1. Introduction
The European Union’s (EU) goals for the increased penetration of renewables in new and renovated buildings have increased the interest of the market and academia in residential-scale renewable-based systems [
1]. In fact, the EU sub-target of an indicative 1.3% yearly increase of renewables share in heating and cooling in residential buildings, calculated over a period of 5 years starting in 2021, highlights the importance of energy transition in residential applications [
2,
3]. Among other renewable energy sources, solar thermal energy has several advantages, including its ease of coupling with heating and domestic hot water (DHW) systems and low installation costs owing to its large market [
4].
As of 2021, renewable heating is almost solely covered by solid biofuels, accounting for an average of 29.2% of the total space heating loads on an EU basis, based on data from Eurostat [
5]. Regarding DHW, 13.9% of the total demand in 2021 was covered by renewables and biofuels [
5]. Among other types of renewables, solar thermal energy can offer a relatively cheap solution to cover part of the DHW and space heating loads, resulting in a larger reduction of conventional fossil fuel-based technologies. In order to minimize the thermal spikes and also tackle the time variability of solar energy, thermal energy storage is extensively applied in coupling to solar harvesting systems [
6].
Thermal energy storage (TES) systems can be distinguished into three types: sensible, latent, and thermo-chemical storage [
7]. In sensible heat storage, there is no phase change in the storage medium. In the simplest configuration, sensible heat storage is realized by a single pressurized tank that is filled with the heat transfer fluid (HTF) that also circulates in the solar collectors. Alternatively, the storage tank can be part of a secondary, intermediate heat transfer circuit that is used for transferring heat from the solar collectors to the consumer/building. In the latter case, the HTF of the primary circuit flows through a helical coil inside the tank and charges it, causing the temperature of the storage medium to increase [
8]. Other types of sensible heat storage include underground heat storage, rock beds, and storage using concrete modules [
9,
10]. In latent heat storage, the storage medium is a phase change material (PCM), which is solidified and melted during charging and discharging phases, respectively [
11]. Because of the large enthalpy change associated with phase change, the energy density of latent heat storage systems is much higher compared to that of sensible storage systems, so they can be more compact [
12]. Finally, thermo-chemical energy storage is based on a reversible endothermal chemical reaction [
13]. The most important advantage of thermo-chemical energy storage is its high storage capacity, which can be several times higher than that of conventional sensible storage systems [
14,
15].
In most solar-driven residential applications, sensible heat storage is used owing to its simplicity, high market availability, and low costs [
16]. Several studies discuss the performance characteristics, modeling aspects, and system integration concepts of sensible storage systems [
17,
18,
19]. Raccanello et al. [
8] evaluated different order models for several types of single-tank storage systems to assess the reliability of simplified modeling approaches for integrating storage tank models into more complex systems without severely affecting the computational cost. Tian and Zhao [
20] conducted a detailed review of different types of solar thermal collectors and high-temperature thermal energy storage systems.
Ismaeel and Yumrutas [
21] simulated the performance of a solar-assisted heat pump connected with an underground thermal energy storage system for wheat drying. For a defined solar field area, the storage tank was sized in order to retain a satisfactory temperature range throughout the year. Syed et al. [
22] experimentally evaluated a solar absorption cooling system coupled with a 2 m
3 stratified storage tank in the city of Madrid, Spain. The operation of a 35 kW nominal capacity absorption chiller, driven by a solar field of 50 m
2 flat plate collectors (FPC), was prolonged by the use of the storage tank, leading to a total daily operation lasting approximately 7.3 h. Karim et al. [
23] evaluated the performance of stratified storage tanks for heating/cooling applications and concluded that tanks with higher height-to-diameter ratios tend to reduce mixing and thus reduce heat losses. In the same direction, Pintaldi et al. [
24] evaluated the energetic performance of sensible and latent heat storage scenarios for solar cooling applications. The analysis found that, for the evaluated scenarios, a minimum specific collector area of 2 m
2 per kW of cooling capacity is required for achieving solar fractions higher than 50%. Jung et al. [
25] assessed control strategies for the optimal operation of a heating system consisting of a heat pump and a thermal storage tank for use in Seoul, South Korea.
