*Article* **Storage Capacity in Dependency of Supercooling and Cycle Stability of Different PCM Emulsions**

**Stefan Gschwander \*, Sophia Niedermaier, Sebastian Gamisch, Moritz Kick , Franziska Klünder and Thomas Haussmann**

> Fraunhofer Institute for Solar Energysystems, Heidenhofstr. 2, 79110 Freiburg, Germany; sophia.niedermaier@ise.fraunhofer.de (S.N.); sebastian.gamisch@ise.fraunhofer.de (S.G.); moritz.kick@ise.fraunhofer.de (M.K.); franziska.kluender@ise.fraunhofer.de (F.K.); thomas.haussmann@ise.fraunhofer.de (T.H.)

**\*** Correspondence: Stefan.gschwander@ise.fraunhofer.de

**Abstract:** Phase-change materials (PCM) play off their advantages over conventional heat storage media when used within narrow temperature ranges. Many cooling and temperature buffering applications, such as cold storage and battery cooling, are operated within small temperature differences, and therefore, they are well-suited for the application of these promising materials. In this study, the storage capacities of different phase-change material emulsions are analysed under consideration of the phase transition behaviour and supercooling effect, which are caused by the submicron size scale of the PCM particles in the emulsion. For comparison reasons, the same formulation for the emulsions was used to emulsify 35 wt.% of different paraffins with different purities and melting temperatures between 16 and 40 ◦C. Enthalpy curves based on differential scanning calorimeter (DSC) measurements are used to calculate the storage capacities within the characteristic and defined temperatures. The enthalpy differences for the emulsions, including the first phase transition, are in a range between 69 and 96 kJ/kg within temperature differences between 6.5 and 10 K. This led to an increase of the storage capacity by a factor of 2–2.7 in comparison to water operated within the same temperature intervals. The study also shows that purer paraffins, which have a much higher enthalpy than blends, reveal, in some cases, a lower increase of the storage capacity in the comparison due to unfavourable crystallisation behaviour when emulsified. In a second analysis, the stability of emulsions was investigated by applying 100 thermal cycles with defined mechanical stress at the same time. An analysis of the viscosity, particle size and melting crystallisation behaviour was done by showing the changes in each property due to the cycling.

**Keywords:** phase-change material; dispersion; thermal-mechanical stability; viscosity; supercooling; nucleating agent; cold storage; battery cooling

## **1. Introduction**

PCM emulsions or, more generally, PCM slurries (PCS) are heat transfer fluids consisting of a PCM that is dispersed in a fluid (in this study, water). Compared to pure water, they showed a high storage capacity due to the usage of the PCM melting and solidification process, in addition to the sensible heat capacity of the water and the PCM. To utilise this benefit, they have to be used within the PCM melting and crystallisation temperature range [1]. This offers the advantage of reduced volumes for energy storage due to a greater storage density [2]. PCS, in general, offer the advantage of being liquid-independent of the PCM's state of matter. This property makes it possible to convey PCS in hydraulic systems equally in both physical states of the PCM. Hence, PCS can act as heat transfer fluids and as a storage fluid. With this functionality, PCS contributes as a storage material to thermal systems that are operated within small temperature differences, such as cooling or temperature-buffering applications. PCS can transfer this heat homogeneously at an almost constant temperature [3]. The fine dispersion of the PCM in water leads to a

**Citation:** Gschwander, S.; Niedermaier, S.; Gamisch, S.; Kick, M.; Klünder, F.; Haussmann, T. Storage Capacity in Dependency of Supercooling and Cycle Stability of Different PCM Emulsions. *Appl. Sci.* **2021**, *11*, 3612. https://doi.org/ 10.3390/app11083612

Academic Editor: Ioannis Kartsonakis

Received: 18 March 2021 Accepted: 12 April 2021 Published: 16 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

high surface-to-volume ratio that contributes to a high heat transfer rate between water and PCM [4,5].

Over time, researchers have investigated different paraffins for use as PCM in emulsions. Already, in the 1990s, Inaba and Morita [6] investigated the application of tetradecane with a melting temperature of 14.75 ◦C in water as a PCM emulsion for cold storage. A mixture of anionic and non-ionic emulsifiers stabilised the emulsion. Paraffins are available over a wide range of melting temperatures. Additionally, melting temperatures between that of pure paraffins can be achieved by mixing two different paraffins [7]. He and Setterwall [7] mixed, for example, n-tetradecane and n-hexadecane and found melting temperatures between those of the pure materials. They also found that a mixture of about 90 mol% n-tetradecane and 10 mol% n-hexadecane showed a melting temperature at 2 ◦C. This makes it possible to adapt the melting point to the demand of different applications by choosing a paraffin with an appropriate melting temperature or mixed paraffins with higher and lower melting temperatures.

According to Montenegro and Landfester [8], n-alkanes with an even number of carbon atoms in the range from 12 to 26 crystallise in the triclinic form, and those with an odd-numbered carbon chain show a phase transition from liquid to a metastable orthorhombic rotator phase and, in the second step, at a lower temperature, a transition to the stable phase. It was also observed that even-numbered n-alkanes show a phase transition to the rotator phase when emulsified in the nanometre scale. Hagelstein und Gschwander [9] observed this as well for octadecane emulsions. It was found that the transition to a rotator phase accounts for 70–75% of the total crystallisation enthalpy. Beside this change in the phase transition, it was observed that, due to the confined volume of the paraffin droplets, it was crystallising at a temperature below its bulk phase-change temperature from solid to liquid. This is due to a separation of the natural given crystallisation seeds [10]. This hysteresis effect is called supercooling. Zhang et al. [11] reported that a PCM emulsion with a melting temperature of 18 ◦C had to be cooled down to 7 ◦C to obtain the full latent heat. Nucleation agents were used to supress the supercooling. Huang et al. [5] developed, for example, a PCM emulsions based on the paraffin blends RT6 with a melting temperature of 6 ◦C, RT10 (10 ◦C) and RT20 (20 ◦C) from Rubitherm Technologies GmbH. Non-ionic emulsifiers—in particular, ethoxylated fatty alcohols—were used for the stabilisation of the PCM droplets/particles. A paraffin with a melting point at 50 ◦C was used as a nucleating agent that was able to decrease the supercooling. After 100 cooling/heating cycles in a test rig, the DSC measurements of the samples revealed an increased supercooling again and indicated an unstable nucleation agent. Lu and Tassou [12] reported the development of a PCM slurry for cooling applications. A paraffin with a high melting temperature was used as a nucleation agent to reduce the supercooling of the emulsified PCM. The storage stability was tested by storing a sample over eight months and applying 50 heating and cooling cycles. The stability was analysed by comparing the DSC curves before and after the experiment. The emulsion showed no reduction of the effectiveness of the nucleation agent, but separation or creaming of the paraffin droplets was observed, which was, in further experiments, supressed by using thickeners. Zhang et al. [2] dispersed n-octacosane with a melting temperature of 60 ◦C for use in heat storage applications. As emulsifiers, they used non-ionic surfactant different combinations of Tween and span emulsifiers with a HLB of 12, as well as hydrophobic SiO<sup>2</sup> particles, as nucleating agent to reduce supercooling. The stability was tested first by storing samples for up to six months and, second, by applying five freezing and melting cycles in which the sample was heated to 70 ◦C and then cooled down again to room temperature. The breaking ratio, which is the ratio between the separated PCM to the whole tested volume, was used to quantify the stability. After the storing experiment, a creaming was observed but no clear trend in the changes of the droplet sizes. In the thermal cycling experiment, separated oil was observed on the samples, and a breaking ratio was determined in a range between 40% and almost 90% for the different emulsions.

Delgado et al. [1] stated that the smaller and more homogenous the particle size distribution is, the more stable a PCM emulsion is for thermal–mechanical cycling and storage time-induced effects. Zhang et al. [13] described the development of PCM emulsions with particle sizes in the nanometre scale using hexadecane as the PCM. The emulsion showed a high stability against phase separation, which was tested under repeated thermal cycles in which the emulsion was cooled below the solidification temperature in an ice bath and heated up again by taking it out of the ice bath and exposing it to room temperature at 25 ◦C. A high stability was proven by applying this procedure. The DSC measurements revealed supercooling, which was reduced in further experiments by using SiO<sup>2</sup> particles, but nevertheless, the supercooling was still in the range of about 5–10 K.

Vorbeck et al. [14] investigated a PCS based on microencapsulated paraffin with a melting range between 22 and 26 ◦C and a melting enthalpy of 50 kJ/kg. The PCS had a fraction of 30 wt.% of microcapsules and was tested on stability within a hydraulic test rig in which the PCS was cycled. The test rig contained plate-type heat exchangers and a centrifugal pump. In every cycle, the slurry was cooled below the solidification temperature and heated up above the melting temperature of the PCM again. Nearly 100,000 cycles were applied without observing a significant number of broken microcapsules. The storage capacity was investigated in a storage tank containing 5 m<sup>3</sup> of the slurry, and the results were compared with experiments using water as the storage fluid in the same storage tank. It was shown that it was possible to store 55% more energy than using water, which leads to a 95-min-longer discharge duration. Biedenbach et al. [15] reported the experimental investigation of a PCM emulsion with 35 wt.% n-octadecane in a 500-L storage tank. The storage capacity was compared to water, which was also investigated in the same storage tank. The PCM emulsion showed a supercooling of 5.7 K. Compared to water, the emulsion offered, in the best-case scenario, an improvement of the storage capacity by a factor of 2.6, operating it between 20 and 27 ◦C, 2.3 in a temperature range between 22 and 29 ◦C and 1.7 between 15 and 29 ◦C. The stability of the emulsion was proven by particle size measurements done with samples taken after filling the storage tank after three and after six days of operation. The particle size was decreased and more uniform after three days of moving to slightly bigger sizes after six days. Delgado et al. [16] investigated a PCM emulsion with a low-cost paraffin and solid content between 59% and 61% in a stirred 46-litre tank with a coiled heat exchanger. A 3.5–5.5 times higher heat transfer coefficient was observed when the emulsion was stirred with 290–600 rpm compared to the non-stirred tank. It was found that the stored energy was 80% higher for an operation temperature between 30 and 50 ◦C and 40% higher when operating it between 30 and 60 ◦C in comparison to water as the storage fluid. Morimoto et al. [17] investigated the influence of the degree of supercooling on the storage capacity in n-hexadecane and n-octadecane emulsions. A freezing ratio was defined as the relation of the heat stored when the emulsion was cooled to a certain temperature before it was molten again to the total storage capacity of the full-phase transition. Dependent on the cooling temperature, the freezing ratio was in a range between 30% and 60% for emulsions having a mean particle size of 0.25 µm.

Research into PCS development is concentrated on the stabilisation of PCM in emulsions against phase separation (creaming) and on the suppression of supercooling. In many publications, the stability against thermal cycling is tested without applying mechanical stress at the same time. There is little information on how supercooling influences the storage capacity of PCM emulsions in comparison to standard heat transfer fluids. In this study, analyses on the storage capacity for emulsions using different paraffins as PCM in comparison to pure water are presented. All investigated emulsions are prepared using the same formulation, including the nucleation agent. Furthermore, an analysis of the stability is undertaken using repeated thermal cycles and shearing the emulsion with a defined shear rate at the same time (thermal–mechanical stress). The influence of supercooling on the storage capacity is analysed by enthalpy curves obtained via DSC measurements. The degradation due to thermal–mechanical stress is tested by using a rotational rheometer

determining the change of viscosity with time, measuring the particle size distribution and the enthalpy via DSC before and after the test.

#### **2. Materials and Methods**

#### *2.1. Materials*

All presented PCM dispersions were prepared by using linear alcohol ethoxylates from Sasol Germany GmbH, Fritz-Staiger-Straße 15, 25541 Brunsbüttel, German as surfactants. The PCMs, which were emulsified, were Parafol 16-97, Parafol 17-97, Linpar 17, Parafol 18-97, Parafol 19-90 and Parafol 19-97, as well as Parafol 20Z, which were also received from Sasol Germany GmbH. For the paraffins of the Parafol trademark, the first digit stands for the number of carbon atoms in the hydrocarbon chain and the second digit for the purity of the material. For example, Parafol 19-90 and Parafol 19-97 are *n*-alkanes with 19 carbon atoms. These two materials differ solely in their purity, with 90% and 97%, respectively. According to Sasol, Linpar 17 and Parafol 20Z are paraffins with high purities, but there are no exact percentages available. In the following, Parafol is abbreviated to P and Linpar to L. The PCM RT 27 is a blend of paraffins that was manufactured by Rubitherm Technologies GmbH, Imhoffweg 6, 12307 Berlin, Germany with an unknown composition. A second blend was examined using 40 wt.% Parafol 16-97 and 60 wt.% Parafol 18-97, resulting in a melting temperature of 18.7 ◦C.

#### *2.2. Formulation of the PCM Dispersions*

The PCM dispersions were all formulated with 35 wt.% of PCM emulsified in deionised water. The concentration of 35 wt.% of PCM shown in previous internal studies a good compromise between capacity, viscosity and stability. For stabilising the PCM droplets/particles in water, 5 wt.% surfactant was added to the formulation. To decrease the supercooling of the PCM dispersion, a derivate of paraffin wax from Sigma-Aldrich Chemie GmbH, Kappelweg 1, 91625 Schnelldorf, Germany with a melting temperature of 105 ◦C was added to the PCM as a nucleation agent before dispersing it in water. The mixture of PCM and nucleation agent was heated to 105 ◦C and stirred for 30 min; then, it was cooled down to 80 ◦C before emulsifying it in water.

The mixture of Paraffin, nucleation agent, emulsifier and water was dispersed for 3 min using a dispersion machine, magicLAB®, from IKA®-Werke GmbH & Co.KG, Janke & Kunkel-Str. 10, 79219 Staufen, Germany which worked with a rotor–stator principle at a speed of 20,000 rpm. In a following step, the PCM emulsion was homogenised in a high-pressure homogeniser APV 2000 from SPX®, Konrad-Zuse-Straße 25, 47445 Moers, Germany at 200 bar. In total, 200 g of every sample was produced containing 70 g of PCM. The initial weight was done with an accuracy of ±50 mg for PCM and water and with ±5 mg for the emulsifiers and nucleation agent. The maximum weighting error was below ±0.07%, and therefore, this error was neglected for the storage capacity analyses.

#### *2.3. Characterisation*

Every PCM dispersion was analysed in terms of its particle size distribution, melting and crystallisation behaviours, enthalpy and stability against thermal–mechanical stress.

