This research evaluates the performance of a Phase Change Material (PCM) battery integrated into the climate system of a new transparent meeting center. The main research questions are: a. “Can the performance of the battery be calculated?” and b. “Can the battery reduce the heating and cooling energy demand in a significant way?” The first question is answered in this document. In order to be able to answer the second question, especially the way the heat loading in winter should be improved, then more research is necessary. In addition to the thermal battery, which consists of Phase Change Material plates, the climate system has a cross-flow heat exchanger and a heat pump. The battery should play a central role in closing the thermal balance of the lightweight building, which can be loaded with hot return or cold outdoor air. The temperature of the battery plates is monitored by multi-sensors and simulated by the use of PHOENICS (Computational Fluid Dynamics) and MATLAB. This paper reports reasonable agreement between the numerical predictions and the measurements, with a maximum variance of 10%. The current coefficient of performance for heating and cooling is already high, more than 27. There is scope for increasing this much further by making use of the very low-pressure difference of the battery (below 25 Pascal), low pressure fans and the ventilation system as a whole.

The demand for low-energy buildings has encouraged the development of new technologies for Heating, Ventilation, and Air Conditioning (HVAC). In this sense, thermal batteries seem to be ideal for dealing with variable heat sources once the systems reach a steadier operation level to save energy [

The literature shows various applications in the built environment sector. However, the focus of most of this literature is on liquid/PCM heat-exchangers, and studies utilizing three phases, gas, liquid, and solid, are somewhat more limited. Much literature on liquid/PCM heat-exchangers evaluates the effect of surface-enlargement like with the addition of fins [

This article focuses on using a PCM in a vertical air-handling unit with Heating, Ventilation, and Air Conditioning (HVAC) in a climate tower. This tower was originally designed as a thermal chimney in which the PCM is used as a buffer to reduce temperature fluctuations of the lightweight building itself [

The aim of the research is to investigate to what extent it is possible to model the existing PCM-battery and how it can be integrated in the building as a whole. CFD is used to get a better understanding of the fundamental background of the physical processes and is a reference for the MATLAB-model. With the MATLAB-model calculation time can be reduced, the energy savings over a whole year can be evaluated and an integration of the model with all the relevant components of the building is possible to find the most optimal control strategy.

The building in which the battery is applied is around 315 m^{2} and is designed to be as transparent as possible for a passive building with active components (see

The thermal battery has 2,106 kg of PCM in the climate tower with a minimum cooling and heating power of 7.2 kW. The storage capacity of the PCM is 181 kWh (18.5 m^{3} natural gas equivalent). Additionally, the heat pump has a nominal cooling power of 30 kW and heating power of 15 kW. This analysis is presented in ^{2}, and so a maximum airflow of 10,800 m^{3}/h produces an air velocity of 2.3 m/s between each panel.

Some (unforeseen) limiting parameters for the airflow are stated below:

The noise level from the fans should not exceed 40 dB(A) according to the Dutch building code. The high-efficiency EBM-Past fans (VBH0450PTTLS K3G450-PAA23-71) in the climate tower have a source noise of 87.6 dB(A) at 8,000 m^{3}/h^{3}/h is still acceptable to the building owner. This amount of air is only used with a very high occupancy of 240 people in a reception-like meeting. During such a meeting, with the fans producing an indoor sound level of 49.2 dB(A), the noise of talking people is comparable. During the day, for this laboratory-like location, a 55 dB(A) noise level at a distance of 40 m is still acceptable to the people living in the surrounding environment. The main problem is not the noise level in general but a disturbing sharp, so-called Aeolian sound via a grille or by the PCM battery. The Aeolian sound is already present at velocities below three m/s, the maximum air velocity via the valve. It is not clear yet how to solve this problem.

The minimum airflow for operating the heat pump is limited to prevent malfunction. Thus, the airflow via the fans cannot be reduced below circa 3,500 m^{3}/h. This will be solved by connecting the heat pump to energy storage in the ground in the near future.

The material in the panels is Calcium Chloride Hexahydrate, which is encapsulated within Crystal Storage Panels (CSP) made of HDPE (high-density polyethylene) material, as shown in ^{3} for the one shown in _{2} by weight [

Description | Value | Unit |
---|---|---|

Dimension (w × l × d) | 275 × 570 × 13 | mm |

Type of filling | PCM 21, 23, 24, and 26 | - |

Mass | 1.8 | kg/panel |

Density | 1,000 | kg/dm^{3} |

Latent heat | 310,000 | J/kg |

Specific heat-solid phase | 2,100 (assumed)* | J/(kgK) |

Specific heat-liquid phase | 2,100 (assumed)* | J/(kgK) |

Thermal conductivity | 1.0 | W/(mK) |

Thermal expansion coefficient |
0.0001 |
m^{3}/m^{3}K^{2}/s |

Material panel (casing) | HDPE | - |

Wall thickness | 0.6 | mm |

Heat transfer coefficient HDPE | 0.5 | W/(mK) |

Specific heat HDPE | 1,400 | J/(kgK) |

Note: *This value has been chosen after starting with values from literature with the results by comparing measurements with CFD simulations.