Çomaklı et al. [
26] evaluated the influence of storage tank sizing on solar water heating systems. The analysis revealed that an increase in the storage tank capacity may enhance the solar collector’s efficiency, but simultaneously, as expected, the average water temperature in the tank decreases. Therefore, an optimal design exists per case; for the scenario of the Turkish DHW standards, a storage tank volume to solar field area ratio of 50–70 L/m
2 was defined as optimal by the authors of this study. Similarly, Li et al. [
27] optimized the storage tank capacity for integration in a solar heating system to be installed in a typical building in Xi’an, China. The optimal storage tank volume to solar field area ratio was determined to be 10–20 L/m
2.
The aforementioned systems focused mostly on short-term storage, which is a totally feasible solution for regions with high solar irradiance, even in the colder months of the year. However, in colder climates (central and northern Europe), solar irradiance is mostly available during hotter periods of the year, and thus there is very limited concurrence with space heating loads. Therefore, the exploitation of solar energy in such regions is only possible with the implementation of seasonal thermal energy storage (STES).
Despite several studies on latent and thermo-chemical storage, sensible storage is the only economically viable solution for STES [
28]. One of the first commercial solar-driven STES was built in Hamburg in 1996 [
29]. The system was based on an underground concrete tank filled with 4500 m
3 of water, achieving in design conditions a solar fraction of 49%. Since then, several studies have been conducted, both through simulations and experiments on STES systems. Terziotti et al. [
30] simulated in TRNSYS the performance of a sand-based STES for a five-story student housing complex located at Virginia Commonwealth University, USA. The simulations revealed that in a building with lower heating loads, the solar fraction reached a value of up to 91%. Sweet et al. [
31] modeled in TRNSYS a solar-powered underground STES system for a residential application in Richmond, Virginia, USA. By evaluating different floor areas of the tested building, it was found that the optimal sizing of the investigated system resulted in a reduction in conventional system consumption of up to 77%. Antoniadis and Martinopoulos [
32] conducted a study in TRNSYS to evaluate a solar-driven system using a STES for a 120 m
2 single-family building in Thessaloniki, Greece. The system was able to reach a solar fraction of 52.3% with respect to the heating loads. Hailu et al. [
33] reported a reduction of more than 40% in the heating loads by the implementation of a solar-driven STES in a two-story house located in Alaska, USA. Hesaraki et al. [
34] evaluated a STES coupled with a heat pump connected to low-temperature space heat emissions for a single-family building located in Stockholm, Sweden. The analysis showed that the optimal ratio of storage capacity to solar field area is approximately 5 m
3/m
2. Li et al. [
35] simulated a solar thermal heat pump system coupled with STES to cover the space heating and DHW loads of a six-story dorm building with a total area of 2252 m
2, located in Beijing, China. The proposed system was found to improve the monthly coefficient of performance (COP) by 12.8% compared to a conventional heat pump system. Drosou et al. [
36] reported the performance of a solar-driven cooling/heating system coupled with an underground STES. The system is used to cover the space heating and cooling loads of a 427 m
2 office building located in Athens, Greece. The measured solar fraction of the system, which uses an absorption chiller and a conventional heat pump, was as high as 70%.
Gabrielli et al. [
37] carried out an optimization procedure to size multi-energy systems coupled with STES by implementing two novel mixed integer linear program models. The proposed methodology was evaluated for the case of a residential application, combining options of seasonal storage technologies, including STES, hydrogen, and battery storage. In fact, the analysis showed that for larger emission savings, thermal storage reported optimal performance only covering the peak loads, while hydrogen seasonal storage was the optimal technology to handle the base loads of the residential building on an annual basis. McKenna et al. [
38] assessed the techno-economic performance of a STES for a typical German residential district. The analysis showed that a 60% renewable heat supply fraction does not significantly increase costs and is therefore viable as a solution.