The particle size distribution was analysed using a laser diffraction particle analyser, LS13320, from Beckman Coulter GmbH, Europark Fichtenhain B 13, 47807 Krefeld, Germany. The measured values represented the particle diameter. The particle size distribution was measured after production and after the thermal–mechanical stability test.

A DSC (Q200 from TA Instruments, Altendorfstraβe 12, 32609 Hüllhorst, Germany) was used to determine the melting/crystallisation temperature and the enthalpy of the fusion/crystallisation. The onset values were chosen as the melting/crystallisation temperature due to its independency of the heating rate. To identify the suitable heating rate for the dispersions, a heating and cooling rate test was performed conformal to the procedure developed in IEA ECES Annex 29/SHC Task 42 [18]. According to the heating rate test, the heating rate was determined to 1 K/min. The measurements were performed in Tzero

hermetic aluminium crucibles at the determined heating rate, and the temperature dynamic range varied depending on the melting point of the PCMs. The sample volume was 10 µL. A flow of dry nitrogen was used to purge the measuring cell with 50 mL/h during the measurements. The error of the DSC measurements was given with ±5% for the enthalpy values and ±0.1 K for the temperature values.

For using PCM emulsions as heat or cold storage fluids, it is important to consider the degree of supercooling, especially when comparing it with water. Supercooling is increasing the necessary temperature range that must be applied to use the full potential of SC. In principle, the larger this temperature range is, the lower is the benefit of PCM systems in comparison to storages using water as the storage fluid. For the determination of SC, a temperature range for charging and discharging is defined as the ∆Tload. The SF is given as the volume-specific sensible and latent heat within the temperature range of ∆Tload in comparison to the volume-specific sensible heat of water in the same temperature range. For the comparison, a value of 4.19 kJ/(kg·K) as the specific heat capacity for water and a density at 20 ◦C of 0.998 kg/L were used [19]. As the material datasheets showed no difference in the densities of the used pure paraffins, density measurements were performed using the Parafol 18-97 emulsion. The density was measured 4 times at 35 ◦C using an Anton Paar DMA 4500 divece from Anton Paar GmbH, Hellmuth-Hirth-Straße 6, 73760 Ostfildern, Germany. It was determined as 0.872 kg/L with a standard deviation of ±12 kg/L considering all 4 measurements. In this study, different temperature points were chosen to obtain different ∆Tload to determine the SC and SF (in the following, called methods a–d):


Tm-end is determined as the temperature where the difference of enthalpy between the crystallisation and melting curve is smaller than 0.2 kJ/kg the first time when subtracting the melting curve from the crystallisation curve, starting from the onset temperature of melting to higher temperatures. Tc-end is determined in the same way but by decreasing the temperature. Tm-start is chosen visually as the temperature of the melting curve were the slope starts to steepen.

The dependency of the available SC on supercooling is illustrated in Figure 1. The graph shows a DSC measurement of a microencapsulated paraffin that is partly crystallised. The material shows two phase transitions. In the first measuring cycle, the material was cooled down to 20 ◦C and heated again to 40 ◦C. In the second cooling, it was only cooled down 25 ◦C and heated from this temperature again (black, dashed line). As visible in this case, the available SC was determined by the crystallisation curve. In the following analyses, this behaviour was considered to determine the SC based on method c for the different PCM emulsions.

The rotational rheometer MCR 502 from Anton Paar GmbH, Hellmuth-Hirth-Straße 6, 73760 Ostfildern, Germany was used to determine the thermal–mechanical stability of the PCM emulsions. The setup allowed us to apply a defined heating and cooling rate to heat and cool a sample within a defined temperature range, as well as to stress the sample with a defined shear rate at the same time. The stability was tested in a concentric cylinder measuring system, consisting of a measuring body and a measuring cup, wherein the measuring body was rotated. This setup was suitable for low viscous fluids and a relatively large amount of sample could be applied. By using cylindrical geometry, the errors caused by the evaporation of water from the sample were reduced, having a minor effect on the measured results. During the measurements, evaporation was additionally reduced with a solvent trap. The PCM emulsion was heated up and cooled down at a defined rate of 0.05 K/s over 100 thermal cycles while being stressed with a constant shear rate of 100 1/s. This test will be called the thermal–mechanical stability test in the following. Before and after the thermal–mechanical stability test, a sample of the cycled PCM emulsion was analysed via DSC, and the particle size distribution was measured using the LS13320.

**Figure 1.** Example for the melting behaviour of a partly crystallised microencapsulated paraffin. The PCM shows two phase transitions in the crystallisation process. The 1st transition is at the higher temperature.

#### **3. Results and Discussion**

The used paraffins are characterised via DSC as well. Table 1 summarizes all DSC results and Figure 2 illustrates the variety of phase transitions and enthalpies within the dependency of temperature. All analysed paraffins show supercooling in the form of a hysteresis. Parafol 16-97 (P16-97), Parafol 18-97 (P18-97) and Parafol 20Z (P20Z) are linear *n*-alkanes, with an even number of 16, 18 and 20 carbon atoms. In the DSC, they show a single-phase transition or two-phase transitions (P20Z) that are very close together when crystallising. The phase transitions are sharp and take place in a narrow temperature range. Linpar 17 (L17), Parafol 17-97 (P17-97), Parafol 19-90 (P19-90) and Parafol 19-97 (P19-97) are *n*-alkanes with an odd number of carbon atoms. These *n*-alkanes show a clearly visible second phase transition in both the melting and crystallisation processes.

**Table 1.** Summary of the most important properties of bulk PCM that are taken for the formulation of PCM dispersions.


**Figure 2.** Enthalpy curves of the bulk materials measured with 1 K·min−<sup>1</sup> via DSC. The shortcuts are given in Table 1.

P17-97 has an onset temperature Tonset (melting point) of 21.2 ◦C, L17 of 20.5 ◦C, P19-90 of 30.5 ◦C and P19-97 of 31.2 ◦C. These paraffins show a smaller hysteresis than most of the pure even-numbered paraffins, but the melting and crystallisation also take place over a broader temperature range. Combined with the two-phase transitions, the temperature difference in an application must then be comparatively large to use the full latent of the storage capacity. For example, P17-97 requires a temperature range of 8 K when only the first phase transition is used and 15 K when the second phase transition should be used as well. The comparison between P19-90 and P19-97 shows how impurities influence the storage capacity and the phase transition temperature. The total storage capacity is reduced from 222 kJ/kg for P19-97 to 207 kJ/kg for P19-90, and the melting temperature (Tonset) decreases from 31.2 ◦C to 30.5 ◦C.

RT27 is a purchasable blend of *n*-alkanes with paraffin, having a different numbers of carbon atoms in their hydrocarbon chains. The melting point Tonset is at 24.5 ◦C, which is close to that of P18-97 (27.3◦C). The phase transition is in a temperature range that is larger than that of P18-97, and the enthalpy for the blend RT27 is 153 kJ/kg, which is 70 kJ/kg less than that of P18-97 (223 kJ/kg).

For the comparison with L17 and P17-97, which are both high-purity odd-numbered paraffins, a blend was mixed using 40 wt.% P16-97 and 60 wt.% P18-97. It has a melting point Tonset of 18.7 ◦C and an enthalpy of 151 kJ/kg. All the investigated blends show a small hysteresis but a broad temperature range for the phase transition in comparison to the pure PCMs.

P20Z is an *n*-alkane with twenty carbon atoms in the hydrocarbon chain. Its melting point Tonset is at 34.3 ◦C, and it offers an enthalpy of 211 kJ/kg. During crystallisation, it also shows two phase transitions, of which the first is the transition to the rotator phase and the second is the transition to the stable triclinic crystal form.

Table 1 summarises the paraffin´s most important properties from the DSC measurements.

#### *3.1. PCM Emulsions*

#### 3.1.1. Particle Size Distribution

All PCM are dispersed with 35 wt.% of PCM in water and homogenised. Through homogenisation, a particle size distribution in the sub-micrometre range is reached, and thus, a high stability against creaming is obtained. Figure 3 depicts the particle size distributions for all emulsions. All sizes are below 1 µm. The emulsions with P19-90 and P20Z have bimodal particle size distributions in the range between 45 and 600 nm, as well as P19-97; all others show a monomodal size distribution in the range between 60 and 500 nm.

**Figure 3.** Particle size distribution of PCM dispersions with 35 wt.% of different PCMs dispersed in water.

#### 3.1.2. Storage Capacities

The DSC results are summarised in Table 2 for all the PCM emulsions. The melting enthalpies are in a range between 53 und 61 kJ/kg, which is on the level expected for oddnumbered paraffins due to the melting enthalpy of the first-phase transition of the used bulk paraffins. The even-numbered paraffin emulsions show a melting enthalpy that is less than 35% of the available bulk paraffin´s melting enthalpy (see HfPCM-1st-35% in Table 2). This is due to the induction of a second-phase transition through the emulsification process. In comparison to the pure paraffins, all emulsions show a high degree of supercooling (∆T = Tm-offset − Tc-onset) that is listed in Table 2 and shown in Figure 4 for all the emulsions. Supercooling is reduced with an increasing melting temperature and samples with lower purity. Additionally, those emulsions prepared using paraffins with an odd number of C atoms in their hydrocarbon chain show a lower degree of supercooling than emulsions with even-numbered paraffins.


**Table 2.** Summary of the main DSC properties for the PCM dispersions, with ∆T = Tm, offset − Tc,-onset.

**Figure 4.** Degree of supercooling and onset temperature of melting for the different emulsions.

Figure 5 depicts the enthalpy curves for all PCM emulsions dependent on the temperature. The graph reveals that all emulsions have at least two-phase transitions in the measured temperature range, except the one prepared with a mixture of P16-97 and P18-97. It might have a second transition below the measured temperature range, but this could not be proved, as it was not possible to cool further down due to the freezing of the emulsion´s water fraction. The curves also show that those emulsions prepared with mixtures of paraffins (P16-97/P18-97, RT27) show broader melting and crystallisation ranges than those using purer paraffins. Figure 6 compares the enthalpy curves of L17, P17-97 and the mixture of P16-97/P18-97 emulsions. It is visible that L17 and the P16-97/P18-97 mixture offer almost the same enthalpy differences in the main transition range, whereas P17-97 melts at a higher temperature and shows a higher change in the enthalpy. Both the bulk PCM L17 and P17-97 have around 210 kJ/kg, while the blend of P16-97/P18-97 is 28% lower with 151 kJ/kg. After emulsification, the enthalpy of this mixture is only about 4.5% lower than that of the L17 and P17-97 emulsions (57.9 k/kg versus 60.6 kJ/kg). A similar effect can be seen for the P18-97 and RT27 emulsions. While the enthalpy of bulk RT-27 is 31% lower, it is only 14% lower when it is emulsified.

Despite the melting enthalpy for the bulk P16-97 being slightly higher (217 kJ/kg) than that of P20Z (211 kJ/kg), emulsified P20Z shows, with 59.0 kJ/kg, a slightly higher melting enthalpy than the emulsified P16-97, which provides 55.6 kJ/kg. One reason for this is the higher share of enthalpy within the second phase transition at a lower temperature for the P16-97 emulsion, which has 12.2 kJ/kg versus 8.6/kJ for the P20Z emulsion.

The storage capacities and temperature ranges for the emulsions, calculated according to methods a–d, are shown in Figure 7 and listed in for method a and b in Table 3 and and for method c and d in Table 4. As expected, the larger the chosen temperature range, the more storage capacity is available, except for the analyses considering a 6 K temperature difference and taking the crystallisation curves into account according to method c. These analyses revealed a much lower storage capacity, as expected from the enthalpy change of the melting curve. Here, those materials with a high degree of supercooling show the lowest storage capacity. The comparisons of the storage capacities with those that would be obtained using water as the storage medium are shown in Figure 8. Here, a massive reduction of the benefits due to the supercooling phenomena and due to broad-phase transition temperature ranges is visible. The analyses that take only the melting curves into account (method a) show the maximum possible SF operating a system at the phase transition from a solid to liquid without any supercooling. Considering only the melting curve with a temperature range of 6 K (according to method b), the benefit is reduced for all materials but shows the realistic maximum benefit compared to water as the storage material in a real storage application. With the given supercooling and melting, as well as crystallisation ranges, P20Z, P19-97 and P17-97 perform best with a SF around 2.5. Considering the 6-K limit according to method c, RT27 offers a 1.8 higher SF than P18-97. Using the full temperature range (method d), P18-97 shows with 2.22 versus 2.06 (RT27) a 7.8% higher SF.

**Figure 5.** Enthalpy curves of the PCM emulsions with a PCM fraction of 35 wt.% each. For each curve, the enthalpy is set to 0 at Tm-end.

**Figure 6.** Enthalpy curves for the P16-97/P18-97, L17 and P17-97 emulsions.

**Figure 7.** Storage capacities according to the different temperature ranges: m-curve (method a), m-curve 6 k (method b), m-c-curve (method c) and m-c curve (method d) (see Section 2.3 Characterisation. Storage factors and temperature differences for the melting process and to acquire the total storage capacity for the first phase-transition m-curve (method a), m-curve 6 k (b), m-c-curve (c) and m-c curve (d) (see Section 2.3 Characterisation).

**Table 3.** Data for the different emulsions for the analyses according to methods a and b.


**Table 4.** Data for the different emulsions for the analyses according to methods c and d.


**Figure 8.** Storage factors and temperature differences for the melting process and to acquire the total storage capacity for the first phase-transition m-curve (method a), m-curve 6 k (b), m-c-curve (c) and m-c curve (d) (see Section 2.3 Characterisation).

