternative designs going forward (e.g., the tradeoff between width and height while accounting for construction and structural constraints).

The FEA modeling work highlighted the importance of initialization and operations in the performance predicted by the FEA. Leveraging some of the experimental data to initialize the domain was critical for ensuring the experimental and FEA results aligned. The commercial-scale results are also highly dependent on the operation cycle assumed—a similar simulation approach detail that was critical to the prototype-scale model validation step. Additionally, the commercial-scale model results required several operating cycles to reach consistent performance. This nuance would be lost if the thermal mass of the storage insulation layers were ignored and if less computationally expensive, steady-state calculations were used. The operating assumptions and/or parameters need to be included in the analysis of all energy storage technologies. The operating cycle used for the commercial-scale model was arbitrary in this study; in the future, an analysis could be coupled with dispatch signals from deployed energy storage and/or a dispatch optimization study for future deployment of particle-based TES.

The FEA model could help inform design studies of particle-based TES systems. While this study focused solely on thermodynamic performance, the economics are important as well. A combined techno-economic analysis could explore the tradeoff between the cost of additional insulation and improved performance. If so, the impact of the storage silo’s design parameters on overall system performance would be prudent to quantify.

Lastly, the model has limitations that could impact the results presented. First, contact resistances are ignored. Contact resistance values could be experimentally measured between each material and then implemented to improve the accuracy of the model. The material properties could be improved with additional data on the temperature-related effects to heat capacity and thermal conductivity of the various material for which they were assumed isothermal. The commercial-scale storage silo model could also be improved through the inclusion of more nuanced geometry (e.g., structural rods). Additionally, the static mesh meant the commercial-scale simulation did not capture the physical process of filling and draining silo during the charging and discharge operating cycles, respectively. The commercial-scale storage silo model ignores any wind effects due to the high-level of variability in wind conditions to which the silo could be subjected. Wind effects would increase the convective coefficient applied to the external faces of the silo decreasing thermal efficiency. Outside of refined thermal performance modeling methods, additional research can focus on coupling this work to thermo-mechanical and/or structure analysis as well as overcoming abrasion; which are all non-trivial challenges to particle-based TES commercialization.

The fabrication, experimental testing, and modeling reported in this study are critical to the development of particle-based TES systems. The study found TES storage silos can be designed to be greater than 97% efficient even at 1200 $∘$C storage temperatures after five days of storage. However, a full techno-economic analysis would be required to determine the “optimal” storage insulation design (i.e., thickness) while accounting for material costs and structural limitations. For example, commercial molten salt tanks used in concentrated solar power plants lose about 1 $∘$C per day while storing much lower temperature materials—a similar economic tradeoff could lead to lower performance from commercial particle-based TES silos. Furthermore, the study highlights the influence of simulation details (e.g., operating cycles and initialization) when validating and analyzing results produced by FEA models. The experiments and modeling together provide a complementary platform to test and analyze TES silo designs, materials, and performance for a range of applications. The applications of which are cross-sector decarbonization of the energy system.

This work was authored in part by the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Funding was provided in part by the DOE Advanced Research Projects Agency–Energy (ARPA-E) DAYS Program under Work Authorization Number 18/CJ000/07/05. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government.

There are no conflicts of interest.

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

$g$ =

gravitational constant (9.81 m/s$2$⁠)

$k$ =

thermal conductivity (W/m-K)

$i$ =

surface of storage, $i∈I=top, side, bottom$

$T$ =

temperature (K)

$x→$ =

position vector (m)

$cp$ =

heat capacity (J/kg-K)

$hi$ =

heat transfer coefficient of side $i$ (W/m$2$-K)

$mp$ =

total mass of particles in silo (kg)

$ni$ =

exponential coefficient of side $i$

$Ci$ =

scaling coefficient of side $i$

$Li$ =

characteristic length of side $i$ (m)

$Tinf$ =

ambient temperature (K)

$Ts$ =

surface temperature (K)

$c¯p,p$ =

average heat capacity of particles (J/kg-K)

 $Tpfinal$ =

average temperature of the particles in the silo at the end of the ‘store” step in the operating cycle (⁠$∘$C)

Ra =

Rayleigh number

   $β$ =

volume expansion coefficient (1/K)

$ν$ =

viscosity (m$2$/s)

$ρ$ =

density (kg/m$3$⁠)

$p$ =

particle property or variable

$s$ =

surface property or variable

  $inf$ =

ambient conditions

ARPA-E =

advanced research projects-energy

DOE =

Department of Energy

FEA =

finite element analysis

GWh$th$ =

gigawatt-hour thermal

LDES =

long-duration energy storage

MW$e$ =

megawatt electric

NIST =

National Institute of Science and Technology

NREL =

National Renewable Energy Laboratory

TC =

thermocouple

TES =

thermal energy storage

TRL =

technology readiness level