In this study, a thermal energy storage system driven by evacuated tube collectors (
Figure 1) is investigated both numerically and experimentally. The distribution of the available solar heat and the heat stored in the STES is realized via a residential heating and DHW distribution system based on a thermally stratified water tank. The tank is working as a diurnal thermal energy storage device, coupled with a natural gas boiler. The solar heat is used to charge the Combi storage tank and/or the STES. Space heating is provided to the building through a floor heating system, while DHW is supplied via a dedicated heat exchanger. STES is used either as a backup or to cover peak loads on days with inadequate solar irradiance. A condensing gas boiler operates as a backup thermal energy source to ensure thermal comfort at any time of the year.
The prediction of the performance of the stratified tank that is coupled with the distribution system is crucial for the overall system efficiency. For this purpose, a simulation model was developed in TRNSYS 18 [
39] along with an experimental test rig. The experimental results are also used to calibrate the simulation model and assess its accuracy.
Thermal modeling of sensible heat storage tanks with TRNSYS software has been extensively investigated in the literature [
40]. The novelty of this study, compared to the existing literature, is the use of experimental data for verifying the performance of the system, which is also based on a non-standard component for modeling the effect of tank stratification. Minimize the simulation errors related to thermal nodes close to the inlet/outlet ports of the stratified tank. Furthermore, most of the existing studies on sensible storage have not considered cases including seasonal storage systems coupled with underfloor heating, which is the scope of the investigated concept.
In that context, this study aims at presenting the development, experimental evaluation, and calibration of the employed simplified stratified tank model parameters in coupling with specific thermal loads associated with the respective DHW and heating demands as defined by the EU standards. As a result, a detailed overview of the model’s accuracy during different mixing, charging, and discharging phases is thoroughly investigated.
3. Results and Discussion
Regarding the control strategy, it was realized with an equation component. The control signal of the boiler’s pump is generated by a Type 1502 thermostat component by comparing the temperature estimation, T
CS4, with a setpoint of 45 °C. During the periods that the HWBP (see
Figure 2) is operating, the three-way valves are rerouting the flow to bypass the boiler. On the contrary, when the HWBP is out of operation, the control of the three-way valves follows the actual system’s operational strategy.
For the charging of the lower parts of the combi-storage tank, which are used for the space heating loads, a dedicated directing valve is used. During periods with no DHW and space heating loads, and while the HWBP is off, the control of the three-way valves directs the flow towards the boiler and then bypasses the space heating heat exchanger to feed the heated flow in the lower parts of the tank. This charging procedure continues until TCS2 reaches the desired setpoint of 45 °C. All different setpoint cases are saved in an equation component, and each dedicated setpoint is used as input for the boiler component, according to the respective mode of operation.
The scope of the experimental test rig was twofold. Firstly, it was developed to assess the potential of the combi-storage tank for efficiently storing heat at different temperature levels and covering the heating and DHW loads of a building. Secondly, the sets of experiments and their results were used to evaluate the accuracy of the non-standard TRNSYS component towards its implementation in more complicated systems.
It is here noted that the error analysis of the experimental results is out of the scope of this study. The analysis regards the comparison of the simulated temperature levels with the values directly measured by the temperature sensors. Hence, the uncertainty of the experimental results is directly related to the accuracy of the temperature sensors reported in
Section 2 and is not propagated through further calculations. As regards the measured water flow rate values, these are given as an input to the simulation model alongside the uncertainty of the respective sensors. With respect to the simulations, they were conducted with a relative convergence tolerance of 0.1% for all model variables.
3.1. Nodes’ Number
As already stated, the non-standard component Type 340 was used for the combi-storage tank simulations, which is based on the ‘MULTIPORT’ store model [
49]. The modeling of the tank’s stratification followed the isothermal nodes approach, dividing the tank into a number of N fully mixed finite volumes and applying respective energy balances [
50]. The nodal approach was preferred over the plug flow approach, as it better simulates the tank’s thermal stratification, according to the study of Allard et al. [
51].
The first step in the experimental evaluation of the developed model concerned the selection of the node number to be used in the model. In order to simulate the loads interacting with the combi-tank in the test rig of
Figure 2, a constant temperature source was implemented in TRNSYS via a Type 534 cylindrical storage tank without any heat exchangers considered [
52]. The node number of the cylindrical storage tank was set at 1, corresponding to a fully mixed tank, while the heat loss coefficient was set at 0. The tank’s volume was assumed to be twenty times larger than the combi-storage tank’s volume. The selection of such a large theoretical volume allowed for a very high thermal inertia of the cylindrical tank compared to the evaluated combi-storage tank, and thus a constant supply temperature defined during simulation. A type 114 constant-speed pump was introduced to set the flow towards the combi-storage tank. An overview of the simplified combi-storage tank model for the nodes’ evaluation is shown in
Figure 7.