#### 3.1.3. Thermal–Mechanical Stability

P16-97, L17, P18-97, P19-90 and P20Z emulsions are chosen for the stability evaluations. All tested PCM emulsions passed this stability test over 100 thermal–mechanical cycles without optically visible changes like phase separation or creaming. Figures 9–13 depict the results for the measured emulsions. The rheological data show different changes in the viscosity over the 100 cycles, which come with the changes in the particle size distribution. In the viscosity measurements, the lower viscosity value indicates the emulsion´s viscosities at 35 ◦C and the upper value those at 5 ◦C, respectively. At the beginning of the tests, the emulsions show viscosities between 20 and 50 mPa·s, except P19-90, which ranges between 25 and almost 95 mPa·s. Looking at the development of viscosity over the 100 cycles, the samples exhibit quite different changes. For P16-97, the viscosity at 35 ◦C is kept more or less constant at 20 mPa·s, while the upper value at 5 ◦C rises very linearly with every cycle, up to 80 mPa·s after 100 cycles. This comes with a change in the particle size distribution, from a monomodal distribution at the beginning to a bimodal distribution with a fraction of smaller sizes and a larger fraction at bigger sizes. L17 shows a rise in viscosity at 35 ◦C from 20 to 32 mPa·s and from almost 55 mPa·s to 80 mPa·s after 100 cycles at 5 ◦C. The particle size increase is minor, and the peak is slightly narrower than at the beginning. P18-97 keeps the lower viscosity at 20 mPa·s, and the upper values are also more or less constant with higher values for single peaks. No significant change in the size distribution is observed. The PCM emulsion P19-90 shows the highest viscosity values in this test. It reaches over 120 mPa·s after 100 cycles. The size distribution also becomes bimodal, but contrary to P16-97, the share that turned to smaller sizes is larger than the one grown to bigger sizes. P20Z also shows a viscosity between 20 and 55 mPa·s at the beginning of the cycle test. While cycled, both values rise, but the upper value at 5 ◦C rises faster than the lower value at 35 ◦C.

**Figure 9.** Thermal–mechanical stability test of Parafol 16-97: (**a**) viscosity measured between 5 and 35 ◦C over 100 cycles, and (**b**) particle size distribution before and after 100 cycles.

**Figure 10.** Thermal–mechanical stability test of Linpar 17: (**a**) viscosity between 5 and 35 ◦C over 100 cycles, and (**b**) particle size distribution before and after 100 cycles.

**Figure 11.** Thermal–mechanical stability test of Parafol 18-97: (**a**) viscosity between 5 and 35 ◦C over 100 cycles, and (**b**) particle size distribution before and after 100 cycles.

**Figure 12.** Thermal–mechanical stability test of Parafol 19-90: (**a**) viscosity between 5 and 35 ◦C over 100 cycles, and (**b**) particle size distribution before and after 100 cycles.

**Figure 13.** Thermal–mechanical stability test of Parafol 20z: (**a**) viscosity between 5 and 35 ◦C over 100 cycles, and (**b**) particle size distribution before and after 100 cycles.

Figure 14 shows the comparison of enthalpy curves determined before and after this test. It is visible that all samples show minor changes in the crystallisation curve except P18-97. P16-97, L17 and P19-90 show slightly but not significantly deeper solidification temperatures after the test that might be an indication of an unstable nucleation seed. P20Z reveals an increase in the phase transition temperatures from a liquid to solid, which could be induced by an increase in particle size. Therefore, in general, it can be stated that those emulsions showing minor changes in the viscosity and in the particle size distribution also revealed minor changes in the phase transition behaviour.

**Figure 14.** Comparison of the melting and crystallisation behaviours before and after the thermalmechanical stability test.

#### **4. Conclusions**

In this study, even- and odd-numbered paraffins were used to prepare PCM emulsions and mixtures. The DSC analyses of the paraffins showed the two-phase transitions of the odd-numbered paraffins. The even-numbered paraffins shown in the DSC were only onephase transitions, except for P20Z, which was a two-phase transition that was close together in the measurements. Furthermore, all used paraffin mixtures showed less enthalpy than the pure materials and a larger temperature range for the phase transition from a solid to liquid and, also, less supercooling. While all used pure paraffins offered total enthalpies in the range between 207 and 223 kJ/kg for the phase transition from a solid to liquid, the enthalpies of the mixtures P16-97/P18-97 and RT27 were 151 kJ/kg and 157 kJ/kg, respectively. This was on the same level with the enthalpies of the first transition of the measured odd-numbered paraffins, which were about 75–77% of their total enthalpy.

The supercooling and storage factors of the PCM emulsions were influenced by the emulsified PCM and nucleation agent. The odd- and even-numbered paraffins showed different crystallisations in the bulk but similar behaviours when emulsified. Odd-numbered paraffins kept the two-phase transitions of the bulk materials, while the investigated evennumbered paraffins showed a second phase transition (different from the bulk materials) that indicated a phase transition from the liquid to a rotator phase (first transition) and from the rotator phase to the stable phase (second transition). This behaviour led to a reduction of up to 27% of the bulk paraffin´s phase transition enthalpy for the first transition, as the emulsion P16-97, for example, showed. It was shown that the purer materials exhibited more supercooling when emulsified than the blends or mixtures. The results revealed that, due to this effect, the mixtures and blends were better in terms of the SF when small temperature intervals were given, which was shown by using a temperature difference of 6 K (comparing RT27 with P18-97 and P19-90 with P19-97). Additionally, the mixture of P16-97 and P18-97 had a SF within a 6-K range that was on the same level as P17-97 (2.11 versus 2.19). If a large melting enthalpy of a bulk material suggests a large SF for a PCM emulsion, these examples showed that this could not be transferred to the PCM emulsion directly, because the supercooling and limited temperature intervals of the applications needed to be taken into account and strongly influenced the SF and, therefore, the advantage of the emulsions over conventional heat transfer fluids.

One aspect of the analysed PCMs that could be transferred to the PCM emulsions was its tendency of supercooling with a decreasing melting temperature. The shorter the PCM´s hydrocarbon chain, the larger the degree of supercooling. Furthermore, the supercooling of the PCM emulsions was higher using purer materials. This effect was shown for Parafol 19-90 and Parafol 19-97, as well as for RT27 and P18-97 emulsions.

Therefore, in terms of costs, it is favourable to have knowledge about the specific temperature interval of an application to consider whether it is more cost-efficient to use not-so-pure but cheaper PCMs providing a good storage capacity within small temperature intervals or to choose the more expensive, purer PCMs to provide more storage capacity when the usable temperature interval is a bit broader.

All the investigated PCM emulsions had similar particle sizes. This indicated that, for the investigated emulsions, the particle size distribution was dependent on the formulation rather than on the used PCM. The thermal–physical properties were measured before and after cycling and showed comparable results for P16-97, L17, P18-97, P19-90 and P20Z emulsions. None of them showed visible instabilities, such as creaming or phase separation, after 100 thermal–mechanical cycles in the rheometer, but a major or minor change in the behaviours and properties was visible for the measured viscosities, particle size distributions and crystallisation behaviours. Especially, the changes in the viscosity and particle size were significant and revealed differences. In comparison, the changes in the melting and solidification behaviours were not so pronounced, and it might be more difficult to draw conclusions about stability from these DSC results. Nevertheless, there were some changes visible that might indicate, e.g., the stability of the crystallisation seed and, due to decreased supercooling, the influence of a change in the particle sizes. In general, all used measuring techniques could reveal a change of properties and, therefore, instability before creaming or phase separation was visible and, therefore, provided an indication of the long-term stability or instability of the PCM emulsions. So far, it is not known how the observed changes affect the stability of these PCM emulsions under real usage or what they mean in terms of useful life applications. Therefore, in future works, the changes of the material properties due to the thermal–mechanical stability test undertaken in the rheometer have to be compared with the stabilities obtained under real applicational conditions, with the aim of developing a model to predict the real stability of PCM emulsions on the basis of this accelerated lab test.

**Author Contributions:** Conceptualization, S.G. (Stefan Gschwander) and S.N.; methodology, S.G. (Stefan Gschwander) and S.N.; investigation, S.G. (Stefan Gschwander) and S.N.; resources, S.G. (Stefan Gschwander); data curation, S.G. (Stefan Gschwander) and S.N.; writing—original draft preparation, S.G. (Stefan Gschwander) and S.N.; writing—review and editing, S.G. (Sebastian Gamisch), M.K., F.K. and T.H.; visualization, S.G. (Stefan Gschwander) and S.N.; supervision, S.G. (Stefan Gschwander) and T.H.; project administration, S.G. (Stefan Gschwander); funding acquisition, S.G. (Stefan Gschwander). All authors have read and agreed to the published version of the manuscript.

**Funding:** The research was funded by the German Federal Ministry of Economics and Energy via several research projects (grant no.: 03ESP357).

**Institutional Review Board Statement:** Excluded.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to restriction given by partners.

**Acknowledgments:** The authors are grateful to the Project Management Jülich for the administrative support.

**Conflicts of Interest:** No conflicts of Interest.

## **Nomenclature**


#### **References**


**David Vérez , Emiliano Borri , Alicia Crespo, Boniface Dominick Mselle , Álvaro de Gracia , Gabriel Zsembinszki and Luisa F. Cabeza \***

> GREiA Research Group, Universitat de Lleida, Pere de Cabrera s/n, 25001 Lleida, Spain; david.verez@udl.cat (D.V.); emiliano.borri@udl.cat (E.B.); alicia.crespo@udl.cat (A.C.); boniface.mselle@udl.cat (B.D.M.); adegracia@diei.udl.cat (Á.d.G.); gabriel.zsembinszki@udl.cat (G.Z.) **\*** Correspondence: luisaf.cabeza@udl.cat; Tel.: +34-973-003-576

**Abstract:** The use of latent heat thermal energy storage is an effective way to increase the efficiency of energy systems due to its high energy density compared with sensible heat storage systems. The design of the storage material encapsulation is one of the key parameters that critically affect the heat transfer in charging/discharging of the storage system. To fill the gap found in the literature, this paper experimentally investigates the effect of the macro-encapsulation design on the performance of a lab-scale thermal energy storage tank. Two rectangular slabs with the same length and width but different thickness (35 mm and 17 mm) filled with commercial phase change material were used. The results show that using thinner slabs achieved a higher power, leading to a reduction in the charging and discharging time of 14% and 30%, respectively, compared with the thicker slabs. Moreover, the variation of the heat transfer fluid flow rate has a deeper impact on the temperature distribution and the energy charged/released when thicker slabs were used. The macro-encapsulation design did not have a significant impact on the discharging efficiency of the tank, which was around 85% for the operating thresholds considered in this study.

**Keywords:** thermal energy storage; latent heat thermal energy storage; phase change materials (PCM); macro-encapsulation; rectangular slab; experimental study

## **1. Introduction**

The use of thermal energy storage (TES) has been proved as an effective way to enhance the penetration of renewable energy into energy systems. Amongst all thermal storage technologies, latent heat thermal energy storage (LHTES) received the attention of several researchers over the last decade due to its high energy density and the wide range of applications [1]. Buildings, for example, represent one of the most common applications of the integration of LTHES as an active or passive system [2–4]. For the active systems, TES can be used in HVAC components or systems to balance the supply of domestic hot water and heating/cooling demand when renewables are used [5,6], or to reduce the energy consumption through peak load shifting [7], or free cooling techniques [8]. On the other hand, passive systems are directly integrated into the building envelope to reduce the energy demand [9,10]. Other common applications where LTHES can be integrated include solar thermal power plants, such as concentrated solar power (CSP) [11], solar cooling applications [12], district heating or cooling [13], waste heat recovery [14], solar process heat [15], or cryogenic applications [16].

The principle behind LHTES is the use of phase change materials (PCM) as the storage medium, allowing to store thermal energy at a nearly constant temperature exploiting the latent heat during the phase transition, for which the most common one is from solid to liquid to minimize the impact of volume expansions [17]. One of the weaknesses of PCM is its low thermal conductivity that negatively affects the thermal power involved in the charging and discharging processes of the energy storage system. Indeed, this represents

**Citation:** Vérez, D.; Borri, E.; Crespo, A.; Mselle, B.D.; de Gracia, Á.; Zsembinszki, G.; Cabeza, L.F. Experimental Study on Two PCM Macro-Encapsulation Designs in a Thermal Energy Storage Tank. *Appl. Sci.* **2021**, *11*, 6171. https://doi.org/ 10.3390/app11136171

Academic Editor: Ioannis Kartsonakis

Received: 25 May 2021 Accepted: 30 June 2021 Published: 2 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

one of the main challenges facing the implementation of PCM in various applications. However, different strategies and techniques that can be used to improve thermal conductivity were investigated in the literature. The main solutions that were extensively studied are the increase in the convection coefficient of heat transfer by means of dynamic systems, the addition of particles (such as carbon elements, metallic particles, and nanoparticles), the inclusion of PCM in a metallic matrix, and the increase in the heat transfer area by using fins, and micro and macro-encapsulation [18–20].

On one hand, PCM micro-encapsulation allows increasing heat transfer surface between the PCM and the heat transfer fluid. However, for PCM microencapsulation, complex and expensive processes are needed, such as spray drying (physical method) or interfacial polymerization (chemical methods) [21]. On the other hand, macro-encapsulation requires a simpler making process resulting in a lower cost [22]. Furthermore, larger sizes of the container also allow an increase in the mechanical stability of systems [23]. Macro-encapsulated PCM can be designed with different geometries mainly based on rectangular [24], cylindrical [25,26], and spherical shapes [27] that can be adapted to different applications. The effect of the design of macro-encapsulation on the heat transfer performance is mostly analyzed by numerical analysis with only a few experimental studies available in the literature, highlighting a research gap. Amongst the experimental studies available, Erlbeck et al. [28] and Al-Yasiri and Szabó [29] experimentally investigated the thermal behavior of concrete blocks with different shapes of microencapsulated PCM. Ismair and Moraes [30] numerically and experimentally evaluated spherical containers made with different geometries and materials and filled with PCM for cold storage domestic applications. This paper experimentally analyzes the effect of two different geometries of macro-encapsulated PCM in rectangular slabs on the performance of an energy storage tank. The analyzed TES tank is part of the generic heating system designed for the EU funded project SWS-HEATING (GA 764025). In particular, the PCM tank is used in the system as a thermal buffer to store the solar energy at low-grade temperature (15 ± 5 ◦C) to be supplied to a novel seasonal TES based on selective water sorbent materials. To the best of the authors knowledge, very few experimental studies on PCM tanks with rectangular slabs were published in the literature. One of the first papers was published by Moreno et al. [31] in which the performance of a TES tank filled with commercial PCM encapsulated in rectangular slabs was compared with the same tank filled with water. The results showed that the energy storage capacity of the tank filled with PCM was increased by 35.5% compared with the same tank filled with water. Another study published by D'Avignon and Kummert [32] reported the results of experimental tests performed to study the behavior of a real-scale PCM storage at different operating conditions. One of the main conclusions from the study was that the PCM hysteresis and sub-cooling effects deviate the expected behavior from the experimental results. Liu et al. [33] used the experimental results obtained from the testing of a PCM tank filled with rectangular slabs containing a PCM with a sub-zero melting temperature (−26.7 ◦C) suitable for refrigerated transport, and glycol as heat transfer fluid. The developed model was based on a one-dimensional approach considering the temperature variations along direction of the heat transfer fluid showing a good agreement with the test. All experimental studies mentioned were carried out using a fixed design of the PCM tank without changing any boundaries related to the geometry or the configuration of the storage tank.

However, the geometrical design of the PCM encapsulation has a large influence on the thermal behavior of the PCM affecting the melting and the solidification process, and consequently the heat transfer [34]. In the case of rectangular shapes, the aspect ratio (height to width ratio) is a parameter that has to be taken into account in the design of TES tanks [21]. This paper shows for the first time a comparison based on experimental results of the thermal behavior of two different designs of macro-encapsulation of rectangular PCM slabs. The behavior of a thermal energy storage tank was analyzed using commercial PCM slabs with different thicknesses. The comparison of the two designs was done in terms of temperature profile, heat transfer rate, and energy obtained during the discharging process. The main results obtained from the experimental tests reported in this paper can be used as a reference for institutions and manufacturers to optimize future designs of PCM tanks. of PCM tanks. **2. Materials and Methods** 

terms of temperature profile, heat transfer rate, and energy obtained during the discharging process. The main results obtained from the experimental tests reported in this paper can be used as a reference for institutions and manufacturers to optimize future designs

#### **2. Materials and Methods** *2.1. Materials*

#### *2.1. Materials* The PCM selected in this experimentation was PlusICE S15 (hydrated salt), supplied

The PCM selected in this experimentation was PlusICE S15 (hydrated salt), supplied by PCM products, United Kingdom [35]. The main thermophysical properties of this material are shown in Table 1. Moreover, water was used as the heat transfer fluid (HTF). by PCM products, United Kingdom [35]. The main thermophysical properties of this material are shown in Table 1. Moreover, water was used as the heat transfer fluid (HTF). **Table 1.** Thermophysical properties of PlusICE S15

**Table 1.** Thermophysical properties of PlusICE S15 [35]. **Properties Value** 

*Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 3 of 14


#### *2.2. Experimental Set-Up 2.2. Experimental Set-Up*  The experiments presented in this paper were carried out at the laboratory of the

The experiments presented in this paper were carried out at the laboratory of the GREiA research group at the University of Lleida in Spain, in a set-up designed to test and characterize latent heat TES systems for mid-low temperature applications (−20 ◦C < T < 100 ◦C). Figure 1 shows a detailed schematic diagram of the experimental set-up composed by a 25 L inertia water tank, whose temperature is controlled by a vapor compression cooling unit (Zanotti model GCU2030ED01B [36]) of 5 kW cooling power, two immersion thermostats (OVAN TH100E-2kW [37], and JP SELECTA-1kW [38]). The set-up also integrates: two variable speed pumps, used to control the flow and inlet temperature at the TES system; and a flow meter Badger meter type ModMAG M1000 [39] with an accuracy of ±0.25 % of the actual value, and the latent heat TES storage. The connections between components were joined using 0.5" diameter copper pipes insulated with 18 × 0.9 mm polyurethane tubes. The data acquisition system used consisted of 3 STEP DL-01 data logger [40] connected to a computer that integrates a system control and data acquisition software (SCADA) developed in InduSoft Web Studio [41]. The data recording interval was set to 10 s. GREiA research group at the University of Lleida in Spain, in a set-up designed to test and characterize latent heat TES systems for mid-low temperature applications (−20 °C < T < 100 °C). Figure 1 shows a detailed schematic diagram of the experimental set-up composed by a 25 L inertia water tank, whose temperature is controlled by a vapor compression cooling unit (Zanotti model GCU2030ED01B [36]) of 5 kW cooling power, two immersion thermostats (OVAN TH100E-2kW [37], and JP SELECTA-1kW [38]). The set-up also integrates: two variable speed pumps, used to control the flow and inlet temperature at the TES system; and a flow meter Badger meter type ModMAG M1000 [39] with an accuracy of ±0.25 % of the actual value, and the latent heat TES storage. The connections between components were joined using 0.5" diameter copper pipes insulated with 18 × 0.9 mm polyurethane tubes. The data acquisition system used consisted of 3 STEP DL-01 data logger [40] connected to a computer that integrates a system control and data acquisition software (SCADA) developed in InduSoft Web Studio [41]. The data recording interval was set to 10 s.

**Figure 1.** Schematic view of the experimental set-up used to perform the experimentation. **Figure 1.** Schematic view of the experimental set-up used to perform the experimentation.

Figure 2 shows the PCM storage tank connected to the experimental set-up. The tests were carried out with two different PCM macro-encapsulation designs, namely, ThinICE and FlatICE (Figure 3). The containers were made in HDPE. The external dimensions of

each design are reported in Figure 3, characterized by presenting similar length and width (A and B), but different thickness, with FlatICE dimensions being double that of ThinICE (C). Furthermore, the (D) dimension reveals that the use of thin macro-encapsulation enabled a larger distance between the slabs, increasing the space that allows circulating the HTF through the TES tank. Considering the aforementioned dimensions shown in Figure 3, the use of ThinICE encapsulation allowed fitting a larger number of slabs inside the tank, but less amount of latent storage material compared with the FlatICE, as shown in Table 2. (A and B), but different thickness, with FlatICE dimensions being double that of ThinICE (C). Furthermore, the (D) dimension reveals that the use of thin macro-encapsulation enabled a larger distance between the slabs, increasing the space that allows circulating the HTF through the TES tank. Considering the aforementioned dimensions shown in Figure 3, the use of ThinICE encapsulation allowed fitting a larger number of slabs inside the tank, but less amount of latent storage material compared with the FlatICE, as shown in Table 2. and FlatICE (Figure 3). The containers were made in HDPE. The external dimensions of each design are reported in Figure 3, characterized by presenting similar length and width (A and B), but different thickness, with FlatICE dimensions being double that of ThinICE (C). Furthermore, the (D) dimension reveals that the use of thin macro-encapsulation enabled a larger distance between the slabs, increasing the space that allows circulating the HTF through the TES tank. Considering the aforementioned dimensions shown in Figure 3, the use of ThinICE encapsulation allowed fitting a larger number of slabs inside the tank, but less amount of latent storage material compared with the FlatICE, as shown in

Figure 2 shows the PCM storage tank connected to the experimental set-up. The tests were carried out with two different PCM macro-encapsulation designs, namely, ThinICE and FlatICE (Figure 3). The containers were made in HDPE. The external dimensions of each design are reported in Figure 3, characterized by presenting similar length and width

Figure 2 shows the PCM storage tank connected to the experimental set-up. The tests were carried out with two different PCM macro-encapsulation designs, namely, ThinICE

*Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 4 of 14

*Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 4 of 14

**Figure 2.** Latent heat TES. **Figure 2.** Latent heat TES. **Figure 2.** Latent heat TES.

**Figure 3.** ThinICE and FlatICE slabs encapsulation. Dimensions in millimeters. **Figure 3.** ThinICE and FlatICE slabs encapsulation. Dimensions in millimeters. **Figure 3.** ThinICE and FlatICE slabs encapsulation. Dimensions in millimeters.



The temperature inside the PCM storage was measured using nine Pt-100 class B, IEC 60751 standard type, with an accuracy of (0.3 + 0.005 · T). The sensors were fixed as shown in Figure 4 to the external surface of three different PCM slabs placed at the bottom, middle, and top of the tank, respectively. Moreover, two additional Pt-100 class A IEC 60751 standard type with an accuracy of (0.15 + 0.002 · T) sensors were placed at the inlets and outlets of the storage tank. The temperature inside the PCM storage was measured using nine Pt-100 class B, IEC 60751 standard type, with an accuracy of (0.3+0.005·T). The sensors were fixed as shown in Figure 4 to the external surface of three different PCM slabs placed at the bottom, middle, and top of the tank, respectively. Moreover, two additional Pt-100 class A IEC 60751 standard type with an accuracy of (0.15+0.002·T) sensors were placed at the inlets and outlets of the storage tank.

**Figure 4.** Temperature sensors location inside the storage tank. **Figure 4.** Temperature sensors location inside the storage tank.

*Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 5 of 14

#### *2.3. Methodology 2.3. Methodology*

The experimental tests consisted of performing four different charging and discharging processes to evaluate the effect of the PCM macro-encapsulation design and the flow rate on the temperature distribution, heat transfer rate, and energy stored/released. At least three repetitions of each process were performed to ensure repeatability. A summary of the flow rates and temperatures used in the experimentation is shown in Table 3. Furthermore, the heat losses in the worst-case scenario analyzed represent 4% of the charging/discharging energy, therefore the analysis of heat losses was not included in the paper. To perform a charging process, HTF was first circulated through the PCM tank until The experimental tests consisted of performing four different charging and discharging processes to evaluate the effect of the PCM macro-encapsulation design and the flow rate on the temperature distribution, heat transfer rate, and energy stored/released. At least three repetitions of each process were performed to ensure repeatability. A summary of the flow rates and temperatures used in the experimentation is shown in Table 3. Furthermore, the heat losses in the worst-case scenario analyzed represent 4% of the charging/discharging energy, therefore the analysis of heat losses was not included in the paper.

all sensors inside the tank reached a temperature of 5 ± 1 °C. Then, the HTF inlet temperature was set at 25 ± 1 °C and the flow rate was set to the corresponding value of the **Table 3.** Summary of the main parameters of the processes.


**Process Slab Type Flow Rate [L/min] Temperature [**°**C] Average Initial Temperature [**°**C] Code**  Charge ThinICE 2 25 5 ± 1 C\_ThinICE\_2L Charge ThinICE 4 25 5 ± 1 C\_ThinICE\_4L Charge FlatICE 2 25 5 ± 1 C\_FlatICE\_2L Charge FlatICE 4 25 5 ± 1 C\_FlatICE\_4L To perform a charging process, HTF was first circulated through the PCM tank until all sensors inside the tank reached a temperature of 5 ± 1 ◦C. Then, the HTF inlet temperature was set at 25 ± 1 ◦C and the flow rate was set to the corresponding value of the experiment shown in Table 3. The charging process was considered complete when the HTF temperature at the outlet of the tank reached 25 ◦C. To perform a discharging process, HTF was first circulated through the PCM tank until all sensors inside the tank reached a temperature of 25 ± 1 ◦C. Then, the HTF inlet temperature was set at 5 ± 1 ◦C and the flow rate was set to the corresponding value of the experiment in Table 3. The discharging process was considered complete when the HTF at the outlet of the tank reached 7 ◦C. This value was used instead of 5 ◦C because a minimum temperature difference of 2 ◦C was assumed between inlet and outlet of the storage tank as a constraint from the demand side.

**HTF Inlet** 

#### *2.4. Uncertainties Analysis*

The impact of the uncertainties in the calculated parameters from the different measurements was evaluated by performing an uncertainty analysis using the Kline McClintock method. The uncertainties of the different monitored parameters are shown in Table 4. HTF specific heat capacity and density were calculated following the correlations presented in Equations (1) and (2) [42]:

$$
\rho\_{HTF} = 1.38 \cdot 10^{-5} \cdot T\_{HTF}^3 - 5.63 \cdot 10^{-3} \cdot T\_{HTF}^2 + 3.6 \cdot 10^{-3} \cdot T\_{HTF}^1 + 1000 \tag{1}
$$

$$\mathcal{C}p\_{\rm HTF} = 2.69 \cdot 10^{-9} \cdot T\_{\rm HTF}^4 - 6.63 \cdot 10^{-7} \cdot T\_{\rm HTF}^3 + 6.67 \cdot 10^{-5} \cdot T\_{\rm HTF}^2 - 2.67 \cdot 10^{-3} \cdot T\_{\rm HTF}^1 + 4.21 \tag{2}$$

**Table 4.** Uncertainties of the different parameters involved in the analyses of the present study.


By applying Equation (3) to the different parameters [43], the uncertainties of the HTF thermophysical properties (density and specific heat) as well as of the heat transfer rates and total stored/released energy were estimated. The uncertainty of the HTF thermophysical properties and heat transfer rates was estimated at each registered time step, and then the mean value was used. Table 5 shows the average uncertainties of the HTF density, specific heat, heat transfer rate, and stored/released energy during the different processes carried out:

$$\mathcal{W}\_{\mathcal{R}} = \left[ \left( \frac{\partial \mathcal{R}}{\partial \mathbf{x}\_1} \cdot \boldsymbol{w}\_{\mathbf{x}\_1} \right)^2 + \left( \frac{\partial \mathcal{R}}{\partial \mathbf{x}\_2} \cdot \boldsymbol{w}\_{\mathbf{x}\_2} \right)^2 + \dots + \left( \frac{\partial \mathcal{R}}{\partial \mathbf{x}\_n} \cdot \boldsymbol{w}\_{\mathbf{x}\_n} \right)^2 \right]^{1/2} \tag{3}$$

where *W<sup>R</sup>* is the estimated uncertainty in the final result, *R* the function which depends on the measured parameters, *x<sup>n</sup>* is the different independent monitored parameters, and *w<sup>x</sup>* is the uncertainties associated to those independent parameters.


**Table 5.** Estimated uncertainties of the HTF thermophysical properties, heat transfer rate, and cumulated energy.

#### **3. Results and Discussion**

#### *3.1. Temperature Evolution during the Charging Process*

Figure 5 shows the charging temperature profile of all sensors placed at the surface of the slabs for the two PCM encapsulation design at different flow rates. To analyze the effect of the encapsulation design in the charging duration, both slabs types were compared at the same flow rate. Due to the higher heat transfer surface and the reduced amount of PCM (30% less according to Table 3) when using ThinICE slabs, at both flow rates, the experiment with the ThinICE slabs reached full charge (T\_out = 25 ◦C) 14% faster than with FlatICE. Furthermore, when analyzing the impact of the flow rate, in both slab designs the experiments show that at 4 L/min the full charge is reached 60% faster than at 2 L/min. up to 15 K between the coldest and hottest regions of the tank. This can be explained by the fact that the tank with FlatICE fits a lower number of slabs, as well as presenting smaller HTF channels compared with the tank with the ThinICE design (Figure 3, Table 2). Therefore, in this tank, the opposition to the HTF flow is higher, enhancing the distribution of the latter towards the regions of the tank where the density is more similar to the HTF inlet one (i.e., upper and middle region of the tank).

*Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 7 of 14

Figure 5 shows the charging temperature profile of all sensors placed at the surface of the slabs for the two PCM encapsulation design at different flow rates. To analyze the effect of the encapsulation design in the charging duration, both slabs types were compared at the same flow rate. Due to the higher heat transfer surface and the reduced amount of PCM (30% less according to Table 3) when using ThinICE slabs, at both flow rates, the experiment with the ThinICE slabs reached full charge (T\_out = 25 °C) 14% faster than with FlatICE. Furthermore, when analyzing the impact of the flow rate, in both slab designs the experiments show that at 4 L/min the full charge is reached 60% faster than at

When comparing temperature distribution inside the tank (Figure 5), a constant stratification profile between the top, middle, and bottom slabs was observed in all the experiments. This effect was more pronounced in the tests performed with FlatICE slabs in which the lower part of the tank takes longer to charge, obtaining a temperature gradient

*3.1. Temperature Evolution during the Charging Process* 

**3. Results and Discussion** 

2 L/min.

**Figure 5.** Charge PCM slab temperature profile for different slabs and HTF flow rates: (**a**) C\_ThinICE and 2 L/min, (**b**) C\_FlatICE and 2 L/min, (**c**) C\_ThinICE and 4 L/min, and (**d**) C\_FlatICE and 4 L/min. Note: The red line denotes the end of the charging experiment (T\_out = 25 °C). The time axis is not presented on the same scale in all the figures. *3.2. Heat Transfer Rate Evolution and Total Energy Stored in the Charging Process*  **Figure 5.** Charge PCM slab temperature profile for different slabs and HTF flow rates: (**a**) C\_ThinICE and 2 L/min, (**b**) C\_FlatICE and 2 L/min, (**c**) C\_ThinICE and 4 L/min, and (**d**) C\_FlatICE and 4 L/min. Note: The red line denotes the end of the charging experiment (T\_out = 25 ◦C). The time axis is not presented on the same scale in all the figures.

Figure 6 presents the evolution of the heat transfer rate (HTR) during the charging process of the four studied cases. Due to the characteristics of the experimental set-up, at the beginning of the experiment, the inlet temperature of the tank oscillated ±2 °C with When comparing temperature distribution inside the tank (Figure 5), a constant stratification profile between the top, middle, and bottom slabs was observed in all the experiments. This effect was more pronounced in the tests performed with FlatICE slabs in which the lower part of the tank takes longer to charge, obtaining a temperature gradient up to 15 K between the coldest and hottest regions of the tank. This can be explained by the fact that the tank with FlatICE fits a lower number of slabs, as well as presenting smaller HTF channels compared with the tank with the ThinICE design (Figure 3, Table 2). Therefore, in this tank, the opposition to the HTF flow is higher, enhancing the distribution of the latter towards the regions of the tank where the density is more similar to the HTF inlet one (i.e., upper and middle region of the tank).

#### *3.2. Heat Transfer Rate Evolution and Total Energy Stored in the Charging Process*

Figure 6 presents the evolution of the heat transfer rate (HTR) during the charging process of the four studied cases. Due to the characteristics of the experimental set-up, at the beginning of the experiment, the inlet temperature of the tank oscillated ±2 ◦C with respect to the desired temperature, affecting the initial peak of the heat transfer rate. However, the inlet temperature stabilized (with Tin standard deviation lower than 0.3) before the temperature inside the tank reaches the latent range of the PCM. The HTR profiles showed an exponential behavior with significantly higher values during the first 20 min of the process when the heat is mainly transferred to the HTF inside the tank and, therefore, rapidly increases its temperature. Afterwards, while the PCM temperature increases, the values of the heat transfer exponentially decrease until minimum values.

faster than the one delivered to the FlatICE\_4L.

lower thickness of the PCM layer using this type of slab.

**Figure 6.** Evolution of the HTF heat transfer rate during the charging processes of the four study cases presented in this study. **Figure 6.** Evolution of the HTF heat transfer rate during the charging processes of the four study cases presented in this study.

respect to the desired temperature, affecting the initial peak of the heat transfer rate. However, the inlet temperature stabilized (with Tin standard deviation lower than 0.3) before the temperature inside the tank reaches the latent range of the PCM. The HTR profiles showed an exponential behavior with significantly higher values during the first 20 min of the process when the heat is mainly transferred to the HTF inside the tank and, therefore, rapidly increases its temperature. Afterwards, while the PCM temperature increases,

During the first 1.5 h of operation, ThinICE\_2L and FlatICE\_2L showed similar HTR values, which indicates that, due to the low flow rate, the heat transfer by convection is low. Therefore, the higher heat transfer surface area existing with ThinICE slabs is not fully exploited. Moreover, after 1.5 h the HTR delivered to the ThinICE\_2L decreases faster than the one delivered to the FlatICE\_2L due to the higher amount of PCM introduced into the tank with FlatICE slabs. At a higher flow rate, heat transfer by convection increases. Therefore, during the first 1.5 h of operation, the bigger heat transfer surface area present with ThinICE\_4L slabs increases its HTR over FlatICE\_4L. After this period, and similar to the results at 2 L/min, the power delivered to the ThinICE\_4L decreased

When analyzing the effect of the flows in each slab type, the influence is greater in the tank with ThinICE slabs obtaining, after the initial peak, up to 0.4 kW more in Thin-ICE\_4L than in ThinICE\_2L. In the case of the tank with FlatICE, this increase drops to 0.1 kW when comparing FlatICE\_4L vs FlatICE\_2L. The latter results corroborate the statement above; the increase in heat transfer by convection, as the flow rate increases, is more pronounced in the tank with ThinICE slabs due to the larger heat transfer surface and the

the values of the heat transfer exponentially decrease until minimum values.

Figure 7 reports the total energy stored for each experiment condition. The results with ThinICE slabs show that the flow variation did not affect the total stored energy, suggesting the correct utilization of the energy storage capacity of the PCM. Conversely, when analyzing the tank with FlatICE slabs, the charging experiments at 4 L/min stored 10% less energy than the same experiment at 2 L/min. This is due to changes in the flow rate distribution between the slab channels inside the tank when increasing the flow rate. At the end of the experiment, (T\_out 25 °C) with FlatICE at 4 L/min, the PCM in the bottom During the first 1.5 h of operation, ThinICE\_2L and FlatICE\_2L showed similar HTR values, which indicates that, due to the low flow rate, the heat transfer by convection is low. Therefore, the higher heat transfer surface area existing with ThinICE slabs is not fully exploited. Moreover, after 1.5 h the HTR delivered to the ThinICE\_2L decreases faster than the one delivered to the FlatICE\_2L due to the higher amount of PCM introduced into the tank with FlatICE slabs. At a higher flow rate, heat transfer by convection increases. Therefore, during the first 1.5 h of operation, the bigger heat transfer surface area present with ThinICE\_4L slabs increases its HTR over FlatICE\_4L. After this period, and similar to the results at 2 L/min, the power delivered to the ThinICE\_4L decreased faster than the one delivered to the FlatICE\_4L.

When analyzing the effect of the flows in each slab type, the influence is greater in the tank with ThinICE slabs obtaining, after the initial peak, up to 0.4 kW more in ThinICE\_4L than in ThinICE\_2L. In the case of the tank with FlatICE, this increase drops to 0.1 kW when comparing FlatICE\_4L vs FlatICE\_2L. The latter results corroborate the statement above; the increase in heat transfer by convection, as the flow rate increases, is more pronounced in the tank with ThinICE slabs due to the larger heat transfer surface and the lower thickness of the PCM layer using this type of slab.

Figure 7 reports the total energy stored for each experiment condition. The results with ThinICE slabs show that the flow variation did not affect the total stored energy, suggesting the correct utilization of the energy storage capacity of the PCM. Conversely, when analyzing the tank with FlatICE slabs, the charging experiments at 4 L/min stored 10% less energy than the same experiment at 2 L/min. This is due to changes in the flow rate distribution between the slab channels inside the tank when increasing the flow rate. At the end of the experiment, (T\_out 25 ◦C) with FlatICE at 4 L/min, the PCM in the bottom slabs of the tank had not completed the phase change (Figure 5) and therefore stored 8% less energy.

less energy.

**Figure 7.** Total energy delivered to the PCM storage tank during the charging processes of the four study cases presented in this study. **Figure 7.** Total energy delivered to the PCM storage tank during the charging processes of the four study cases presented in this study.

slabs of the tank had not completed the phase change (Figure 5) and therefore stored 8%

#### *3.3. Temperature Evolution during the Discharging Process 3.3. Temperature Evolution during the Discharging Process*

Figure 8 shows the discharging temperature profile of all sensors placed at the surface of the slabs at two different mass flow rates. Analyzing the influence of the PCM encapsulation design on both flow rates, the tank with ThinICE slabs finished the discharging process 30% faster than with FlatICE slabs. This can be explained by the higher heat transfer surface and the lower amount of PCM (30% less according to Table 3) when using ThinICE. Moreover, when analyzing the influence of the flow rate, Figure 8 shows that for both slab types at 4 L/min the experiments were completed 50% faster than at 2 Figure 8 shows the discharging temperature profile of all sensors placed at the surface of the slabs at two different mass flow rates. Analyzing the influence of the PCM encapsulation design on both flow rates, the tank with ThinICE slabs finished the discharging process 30% faster than with FlatICE slabs. This can be explained by the higher heat transfer surface and the lower amount of PCM (30% less according to Table 3) when using ThinICE. Moreover, when analyzing the influence of the flow rate, Figure 8 shows that for both slab types at 4 L/min the experiments were completed 50% faster than at 2 L/min. *Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 10 of 14

**Figure 8.** Discharge process PCM slab temperature profile for different slabs and HTF flow rates: (**a**) ThinICE and 2 L/min, (**b**) FlatICE and 2 L/min, (**c**) ThinICE and 4 L/min, and (**d**) FlatICE and 4 L/min. Note: The red line denotes the end of the charging experiment (T\_out = 25 °C). The time axis is not presented on the same scale in all the figures. **Figure 8.** Discharge process PCM slab temperature profile for different slabs and HTF flow rates: (**a**) ThinICE and 2 L/min, (**b**) FlatICE and 2 L/min, (**c**) ThinICE and 4 L/min, and (**d**) FlatICE and 4 L/min. Note: The red line denotes the end of the charging experiment (T\_out = 25 ◦C). The time axis is not presented on the same scale in all the figures.

*3.4. Heat Transfer Rate Evolution and Total Energy Released in the Discharging Process* 

mentation the inlet temperature of the tank oscillates ±2 °C around the desired temperature, affecting the initial peak of power. However, the inlet temperature stabilizes (T\_in standard deviation lower than 0.3 °C) before the temperature inside the tank reaches the

FlatICE\_2L. Moreover, due to the lower amount of PCM in the storage tank with ThinICE slabs, after 1.5 h the HTR delivered by ThinICE\_2L decreases faster than the one delivered by FlatICE\_2L. At 4 L/min, after the initial peak and during the first 1.5 h, similar results to 2 L/min are obtained. Moreover, after 1.5 h the HTR of ThinICE\_4L drastically decreases, therefore for the next 2 h (from 1.5–3.5 h) FlatICE\_4L maintains an HTR up to 0.4

The HTR evolution during the discharging process for all the experimental cases is shown in Figure 9. In all the experiments the profiles showed a similar trend. Significantly higher values were obtained during the first 20 min of the process when the heat is mainly transferred from the HTF inside the tank followed by an exponential decrease while the PCM decreases its temperature until minimum values are reached. Furthermore, in this case, due to the characteristics of the experimental facility, at the beginning of the experi-When comparing temperature distribution inside the tank (Figure 8), a similar temperature profile between the top, middle, and bottom slabs was observed in all the experiments performed with ThinICE slabs. Moreover, this behavior changed with the use of the Flat-ICE, where the stratification and the profile temperature inside the tank depends on the mass flow rate.

kW higher than ThinICE\_2L.

#### *3.4. Heat Transfer Rate Evolution and Total Energy Released in the Discharging Process*

The HTR evolution during the discharging process for all the experimental cases is shown in Figure 9. In all the experiments the profiles showed a similar trend. Significantly higher values were obtained during the first 20 min of the process when the heat is mainly transferred from the HTF inside the tank followed by an exponential decrease while the PCM decreases its temperature until minimum values are reached. Furthermore, in this case, due to the characteristics of the experimental facility, at the beginning of the experimentation the inlet temperature of the tank oscillates ±2 ◦C around the desired temperature, affecting the initial peak of power. However, the inlet temperature stabilizes (T\_in standard deviation lower than 0.3 ◦C) before the temperature inside the tank reaches the latent range of the PCM. After the initial peak and during the first 1.5 h of operation, ThinICE\_2L shows slightly better performance getting up to 0.1 kW more HTR than FlatICE\_2L. Moreover, due to the lower amount of PCM in the storage tank with ThinICE slabs, after 1.5 h the HTR delivered by ThinICE\_2L decreases faster than the one delivered by FlatICE\_2L. At 4 L/min, after the initial peak and during the first 1.5 h, similar results to 2 L/min are obtained. Moreover, after 1.5 h the HTR of ThinICE\_4L drastically decreases, therefore for the next 2 h (from 1.5–3.5 h) FlatICE\_4L maintains an HTR up to 0.4 kW higher than ThinICE\_2L. *Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 11 of 14

**Figure 9.** Evolution of the HTF heat transfer rate during the discharging processes of the four study cases presented in this study. **Figure 9.** Evolution of the HTF heat transfer rate during the discharging processes of the four study cases presented in this study.

Figure 10 reports the total energy released for each experiment conditions, and the percentage it represents with respect to the energy stored in the charging process (Figure 7) in each case. Analyzing the effects of flow rate, similar to the charging, both experiments with ThinICE slabs showed a comparable energy release, suggesting a correct utilization of the energy stored in the PCM. In the case of FlatICE slabs, experiments at 2 L/min released 10% more energy compared with 4 L/min. This is supported by the fact that in the charging process the tank at 2 L/min manages to store 10% more energy than at 4 L/min (Figure 5). In addition, it is interesting to note that in all cases of the selected operating Figure 10 reports the total energy released for each experiment conditions, and the percentage it represents with respect to the energy stored in the charging process (Figure 7) in each case. Analyzing the effects of flow rate, similar to the charging, both experiments with ThinICE slabs showed a comparable energy release, suggesting a correct utilization of the energy stored in the PCM. In the case of FlatICE slabs, experiments at 2 L/min released 10% more energy compared with 4 L/min. This is supported by the fact that in the charging process the tank at 2 L/min manages to store 10% more energy than at 4 L/min (Figure 5). In addition, it is interesting to note that in all cases of the selected operating threshold approximately 85% of the energy stored in the tank was discharged.