In order to find the optimal node number, three relevant charging experiments were conducted and parametrically simulated in order to evaluate the model’s reliability for various numbers of nodes. The operating conditions of the tests are summarized in
Table 3.
In Test No. 1, a charging phase of the combi-storage tank was considered. An initial temperature of 25 °C was set within the tank, while the hot charging stream was supplied by the upper port (port 5 based on
Figure 2) at a temperature of 65 °C. Test 1 was completed when the lower layers of the tank reached a temperature of 60 °C.
Test No. 2 involved a charging process similar to Test No. 1, with the charging stream again connected to the upper port of the tank, supplying water at 65 °C. The difference in Test No. 2 was the fact that the tank’s initial temperature was 45 °C, while the charging stream had a higher flowrate. Finally, Test No. 3 investigated the charging of the tank at a lower temperature (45 °C) with the supply stream to feed the tank in the middle level of the tank (port 3 based on
Figure 2).
Eight different simulation cases with a number of nodes equal to 10, 20, 30, 60, 90, 120, 150, and 180 nodes, respectively, were compared to three different charging experiments. The Root Mean Square Error (RMSE) was estimated, as shown in Equation (1), by comparing each temperature simulation result for the five different ports at the time the test’s completion setpoint is reached with the experimental results for all three experiments. The RMSE results for each number of nodes tested and for the three tests are presented in
Figure 8. As can be observed, for a number of nodes higher than 60, the improvement is relatively small. In fact, this comes in agreement with the study by Wischhusen [
53], which mentions that for a relatively low number of nodes (
n < 20), the buoyancy effects harm the accuracy of the model [
54]. Hence, to minimize, as much as possible, the calculation time, the number of nodes used in the following tests was set equal to 60. The final model parameters are reported in
Table A4 of the
Appendix A.
3.2. DHW Demand Profile Sensitivity Analysis
As already stated in the previous section, the DHW profile was modeled based on the Type 14 “forcing function water draw”. In order to model the profile, the daily interval is divided into 96 equal periods. The DHW demand is assumed to be generated at the beginning of each period.
For the test rig’s case, under a constant flowrate of 0.18 L/s and with a temperature rise of 25 K, the equivalent liters of hot water, corresponding to an energy consumption of 5.845 kWh (
Table A1), were equal to 200.4 L. In order to compare the model’s DHW profile area over a 24-h period with the aforementioned target volume, a quantity integrator (Type 24) connected with a scope (Type 76) was used.
When a timestep of 60 s was used, an absolute difference of 145.2 L was observed (the absolute difference between the estimated volume by TRNSYS and the equivalent volume of 200.4 L). This large deviation is owed to the calculation approach of the DHW profile by the TRNSYS algorithm. In fact, the algorithm calculates the value of the forcing function at the beginning and ending of the time step, joining the two points via a line. Hence, the rectangular area of the forcing function is transformed into a trapezoid, leading to an overshooting of energy consumption. In order to overcome this deviation, two approaches could be applied:
In order to make clearer the impact of the geometrical reforming in the correction of the estimated equivalent DHW volume, a qualitative example of the estimation derived from the original rectangular-shaped tapping demand and the corresponding estimation by the geometrical reforming are shown in
Figure 10a,b, respectively.
Based on the above, the time duration (EB) and the time step used in the DHW profile need to be determined to minimize the error with the trapezoid reforming. Hence, a dedicated sensitivity analysis was conducted. Four different (EB) duration values were evaluated, namely 0 s, 5 s, 10 s, and 20 s. By applying the four different durations under seven different time steps (1, 5, 10, 20, 30, 50, and 60 s), the results presented in
Figure 11 were obtained. As can be seen in
Figure 11, the optimal results were obtained for a time duration (EB) of 10 s, while for all the rest of the time durations, smaller time steps enhanced the calculation deviation. On the other hand, for time steps larger than 30 s, the smaller deviation was recorded for a time duration of 0 s (i.e., without reforming). However, this is attributed to the fact that an entire tapping use (at around 21:15) failed to be calculated. In order to prove this conclusion, the water draw profile of
Table A3 was used as an input scenario for a timestep of 30 s and for EB = 0, as depicted in the dedicated run of
Figure 12 (denoting with red the missed tapping). Regarding the time step for a specific time duration, time steps of equal or lower value than the time duration result in a constant result. On the contrary, for larger time steps, the result is unstable.
However, using the trapezoid reforming, the time period during which the demand is non-zero is larger than the time period (BC). In order to bypass this setback, the EA linear fraction is shifted parallelly until E==B. A second sensitivity analysis was conducted with the adapted trapezoid under the same time-step scenarios as in
Figure 11. In this sensitivity analysis, the time duration EB was considered under four scenarios: 0 s, 1 s, 5 s, and 10 s. The control signal of the DHWP was multiplied by the pump’s constant flowrate, and the result was integrated over a single-day period. The results are depicted in
Table 4. As shown, the accuracy of the results is heavily influenced by the time step and the use of the reforming, while it is not affected by the time duration (EB). This conclusion can be more easily visualized in
Figure 13, where the relative error with and without the application of geometrical reforming is shown. Based on the above results, a value of 10 s was selected as the optimal time duration (EB), while the time step was selected to be 10 s as well, in order to minimize the computational time. These settings were applied in all the case studies presented in the next section.
3.3. Charging of Entire Tank
Within the context of the combi-storage tank use for DHW and space heating, four additional experiments were conducted and are presented in the following three subsections. The reason for these experiments was twofold: the further validation of the simulation model and the experimental assessment of the proposed combi-storage tank for use in residential space heating and DHW applications.
The first experiment involved the charging of both sections of the tank in two subsequent steps, as shown in
Figure 14. In the first step (from 0s up to 3170 s), the upper part of the tank was charged via port 4 (
Figure 2). The flowrate of the HWBP was set at 0.14 L/s, while the gas boiler setpoint was equal to 60 °C. The initial conditions in the combi-tank included a uniform temperature of approximately 36 °C, and the target values were 58 °C for the temperatures
and
. In the second step of the test (from 3170 s up to the end), after the upper part was charged, the lower part of the tank, dedicated to covering the space heating loads, was also heated via port 1 with a flow rate of 0.25 L/s. The respective boiler setpoint was set at a water temperature of 50 °C.
As shown in
Figure 14, in the first step of the charging process, there is a good agreement between the experimental and the simulation results after approximately 1800 s. By that time, all five experimentally measured temperatures fall within the range of the estimated values by the simulations, for considered errors of ±0.6 °C. As the heat is introduced in the tank via port 4, temperatures,
and
. rise from the start in a close range, while
has a delayed increase in the temperature. Finally,
and
are practically not influenced throughout this step, revealing a good stratification behavior of the tank. Once the second step is initiated, and heat is fed via the tank’s lower port (port 1),
and
show a sharp rise; on the other hand,
is also influenced, tending to mix with the lower parts of the tank and therefore reducing its temperature. The stepwise change in the flowrates results in a deviation between the model and the experiments, in particular for
, which re-converges after approximately 1000 s. Finally,
and
. tend to be unaffected by the space heating charging, a behavior that comes in agreement with both experimental and simulation results.
3.4. Space Heating Test
The next set of experiments aimed at testing the space heating performance of the experimental system and the simulation model, respectively. The supply flow rate for this experiment was equal to 0.14 L/s, corresponding to a temperature difference equal to 10 °C (considering a supply temperature from the tank of 43 °C and a return temperature of 33 °C). The total thermal power corresponded to approximately 6 kW. This power supply to the heating system was estimated to be the maximum thermal power needed, based on typical nZEB building simulations [
46,
55]. When space heating is supplied exclusively from the combi-storage tank, the space heating supply temperature to the heat exchanger (
Figure 2) is regulated by a mixing valve, so that the inlet to the hypothetical floor heating system is kept at 38 °C. As can be observed by the results of
Figure 15, although the model accurately predicts the final temperature profiles of the tank, there are some deviations occurring due to the lack of detailed inertia modeling of both the tank and the rest of the system components. This is especially evident as far as
is concerned, which is influenced the most by inertia phenomena, as was also observed in the experimental data. In this scenario,
shows a stronger trend to mix with the lower section of the tank compared to the simulation. Thus, the simulation model presents a slightly better stratification behavior than the experimental procedure actually revealed.