**Figure 10.** Total energy released by the PCM storage tank during the discharging processes of the

Macro-encapsulation of phase change materials (PCM) represents one of the most widely used techniques for the implementation of latent heat thermal energy storage systems. The design of the macro-encapsulation is fundamental to archive the best compromise between optimal heat transfer performance and energy stored. However, current literature lacks experimental data on the effect of macro-encapsulation in the performance

threshold approximately 85% of the energy stored in the tank was discharged.

9106

85 %

8218

83 %

7837 7555

86 % 82 %

D\_ThinICE\_2L D\_ThinICE\_4L D\_FlaICE\_2L D\_FlaICE\_4L

four study cases presented in this study.

of latent heat thermal energy storage.

**4. Conclusions** 

Energy [kW]

**Figure 10.** Total energy released by the PCM storage tank during the discharging processes of the four study cases presented in this study. **Figure 10.** Total energy released by the PCM storage tank during the discharging processes of the four study cases presented in this study.

**Figure 9.** Evolution of the HTF heat transfer rate during the discharging processes of the four

012345678

Time [hours]

threshold approximately 85% of the energy stored in the tank was discharged.

Figure 10 reports the total energy released for each experiment conditions, and the percentage it represents with respect to the energy stored in the charging process (Figure 7) in each case. Analyzing the effects of flow rate, similar to the charging, both experiments with ThinICE slabs showed a comparable energy release, suggesting a correct utilization of the energy stored in the PCM. In the case of FlatICE slabs, experiments at 2 L/min released 10% more energy compared with 4 L/min. This is supported by the fact that in the charging process the tank at 2 L/min manages to store 10% more energy than at 4 L/min (Figure 5). In addition, it is interesting to note that in all cases of the selected operating

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Time [hours]

ThinICE\_4L ThinICE\_2L FlatICE\_4L FlatICE\_2L

#### **4. Conclusions 4. Conclusions**

study cases presented in this study.

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

Power [kW]

0

1

2

Power [kW]

Heat transfer rate [kW]

3

4

5

Macro-encapsulation of phase change materials (PCM) represents one of the most widely used techniques for the implementation of latent heat thermal energy storage systems. The design of the macro-encapsulation is fundamental to archive the best compromise between optimal heat transfer performance and energy stored. However, current literature lacks experimental data on the effect of macro-encapsulation in the performance Macro-encapsulation of phase change materials (PCM) represents one of the most widely used techniques for the implementation of latent heat thermal energy storage systems. The design of the macro-encapsulation is fundamental to archive the best compromise between optimal heat transfer performance and energy stored. However, current literature lacks experimental data on the effect of macro-encapsulation in the performance of latent heat thermal energy storage.

of latent heat thermal energy storage. This paper analyzed, through an experimental study, the effect of the design of macroencapsulated PCM on the thermal behavior of a latent heat thermal energy storage tank during both the charging and discharging processes. In this study, external dimensions of the energy storage tank were fixed and two different types of commercial slabs with different thickness filled with the same PCM were tested. The results could be particularly useful to evaluate the best configuration of storage medium when the storage tank is limited with a fixed volume.

The results were compared in terms of temperature profile, heat transfer rate, and energy stored/released. The results and the conclusions obtained from this study can be applied to similar configuration of the PCM storage that aim to use rectangular macroencapsulated slabs as storage medium. The lesson learnt from this study suggests that macro-encapsulation design has a relevant impact on the heat transfer during both charging and discharging processes, so the design of the TES unit should be done and analyzed according to the requirements of the application.

The use of a thinner macro-encapsulation design (ThinICE) allowed fitting a larger number of slabs inside the tank. However, the higher amount of encapsulation material and the larger distance between the slabs (i.e., higher HTF channels height) resulted in a 30% less amount of PCM introduced inside the tank with this encapsulation design.

With ThinICE slabs, the temperature profiles were less affected by the influence of the mass flow rate, promoting a stratified temperature profile inside the tank in both the charging and discharging processes. Using FlatICE, this effect is more pronounced at low flow rates due to the smaller height of the channels that obstructed the flow at the bottom of the tank during charging and at the top of the tank during discharging. However, at high flow rates, the stratification is reduced with the use of thicker slabs, especially during the discharging process.

In all the discharging tests, when the outlet temperature of the tank reached 7 ◦C, approximately 85% of the energy previously stored in the tank was discharged.

The effect of increasing the heat transfer surface using ThinICE slabs on the power delivered by the storage tank is mostly appreciated at a higher flow rate where the heat transferred by convection is higher. Furthermore, using thinner slabs, the higher heat transfer surface area achieves a higher discharging power but is delivered for a shorter period of time. Therefore, for longer discharging periods and for higher storage capacity given a fixed volume of storage tank, the use of FlatICE should be preferred.

**Author Contributions:** Conceptualization, D.V., E.B., and L.F.C.; methodology, D.V., E.B., and G.Z.; formal analysis, D.V., E.B., A.C., B.D.M., Á.d.G., G.Z., and L.F.C.; investigation, D.V., A.C., and E.B.; resources, L.F.C.; data curation, L.F.C.; writing—original draft preparation, D.V. and E.B.; writing review and editing, D.V., E.B., A.C., B.D.M., Á.d.G., and G.Z., L.F.C.; visualization, D.V.; supervision, L.F.C.; project administration, L.F.C.; and funding acquisition, L.F.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project has received funding from the European Union's Horizon 2020 research and innovation program under grant agreement No. 764025 (SWS-HEATING). This work was partially funded by the Ministerio de Ciencia, Innovación y Universidades de España (RTI2018-093849-B-C31— MCIU/AEI/FEDER, UE) and by the Ministerio de Ciencia, Innovación y Universidades—Agencia Estatal de Investigación (AEI) (RED2018-102431-T). This work is partially supported by ICREA under the ICREA Academia program.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** The authors would like to thank the Catalan Government for the quality accreditation given to their research group (2017 SGR 1537). GREiA is a certified agent TECNIO in the category of technology developers from the Government of Catalonia. Boniface Dominick Mselle would like to thank Programa Santander PredocUdL for his research fellowship. Alicia Crespo would like to acknowledge the financial support of the FI-SDUR grant from the AGAUR of the Generalitat de Catalunya and Secretaria d'Universitats i Recerca del Departament d'Empresa i Coneixement de la Generalitat de Catalunya.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Hitoshi Kiyokawa <sup>1</sup> , Hiroki Tokutomi <sup>1</sup> , Shinichi Ishida <sup>2</sup> , Hiroaki Nishi <sup>3</sup> and Ryo Ohmura 1,\***

	- Japan; west@sd.keio.ac.jp

**Abstract:** Kinetic characteristics of thermal energy storage (TES) using tetrabutylammonium acrylate (TBAAc) hydrate were experimentally evaluated for practical use as PCMs. Mechanical agitation or ultrasonic vibration was added to detach the hydrate adhesion on the heat exchanger, which could be a thermal resistance. The effect of the external forces also was evaluated by changing their rotation rate and frequency. When the agitation rate was 600 rpm, the system achieved TES density of 140 MJ*/*m<sup>3</sup> in 2.9 h. This value is comparable to the ideal performance of ice TES when its solid phase fraction is 45%. UA/V (U: thermal transfer coefficient, A: surface area of the heat exchange coil, V: volume of the TES medium) is known as an index of the ease of heat transfer in a heat exchanger. UA/V obtained in this study was comparable to that of other common heat exchangers, which means the equivalent performance would be available by setting the similar UA/V. In this study, we succeeded in obtaining practical data for heat storage by TBAAc hydrate. The data obtained in this study will be a great help for the practical application of hydrate heat storage in the future.

**Keywords:** lathrate hydrate; thermal energy storage; tetrabutylammonium acrylate (TBAAc); crystal growth; ultrasonic vibration

#### **1. Introduction**

The use of sustainable, renewable energy sources is becoming increasingly important as global attention on environmental issues increases. However, the amount of electricity generated by renewable energy sources such as wind and solar power fluctuates greatly depending on the natural environment. In recent years, energy consumption has been increasing. It is essential to have a technology to fill the gap between the amount of electricity generated and the demand. These technologies would contribute to the load leveling [1].

In particular, the recent development in information and communication technology (ICT) (e.g., Internet of Things, 5G communications, cloud computing, big data) has produced a great demand for data centers (DCs). In 2020, the rapid spread of remote working due to the pandemic also boosted demand for DCs. The industrial use of these technologies is also driving this trend [2,3]. Masanet et al. [4] estimated that the electricity use in DCs accounted for 1% of worldwide electricity use in 2020. This trend will make DCs even more energy-intensive and will increase the cost of cooling the heat generated and the amount of electricity used [5]. Considering the generalization of ICT and the widespread use of the fifth-generation mobile communication system, the amount of communication will further increase worldwide. The importance of data centers will increase to support the exchange of huge amounts of data. A suitable cooling system will be able to keep the appropriate temperature. In addition, such systems are significant to realize the zero-downtime of a DC operation with low cost.

**Citation:** Kiyokawa, H.; Tokutomi, H.; Ishida, S.; Nishi, H.; Ohmura, R. Thermal Energy Storage Performance of Tetrabutylammonium Acrylate Hydrate as Phase Change Materials. *Appl. Sci.* **2021**, *11*, 4848. https:// doi.org/10.3390/app11114848

Academic Editor: Ioannis Kartsonakis

Received: 17 April 2021 Accepted: 24 May 2021 Published: 25 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

DC cooling systems are classified into two types. One is an air-based cooling system [6,7]. The other one is a liquid cooling system [8]. As for air-based system, the heat emitted from the computer is removed by air flow. The main drawback of this system is a low thermal conductivity, less than 0.03 W m−<sup>1</sup> K−<sup>1</sup> [9]. On the other hand, in a liquid cooling system, the heat is removed by the thermal conduction and convection into the liquid, mostly water. A liquid cooling system is inclined to be bigger and heavier because that system utilizes only sensible heat, which has low energy density. Thermal energy storage technology is used to store the heat energy with the heat capacity of substance. When the demand of the electricity is low, the surplus electricity is stored as cold energy. Then, the stored energy is changed to electricity when the demand increases, for example, daytime peak in power usage [10,11].

Phase Change Materials (PCMs) can be a competitive system for thermal energy storage because they can also utilize latent heat, which has a higher energy density [12–16]. Organic compounds and water are representative examples of PCMs for DC cooling. As for organic PCMs, they often have problems in terms of safety. For example, paraffin is known as an organic PCM, but it is not safe because of its flammability [17]. Water is not suitable because its phase change point is outside the range of operation temperature in DCs, 288 K to 305 K [18].

Utilizing clathrate hydrate for a thermal energy storage medium has been proposed as a better idea than water or an ice-based medium [19–23]. Clathrate hydrates, ice-like compounds, are formed when some gases have contact with water or ice under high pressure and/or low temperature [24]. The gases encapsulated in clathrate hydrates are called guest compounds. Clathrate hydrates have enormous potential for industrial usage thanks to their thermodynamic properties [19,22,25–28]. Considering the application for thermal energy storage medium, equilibrium temperature should be suitable for the DC operation temperature. Favorable equilibrium temperature can be obtained appropriately by selecting the guest compounds [29–31].

Until now, only tetrabutylammonium bromide (TBAB) hydrate has been commercially applied as a hydrate PCM for air conditioning. The thermodynamic properties of TBAB hydrate have been reported in the previous studies. TBAB hydrate has an equilibrium temperature at an atmospheric pressure of 285.9 K [32], a dissociation heat of 193 kJ/kg [33], and the thermal conductivity of 0.35 W m−<sup>1</sup> K−<sup>1</sup> [34]. As reported by ASHRAE Technical Committee [18], the appropriate temperature range of DC cooling is 288 K to 305 K. This is why TBAB hydrate is not favorable for DC cooling.

Sakamoto et al. [35] studied the thermodynamic properties of tetrabutylammonium acrylate (TBAAc) hydrate. According to the study, TBAAc hydrate had the highest equilibrium temperature of 291.5 K at *w*TBAAc = 0.36, where *w*TBAAc is the mass fraction of TBAAc. The greatest dissociation heat, a latent heat, of 195 kJ/kg was obtained at *w*TBAAc = 0.33 under ambient pressure. As for the thermal conductivity of ionic semiclathrate hydrates, Fujiura et al. [34] studied the value of tetrabutylammonium bromide (TBAB) hydrate and tetrabutylammonium chloride (TBAC) hydrate, which have similar characteristics with TBAAc hydrate. They reported that TBAB hydrate and TBAC hydrate have thermal conductivity of approximately 0.40 W m−<sup>1</sup> K −1 . According to the report by ASHRAE Technical Committee [18], the cooling system for DCs is required to be set from 288 K to 305 K. From the aspect of operation temperature in DCs, it can be said that TBAAc hydrates are suitable for utilizing in DC cooling as PCMs. Also, the dissociation heat of the TBAAc is larger than that of TBAB hydrate, which has been commercialized in the industrial area [33,35]. Given these advantages, TBAAc hydrates could be better PCMs for DC cooling. This is why it has significant meaning to practically evaluate the performance of TBAAc hydrates as PCMs. As a step for industrial use of TBAAc hydrate, we performed kinetic thermal energy storage experiments with TBAAc hydrate.

Based on these results, we obtained the data of energy storage density and energy storage rate during the experiments. During the experiments, we also found the adhesion of the hydrate to the surface of the heat exchanger. As the adhered hydrate is the thermal resistance to the further hydrate formation, we used an agitator or an ultrasonic transducer to break the adhesion. By comparing the data with and without external forces, such as the agitation and ultrasonic waves, we obtained the practical data for a hydrate-based thermal energy storage system in DC cooling.

#### **2. Materials and Methods**

#### *2.1. Materials*

The details on reagents used in this study were summarized in Table 1. Tetrabutylammonium Acrylate (TBAAc) aqueous solution was obtained by neutralizing Tetrabuthylammonium hydroxide (TBAOH) aqueous solution, 1.52 kg, (0.40 mass fraction, Sigma Aldrich Xo. LLC, Saint Louis, MO, USA) and Acrylic acid, 0.17 kg, (0.99 mass fraction in liquid reagent, Sigma Aldrich Xo. LLC, Saint Louis, State of MO, USA). Also, we added laboratory-made H2O, 0.35 kg. The mass fraction of TBAAc (*wTBAA<sup>c</sup>* )) was 0.36. The masses of all reagents were measured by an electronic balance (GF-600, A&D Co. Ltd., Tokyo, Japan) with an expanded uncertainty of ± 0.004 g (coverage factor, *k* = 2).