3.5. DHW Test
In order for the DHW tests to be conducted, the combi-storage tank is priorly fully charged (
and
equal to 58 °C). The first test conducted with respect to DHW investigated the maximum power (worst case) scenario, which corresponded to the largest power consumption of the used water profile (as mentioned in
Section 2.4, load profiles for cycle no. 2 of the European Standard [
41] were used). Hence, an 18.81 kW load was considered for a total duration of 18.65 min, which corresponds to the total consumed energy of 5.85 kWh. The boiler was turned on when
55.5 °C, with a nominal DHW flowrate of 0.3 L/s and a ΔΤ equal to 15 Κ. The above selection in the ΔΤ and the corresponding flowrate were partially dictated by the available cold-water supply in Athens, Greece, during the time of the experiments (July 2020) and were equal to approximately 27 °C. The nominal supply and return temperatures were set equal to 58 °C and 33 °C, respectively.
As seen in
Figure 16, the used working conditions maintained the temperatures
and
above 42 °C during the experiment, despite the low temperature return of the DHW consumption. However, in the upper part of the tank, temperatures decreased rapidly, resulting in the need for additional heat input after 490 s. With respect to the simulation model results, as shown in
Figure 16, the rapid temperature changes, in particular,
and
, were not accurately predicted by the model, which required approximately 1000 s and close to steady state conditions in order to converge at an acceptable level. These results highlight that such storage tank models can be used for steady-state (on-/off-design) simulations, which in particular for solar-driven systems commonly have hourly or larger timesteps but should be avoided for dynamic and semi-dynamic simulations as they tend to deviate considerably.
In order to find a suitable switch-on temperature criterion for the boiler that minimizes the number of cut-off/switch-on cycles while ensuring a reliable supply temperature for SH and DHW, the combi-storage tank’s behavior was further investigated by conducting a second DHW test. In this case, a combination of both and ( < 53 °C) was introduced in the control criteria in order to be able to deal with the temperature drop due to DHW consumption and provide hot water at acceptable temperature levels. Moreover, two alternative criteria were set for the boiler to be turned on in order to deal with the temperature drop due to heat losses and internal mixing ( < 50 °C or < 56 °C). Eventually, the boiler was turned on in case any of the aforementioned three criteria were not satisfied. The boiler was then turned off again when was heated above 58 °C.
For the purposes of this experiment, an equivalent profile was created by grouping the original DHW profile of the European standard’s consumptions into eight different groups, as presented in
Table A3 of the
Appendix A. Two complete cycles were executed in the experiment to ensure that the performance of the tank was adequate, even if it was partially discharged at the beginning of the cycle.
The results of this experiment are presented in
Figure 17 along with the respective simulation model results. The used criteria are considered to be adequate, as
remains well above 50 °C in all cases, which is an acceptable scenario for DHW use. However, as can be observed, the boiler is needed to operate with relatively high temperature input and for small ΔΤ (minimum
was 47.1 °C and minimum
was 45.9 °C), not allowing the condensation to take place and thus dramatically decreasing the boiler’s efficiency. With respect to the simulation model, similarly to the previous cases, the transient states are not adequately predicted by the model, with the temperatures of the upper levels of the tank being constantly over-predicted and the lower levels’ temperatures being under-predicted. This longer experiment, with more fluctuations, highlights again the incompatibility of the model for transient simulations. However, it has to be noted that in cases where transient phenomena are not of concern and the time-step of the simulations is larger, the model can adequately predict the storage’s behavior since, as shown in
Figure 17, eventually the temperature profiles converge to the temperatures measured during the experiments.