#### *2.2. Apparatus*

The schematic diagram of the experimental apparatus is illustrated in Figure 1. The insulating container's width, length and height are all 150 mm. Cooling water was circulated inside a copper coil tube (inner diameter 6.35 mm, outer diameter 4.35 mm) for heat exchange. This coil's diameter, pitch and number of turns are 100 mm, 10 mm, and 7 times, respectively. The temperature and flow rate in the coil was controlled by a chiller (CTP-3000 EYELA, Tokyo Rikakikai Co. Ltd., Tokyo, Japan) and flowmeter with a precision needle valve (RK 1250, KOFLOC Kyoto, Kyoto, Japan). As shown in Figure 1, the temperature of cooling water and TBAAc were measured by three platinum resistance temperature detectors (222-055, Electronic Temperature Instrument Ltd., West Sussex, UK) with an uncertainty of ± 0.1 K (*k* = 2). The surface of the coil would be covered with hydrate during the experiment. This hydrate prevents the heat exchange between the heat exchanger and thermal energy storage medium. We used an ultrasonic transducer (HEC-45282, HONDA Electronics Co. Ltd., Tokyo, Japan) or an agitator (SM-103D, Kenis Ltd., Osaka, Japan) to prevent the hydrate deposition.

#### *2.3. Procedures*

We made 2.0 kg of TBAAc aqueous solution by mixing the above reagents and put them in the container. The entire system was confirmed to be stationary at 298 K. *T*in (outlet of the chiller) and flow rate, . *Vw*, of the cooling water were controlled at 291 K and 5.0 <sup>×</sup> <sup>10</sup>−<sup>4</sup> <sup>m</sup>3/min. Then, we started the thermal energy storage with TBAAc hydrate. We set t = 0 when we started the circulation of the cooling water. The time evolution of *T*in, *T*out (temperature at inlet of the chiller) and *T*PCM were measured every 10 s to evaluate the kinetic aspect of TBAAc hydrate as a thermal energy storage medium.

We performed such experiments in three systems. First, we did not add any external forces to the apparatus. Second, we used a rotating-impeller agitator to break adhesions between the hydrate and the coil. Third, we used the ultrasonic wave instead of mechanical agitation. In the system where the agitator was used, we also investigated the effect of rotation rate of the agitator by changing the rate to 100 rpm, 300 rpm and 600 rpm. Similarly, we investigated the effect of the frequency of ultrasonic waves on the system by changing the frequency. The frequency was set at either 28 kHz or 56 kHz.

**Figure 1.** Schematic diagram of the apparatus.

#### **3. Results and Discussion**

#### *3.1. The System without Any External Forces (Static System)*

Initially, we reported the thermographs of Tetrabutylammonium Acrylate (TBAAc) hydrate obtained by a differential scanning calorimeter (DSC). Figure 2 shows the thermographs, which were measured at *wTBAAc* = 0.33 and 0.36. A single peak was observed, and the existence of ice was not identified at *wTBAAc* = 0.33.

As for the system without any external forces (static system), we performed thermal energy storage experiments. We measured Tin, Tout, and TPCM. Figure 3 shows the time evolution of each temperature. Time zero indicates the time at which the circulating water is started to flow inside the coil. As shown in Figure 3, TPCM decreased until t = 1.5 h and increased from t = 1.5 to t = 4 h. When TPCM is increasing, TBAAc hydrate formation would occur because the TPCM was 10 K lower than its equilibrium temperature, 291.5 K [35]. After the hydrate formation, the heat generation from the hydrate formation occurred and exceeded the endothermic of the heat exchanger. This was why TPCM rose from t = 1.5 to t = 4 h. TPCM moderately decreased again at t = 4 h when TPCM was 288 K. At this moment, the formed TBAAc hydrate gradually grew around the heat exchanger as shown in Figure 4d. This adhesion would be the thermal resistance between the heat exchanger and thermal energy storage medium much lower than before. Then, the hydrate formation rate decreased due to the increased thermal resistance, and TPCM decreased because the endothermic of the heat exchanger exceeds the heat generation from hydrate formation.

Figure 4a–d illustrates the optical observation of the time evolution of the hydrate formation. The hydrate was formed on the surface of the heat exchanger. Heat exchange occurs on the heat exchanger surface, which leads to hydrate formation occurring there. From t = 1.5 h to t = 2 h, the hydrate gradually grew and started covering the coil surface. The surface of the coil was fully covered by the hydrate around t = 4 h. After t = 7 h, optical changes could not be observed. In addition, we found that the TBAAc hydrate

kept growing from the surface of the coil to the bulk of the solution. There was a 3 cm gap between the surface of the coil and the wall of the container. This discovery of the hydrate growth length can be a tip to design a static thermal energy storage system. Thermal energy storage rate, . *q*, and density, *q*, were calculated by the formulas (1) and (2). What every character stands for is written in Table 2. On the right-hand side of (2), the thermal energy storage density due to sensible heat is subtracted from the thermal energy storage density, which is obtained by integrating the thermal energy storage rate.

$$
\dot{q} = \rho\_w \dot{V}\_w c\_w (T\_{out} - T\_{in}) / V\_{\rm PCM} \tag{1}
$$

$$\overline{q} = \sum\_{t=0}^{t} \dot{q}dt - \rho\_{\text{PCM}} c\_{\text{PCM}} (T\_{\text{PCM},0} - T\_{\text{PCM}}) \tag{2}$$

**Figure 2.** The DSC heating curves measured at *wTBAAc* = 0.33 and 0.36. •: *wTBAAc* = 0.36, ◆: *wTBAAc* = 0.33.

**Figure 3.** Time evolutions of the temperature at inlet of the chiller (Tin), outlet (Tout) of the chiller and thermal energy storage medium (TPCM) in the static system.

**Figure 4.** Crystal growth of TBAAc hydrate on the surface of the heat exchanger coil. (**a**–**d**) illustrate the optical measurement of the hydrate formation at t = 2, t = 2.5, t = 4 and t = 7, respectively.


**Table 2.** The value of each parameter in the calculation.

The obtained thermal energy storage density represents the heat storage density due to heat of hydrate formation and decomposition. The density of the PCM is assumed to be 1.0 kg m−<sup>3</sup> . The value of each parameter used in the calculation is shown in Table 2. The physical property of water was referred to REFPROP ver. 9.1 [36].

Figure 5a illustrates the time evolution of the thermal energy storage rate in the static system. Since . *q* is proportional to the gap between Tin and Tout, . *q* varied the same way with Tout. Immediately after the start of the experiment, . *q* decreased as Tout decreased, and as Tout approached Tin, . *<sup>q</sup>* approached 0. After that, . *q* increased at t = 1.5 h and reached around 9 kW/m<sup>3</sup> . This is due to the increase in Tout caused by the heat generation from hydrate formation and growth, as described above. After t = 3 h, as Tout descended to be tangent to Tin, . *q* descended to 0 with some fluctuations. Figure 5b illustrates the time evolution of thermal energy storage density. Since *q* indicates the thermal energy density by hydrate growth and decomposition, there is no change right after the experiments start. As the hydrate grew, thermal energy storage density increased. The rate of increase gradually became slower. The hydrate formation rate got slower due to the decreased heat conductivity around the heat exchanger coil. This is why the increasing rate of thermal energy storage density decreased. Thermal energy density reached 96 MJ/m<sup>3</sup> in 10 h.

**Figure 5.** Time evolution of thermal energy storage rate and density in static system. (**a**,**b**) depict the thermal energy storage rate and density, respectively.

It is known that the solid phase fraction of ice thermal storage increases significantly with time up to around 30%, but the increasing rate decreases after about 40–50%. Therefore, it is favorable to design a thermal energy storage system that has a solid phase fraction of 40–50% [37]. For example, when the solid phase fraction is 45% in an ice thermal energy storage system, thermal energy storage density is approximately 140 MJ/m<sup>3</sup> . Oyama et al. [33] reported that TBAB hydrate, which is the only commercially available hydrate PCM, has the heat for formation and decomposition of approximately 200 MJ/m<sup>3</sup> . Assuming the TBAAc hydrate has 200 MJ/m<sup>3</sup> of heat in hydrate formation and decomposition, the thermal energy storage density corresponds to a solid phase fraction of 48%.

#### *3.2. In the System with Mechanical Agitation*

As written in 3.1., the surface of the heat exchanger coil was covered by hydrate. This hydrate adhesion increased thermal resistance. To solve this problem, mechanical agitation was used. In addition, to investigate the effect of the rotation rate, we changed the rate to 100 rpm, 300 rpm and 600 rpm. Figure 6a–c shows the time evolution of Tin, Tout and TPCM in each rotation rate. Time zero was defined as the time at which the circulating water is started to be flown. In every rotation rate, TPCM decreased twice after the experiment started and temporarily increased. In the agitated system, TPCM increased earlier than the static system. As the rotation rate became quicker, TPCM increased earlier. When TPCM increased, the temperature was around 282 K, which is the same as the static system. This result indicates that the forced convection generated by the agitation increased the convective heat transfer coefficient, which resulted in faster cooling of the heat storage medium and thus faster nucleation. TPCM increased rapidly more than the static system, which was almost vertically to around 288 K. This would be due to the increase in heat and mass transfer rates caused by the forced convection generated by the agitation, which led to the increase in the hydrate growth rate.

TPCM changed from a rise to a fall around 288 K and then descended slowly. As the hydrate grew, the viscosity of the heat storage medium increased and the forced convection was weakened, which lowered the hydrate formation rate. TPCM decreased because the hydrate growth rate became smaller, and the heat absorption by the heat exchanger exceeded the heat generated from hydrate formation.

Figure 7a–c shows the pictures of crystal growth in the system when the mechanical agitation rate was 300 rpm. At t = 0.6 h, numerous small crystals began to float in the aqueous solution. After t = 0.7 h, hydrate was formed in the entire aqueous solution. The hydrate nucleation occurred in the entire aqueous solution. This was because of the forced convection generated by the agitation, which stripped the hydrate from the heat exchange coil and suspended it in the aqueous solution. This convection increased the hydrate growth rate as well. Within a few minutes, the heat exchange coil was only faintly visible

by the hydrate, as shown in Figure 7b. The hydrate formation has progressed and the viscosity increased as shown in Figure 7c, after which little change was observed.

**Figure 6.** Time evolutions of the temperature at inlet of the chiller (Tin), outlet (Tout) of the chiller and thermal energy storage medium (TPCM) in the system with the mechanical agitation. (**a**–**c**) depict each curve at the agitation rate of 100 rpm, 300 rpm and 600 rpm, respectively.

**Figure 7.** Optical observation changing in the container with mechanical agitation of 300 rpm. (**a**–**c**) are the pictures at 0.6 h, 0.7 h and 1 h after cooling water started to be circulated, re-spectively.

We visually observed that TBAAc hydrate formed and grew to the bulk of the solution. The gap between the heat transfer surface and the corner of the container was about 5 cm. In other words, when the cooling temperature was 10 K lower than the equilibrium temperature, TBAAc hydrate formed and grew at a distance of more than 5 cm from the heat transfer surface in the agitated system. The same tendency was observed at the rotation rate of 100 rpm and 600 rpm.

Figure 8a shows the time evolution of thermal energy storage density. To make a comparison with the static system, the time evolution of the static system is also shown in Figure 8a. Regarding the thermal energy storage rate, . *q* is proportional to the difference between the inflow temperature Tin and the outflow temperature Tout. This was why . *q* had almost the same trend as the Tout change in the static system. As Tout increased rapidly more than the static system, . *q* increased rapidly more than the static system as well, rising almost vertically. The maximum thermal energy storage rate at that time was about 6 to 8 times faster than that of the static system. The larger the rotation rate of the agitation increased the value of . *<sup>q</sup>*. After a rapid increase, the change in . *q* suddenly turned into a downward movement and gradually descended to 0. This was because the hydrate had grown, and the viscosity had increased, which resulted in less convection.

**Figure 8.** Time evolution of thermal energy storage rate and density in the agitated system (**a**,**b**) depict the thermal energy storage rate and density, respectively.

As shown in Figure 8b, *q* in the agitated system increased earlier than that in the static system, and the higher rotation rate of the agitation increased *q* earlier. This was because the temperature of the thermal energy storage medium was lowered quickly by forced convection, and thus hydrate nucleation also occurred quickly. The thermal energy storage density *q* was larger than that of the static system and increased by a higher rotation rate. This was because the forced convection caused by the agitation allowed hydrate to be formed in the entire bulk of the solution. In the agitated system with the rotation rates of 100 rpm, 300 rpm and 600 rpm, thermal energy density reached 121 MJ/m<sup>3</sup> , 159 MJ/m<sup>3</sup> and 183 MJ/m<sup>3</sup> in 10 h, respectively. Similarly, assuming the TBAAc hydrate has 200 MJ/m<sup>3</sup> of heat in hydrate formation and decomposition, the thermal energy storage density corresponds to a solid phase fraction of 61%, 80% and 92%, respectively.

#### *3.3. In the System Using an Ultrasonic Transducer*

In this section, we discuss the results in the system with ultrasonic vibration. The inflow temperature Tin, outflow temperature Tout and thermal energy storage medium temperature TPCM were measured. Figure 9a,b shows the time evolutions of Tin, Tout and TPCM in the ultrasonic vibration system (28 kHz and 56 kHz). Time zero was defined as the time at which the circulating water was started to flow.

**Figure 9.** Time evolutions of the temperature at inlet of the chiller (Tin), outlet (Tout) of the chiller and thermal energy storage medium (TPCM) in the system with ultrasonic vibration. (**a**,**b**) depict each curve at the frequency of 28 kHz and 56 kHz, respectively.

The slope of the TPCM first decrease was similar to that of the static system. However, when the slope changed from negative to positive, the TPCM was about 284 K, which was higher than that of the static and agitated system. This is because the hydrate nucleation occurred at a higher temperature than in the static system. In an ultrasonic vibration system, hydrate nucleation would occur at a higher temperature due to the effects of cavitation or acceleration of ultrasonic vibration.

The slope of the TPCM increase is not much different from that of the static system at 28 kHz and appears to be more gradual at 56 kHz. This was because of relatively high TPCM, which results in a relatively slow hydrate growth rate and low heat of formation during hydrate nucleation. The increase in TPCM changed from a rise to a fall at around 288 K, and then TPCM slowly decreased. This was because the slope of the TPCM change turned from positive to negative because hydrate was generated and covered the heat exchange coil surface as the static system.

Figure 10a–c shows the pictures of the crystal growth during this system. As shown in Figure 10a–c, hydrate was formed on the heat transfer surface near the bottom of the container. As time went by, the formed hydrate piled up on the bottom of the container. Figure 11a shows time evolutions of thermal energy storage rate in the system using the ultrasonic vibration, which depends on the frequency of 28 kHz and 56 kHz. The thermal energy storage rate . *q* in the static system is also shown for comparison. Time zero was defined as the time at which the circulating water is started to flow. Here, . *q* was proportional to the difference between Tin and Tout and thus shows almost the same trend as Tout change in the static system. In addition, the slope of change in the thermal energy storage rate was similar to that of the static system. However, as for the 28 kHz ultrasonic vibration, the maximum value of . *q* after the increase was about 1.5 times larger than that of the static system. This was because the hydrate was detached from the heat exchange coil by cavitation, which increased the growth rate of the hydrate. On the other hand, the maximum value of . *q* at 56 kHz was almost the same as that of the stationary system. Considering the size of hydrate adhesion, the cavitation effect is stronger at lower frequencies. This was why the maximum value of . *<sup>q</sup>* at 56 kHz was larger than that of . *q* at 28 kHz. After the peak of . *<sup>q</sup>*, the ultrasonic vibration system maintained a larger . *q* for a relatively longer time than the static system. This was because the hydrate growth was continued for a relatively long time after the hydrate covered the heat exchange coil due to the effects of cavitation. Figure 11b shows the time evolution of the thermal energy storage density. The result of the static system is also shown in Figure 11b. Time zero was defined as the time at which the circulating water is started to flow. As shown in Figure 11b, *q* in the ultrasonic vibrated system started increasing a bit earlier than the static system. This was because the cavitation enabled the hydrate to be formed at a higher temperature. The slope of *q* change in the ultrasonic vibration system is almost the same as with the static system. However, the bigger slope could be kept longer than the static system. This was also

because the hydrate growth could be kept by the effect of the cavitation. In the ultrasonic vibration system with a frequency of 28 kHz and 56 kHz, thermal energy storage density reached 126 MJ/m<sup>3</sup> , 111 MJ/m<sup>3</sup> in 10 h, respectively. Similarly, assuming the TBAAc hydrate has 200 MJ/m<sup>3</sup> of heat in hydrate formation and decomposition, the thermal energy storage density corresponds to a solid phase fraction of 63% and 55%, respectively.

**Figure 10.** Optical observation changing in the container with ultrasonic vibration of 28 kHz. (**a**–**d**) are the pictures at 2 h, 2.5 h, 3 h and 5 h after cooling water started to be circulated, respectively.

**Figure 11.** Time evolutions of the thermal energy storage rate and density in the system using an ultrasonic vibration. (**a**,**b**) depict the thermal energy storage rate and density, respectively.

#### *3.4. Comparison in All Systems*

Figure 12a shows a comparison in the time evolution of the thermal energy storage rate in all systems. In the agitated system, the thermal energy storage rate, . *q*, increased earlier than the other systems; . *q* increased at 1 h, 0.6 and 0.4 h for a rotation rate of 100 rpm, 300 rpm and 600 rpm, respectively. The forced convection generated by the agitation increased the convective heat transfer coefficient. This was why the temperature of the thermal energy storage medium dropped relatively earlier and hydrate formation occurred earlier, resulting in the early increase of . *q*. In addition, this forced convection caused rapid hydrate formation in the entire container, which released huge heat in a short period. This was why . *<sup>q</sup>* increased vertically. As the rotation rate increased, . *q* increased more. The maximum values of . *q* after the vertical increase were 47 kW m−<sup>3</sup> , 54 kW m−<sup>3</sup> and 69 kW m−<sup>3</sup> for a rotation rate of 100 rpm, 300 rpm and 600 rpm, respectively. These values are 5 to 8 times larger than that of a static system.

**Figure 12.** Time evolutions of thermal energy storage rate and density in all systems. (**a**,**b**) depict the thermal energy storage rate and density, respectively.

In the ultrasonic vibration system, the maximum of . *q* was 14 kW m−<sup>3</sup> , which is about 1.5 times larger than that of the static system, when the frequency was 28 kHz. This was because cavitation detached the adhesion between hydrate and heat exchanger coil, which could be the thermal resistance. As for the system with 56 kHz frequency, the maximum values of . *q* were almost the same as in the static system. This was because the effect of the cavitation was not as big as 28 kHz [38]. In summary, the agitation increased the rate of heat and mass transfer by creating forced convection, thereby shortening the hydrate formation time and increasing the hydrate growth rate. On the other hand, ultrasonic vibration kept the hydrate growth longer by the effect of cavitation.

Figure 12b shows a comparison in the time evolution of thermal energy storage density, *q*, in all systems. In the agitated system, *q* increased earlier and quicker than the other systems. This was because TPCM decreased quicker by the forced convection, which caused the quicker hydrate formation and growth for the entire container. In the ultrasonic vibration system, the *q* increase appeared to be relatively longer than the other systems. This was because the cavitation enabled hydrate to grow after hydrate covered the surface of the heat exchanger coil. The static system, the agitated system (rotation rate of 100 rpm, 300 rpm and 600 rpm), and the ultrasonic vibrated system (frequency of 28 kHz and 56 kHz) had the thermal energy storage density of 96 MJ/m<sup>3</sup> , 121 MJ/m<sup>3</sup> , 159 MJ/m<sup>3</sup> , 183 MJ/m<sup>3</sup> , 126 MJ/m<sup>3</sup> , and 111 MJ/m<sup>3</sup> , respectively. Assuming TBAAc hydrate has the heat for formation and decomposition of approximately 200 MJ/m<sup>3</sup> , each solid phase fraction is 48%, 61%, 80%, 92%, 63%, and 55%, respectively.

#### *3.5. For Industrial Utilizing of TBAAc Hydrate as PCM*

As written in Section 3.1., a range of the practical solid-phase fraction is from 40 to 50% [37]. The thermal energy storage density of ice is 140 MJ m−<sup>3</sup> when the solid phase fraction is 45%. As for TBAB hydrate, it has a thermal energy storage density of 100 MJ/m<sup>3</sup> when the solid phase fraction is 50% [33]. We regarded these values as a target thermal energy storage density. Given the thermal energy storage during nighttime of 10 h, the required time for thermal energy storage should be shortened. This was why the time to reach the target values was compared to each other system. In addition, the amount of the used electric power was estimated for more practical evaluation.

Table 3 shows the time that was needed to achieve the target thermal energy storage density. As for the target of 140 MJ/m<sup>3</sup> , it was achieved only when the rotation rates of agitation were 300 rpm and 600 rpm. When the rotation rates were 300 rpm or 600 rpm, it was 5.9 h or 2.9 h to achieve the target, respectively. Concerning the consumed electric power to achieve the target, it was 58 MJ/m<sup>3</sup> and 29 MJ/m<sup>3</sup> , respectively. As for the target of 100 MJ/m<sup>3</sup> , it was achieved in all systems where an external force was applied. When the rotation rates were 300 rpm and 600 rpm, it took about a third of the time of

the ultrasonic vibration system. However, the consumed electric power by the external force to achieve the target was 45 MJ/m<sup>3</sup> , 24 MJ/m<sup>3</sup> , 15 MJ/m<sup>3</sup> for the agitation system (100 rpm, 300 rpm, 600 rpm), and 5 MJ/m<sup>3</sup> for the ultrasonic vibration system at both frequencies (28 kHz, 56 kHz), respectively. Therefore, when the target of thermal energy storage density is 100 MJ/m<sup>3</sup> , the agitation system (300 rpm, 600 rpm) is more suitable for storing thermal energy in a short time, and the ultrasonic vibration system is more suitable as an external force for reducing the amount of electricity used.


**Table 3.** The time to achieve the target thermal energy storage density.

In this study, the thermal energy storage density in the agitated system was greatly improved by adding agitation. On the other hand, the amount of electricity used for agitation became larger than other systems. The agitation intensity was almost the same as that in a general low-viscosity agitation. However, the agitation Reynolds number was 25,000 at 600 rpm. This was due to the fact that the impeller power number of the three propeller blades was 1, which was smaller than that of other agitation blades [39]. Therefore, even if the agitation intensity was a general value, the agitation Reynolds number became larger, and energy consumption also became larger. It will be possible to optimize the rotations and electric power consumption by selecting a more suitable agitator at the similar experimental apparatus. The thermal energy storage density in the ultrasonic vibration system was not as large as that in the agitation system. The reason for this result is that it was difficult to resonate the ultrasonic transducer. To resonate the ultrasonic transducer, it is necessary to provide a natural frequency, but this frequency varies slightly depending on the environment in which the ultrasonic transducer is placed, for example, the materials in contact. Also, the natural frequency is also expected to change during the transformation of the aqueous solution into hydrate. Therefore, it is possible that the ultrasonic transducer was not able to resonate properly in this experimental system. A possible solution to solve this problem is to use a device that automatically follows the resonant frequency and keeps giving the resonant frequency to the transducer. Since the power consumption for ultrasonic vibration is about one-tenth of the power used for agitation, it is thought that the ultrasonic vibration system could be the most favorable way by solving the above problems.

#### *3.6. The Evaluation of the Heat Exchanger*

The thermal energy storage rate by the heat exchanger can be calculated as in Equation (3), where U is the heat transfer coefficient, A is the surface area of the heat exchange coil, and VPCM is the volume of the thermal energy storage medium. The logarithmic mean temperature difference, ∆Tlm, is calculated by Equation (4). As can be seen from Equation (4), the logarithmic mean temperature difference, ∆Tlm, is the average of the temperature difference between the thermal energy storage medium and the cooling water using the logarithm. This logarithmic mean temperature difference was substituted into Equation (3) to calculate the heat transfer coefficient for the experiments under each condition. The heat transfer coefficient was calculated to be about 104–10<sup>5</sup> W m−<sup>2</sup> K. The A/VPCM in this experiment was 15 m2/m<sup>3</sup> , and the UA/VPCM value was calculated to be around 103–10<sup>4</sup> W m−<sup>3</sup> K. For instance, a plate heat exchanger has a UA/VPCM that is similar to this value [40]. As the UA/V increases, the amount of heat exchange will also increase. It realizes a higher thermal energy storage density. In this study, the UA/V of the heat exchanger was comparable to that of other common heat exchangers [40]. This result demonstrates that the comparable thermal energy storage density in this study can be obtained using other common heat exchangers.

For the practical use of thermal energy storage using TBAAc hydrate on a larger scale, it is also expected that the comparable thermal energy storage density in this study would be available by using a heat exchanger which has the same level UA/V.

$$
\dot{q} = UA\Delta T\_{\rm lm} / \mathcal{V}\_{\rm PCM} \tag{3}
$$

$$
\Delta T\_{\rm lm} \frac{T\_{\rm out} - T\_{\rm in}}{\ln \frac{T\_{\rm PCM} - T\_{\rm in}}{T\_{\rm PCM} - T\_{\rm out}}} \tag{4}
$$

#### **4. Conclusions**

Kinetic characteristics of thermal energy storage using tetrabutylammonium acrylate (TBAAc) hydrate were practically and experimentally evaluated for practical use as Phase Change Materials (PCMs). During the experiments, the adhesion of hydrate on the surface of the heat exchanger coil was found, which increased the thermal resistance between the heat exchanger and thermal energy storage medium. This was why, as the external forces, we added a mechanical agitation (rotation rate of 100 rpm, 300 rpm, and 600 rpm) and ultrasonic vibration (frequency of 28 kHz and 56 kHz) on each system. It was revealed that the external forces improved the thermal energy storage kinetic characteristics. When the agitation rate was 600 rpm, the system achieved thermal energy storage of 140 MJ*/*m<sup>3</sup> in 2.9 h. This value is comparable to the ideal performance of ice thermal energy storage when its solid phase fraction is 45%. The energy consumption for agitation was 28 MJ*/*m<sup>3</sup> to achieve this value. The thermal energy storage of 100 MJ*/*m<sup>3</sup> , which is a TBAAc solid phase fraction of 50%, was achieved in 1.5 h and 5.8 h with mechanical agitation (600 rpm) and ultrasonic vibration (28 kHz), respectively. The energy consumptions to achieve this target value were 15 MJ*/*m<sup>3</sup> and 5.2 MJ*/*m<sup>3</sup> , respectively. In summary, the agitation increased the rate of heat and mass transfer by creating forced convection, thereby shortening the hydrate formation time and increasing the growth rate of the hydrate. In addition, ultrasonic vibration could keep the hydrate growth time longer by the effect of cavitation.

In this system, the UA/V (U: thermal transfer coefficient, A: surface area of the heat exchange coil, V: volume of the thermal energy storage medium) was approximately 1.0 <sup>×</sup> <sup>10</sup>−4–10−<sup>3</sup> W m−<sup>3</sup> K. This UA/V of the heat exchanger was comparable to that of other common heat exchangers. Even if the system would be scaled up and different from this study, the same level of thermal energy storage as this study could be obtained by using a heat exchanger of which UA/V is comparable to this study. As for the ultrasonic vibration system, thermal energy storage density was one-tenth of the agitated system. A device that enables an oscillator to resonate and break the hydrate adhesion is necessary to increase the thermal energy storage density. Then, that system would be most favorable in terms of the storing performance and energy power consumption.

**Author Contributions:** Conceptualization, H.K., H.T. and R.O.; methodology, H.T., S.I., H.N. and R.O.; formal analysis, H.T., S.I., H.N. and R.O.; investigation, H.K., H.T., S.I., H.N. and R.O.; resources, H.T. and R.O.; data curation, H.K., H.T., S.I., H.N. and R.O.; writing—original draft preparation, H.K. and R.O.; writing—review and editing, H.K. and R.O.; visualization, H.T. and R.O.; supervision, R.O.; project administration, R.O.; funding acquisition, R.O. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work has been supported by a Keirin-racing-based research promotion fund from the JKA Foundation (Grant Number: 2020M-195) and a part of Low Carbon Technology Research and Development Program for "Practical Study on Energy Management to Reduce CO<sup>2</sup> emissions from University Campuses" from the Ministry of the Environment, Japan.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors would like to thank Haruki Sato, professor emeritus and Masao Takeuchi, Keio University, for the encouragement on this work.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