Three approaches are used to evaluate the PCM battery performance. First, the temperature of the PCM panels is evaluated experimentally, relying on the set of sensors installed in situ, as described in

The PCM surface, supply, and exhaust air temperatures are measured, as are the air flow rate and relative humidity. These are collected at 5-min intervals by a set of sensors installed in the system. Additionally, temperature sensors are installed at the inlet and outlet of each sub-system, such as the heat exchanger, the PCM battery, and the heat pump. The specifications of the installed temperature and air velocity sensors and their operational features are listed in

Temperature sensor | Air velocity sensor (pressure) | ||
---|---|---|---|

Parameter | Description | Parameter | Description |

Type of sensor | NTC10K | Type of sensor | Belimo 22ADP-184 |

Range | −30°C/150°C | Characteristic | 0–10 V = 0–2500 Pa |

Resistance | 176.68–184.80 Ω | K-factor | 240* |

The air velocity is calculated from the pressure difference and the K-factor, which is defined by^{3}/h. For instance, with a K-factor of 240 and a pressure difference of 40.1 Pa, the airflow ^{3}/h. In

The 1,170 panels in the PCM battery are divided into two decks separated by a metallic plate, and panels fill the unused third deck with a lower phase-change temperature. The temperature of the PCM battery is measured with 18 sensors, of which 9 are located on the upper deck and 9 on the lower deck. The temperature sensors are placed within the 4 mm gap between the panels in each deck. The sensors are located to facilitate the temperature profile measurement over the entire PCM battery, as shown in

The overall temperature of the PCM battery is calculated by taking an average of the measurements provided by the 18 sensors. The measured data are provided in real-time by a monitoring interface that can be accessed online through its corresponding website (see

A transient, three-dimensional CFD model of the PCM battery is developed using the general-purpose CFD code PHOENICS to solve the conservation equations for mass, momentum and energy. The battery comprises the liquid and solid regions of the PCM which are housed within solid casings, and these in turn undergo conjugate heat transfer with the air flow in the adjacent channels. The PCM’s undergo melting and solidification processes with natural convection in the melt, and they are modelled using the enthalpy-porosity formulation of Voller et al. [

In the air channels between the PCM panels, the following uniform values are used for the physical properties of air: k = 0.023 W/mK, C_{pa }= 1,005 J/kgK, ρ = 1.189 kg/m^{3} and ν = 1.154.10^{−5} m^{2}/s.

The temperature difference between the PCM panel and the air and the convective heat transfer coefficient (

For laminar flow, the following expressions are in use for making estimates of the heat transfer coefficient:

However, even at Reynolds numbers as low as 800, perturbations in the channel flow can initiate a transition to a turbulent regime [^{2}K. A high convective heat transfer coefficient can lead to a rapid rise in the air temperature in the cavity. When the airflow is laminar, the ^{2}K.

The value of

With 100% heat transfer efficiency and assuming a linear rise of temperature in the cavity, the convective heat transfer coefficient from a steady-state CFD simulation is calculated as ^{2}K. This ^{2}K) and turbulent (18.3 W/m^{2}K) flow.

Turbulence is ignored in the PCM melt, whilst the Chen-Kim k-ε model [_{c} = k/δ where δ is the wall-to-node distance. Suppose the near-wall flow is in the turbulent regime, then the local heat-transfer coefficient is computed from a wall function, so that h_{c} = St ρ U_{r} C_{p} where U_{r} is the near-wall air velocity and St is the local Stanton number, which is computed from a modified Reynolds analogy [

The foregoing mathematical model has been incorporated into the general-purpose, commercial CFD code PHOENICS for solution on a structured Cartesian grid using a finite-volume numerical method. Fully implicit backward differencing is employed for the transient terms and central differencing for the diffusion terms. The convection terms are discretized using hybrid differencing [

The set of finite volume equations are solved iteratively using the SIMPLEST [

The numerical solution procedure requires appropriate relaxation of the flow variables in order to procure convergence. Two types of relaxation are employed, namely inertial and linear. The former is normally applied to the velocity variables, whereas the latter is applied to all other flow variables, as and when necessary.

The convergence requirement is that for each set of finite volume equations the sum of the absolute residual sources over the whole solution domain is less than 1% of reference quantities based on the total inflow of the variable in question. An additional requirement is that the values of monitored dependent variables at a selected location do not change by more than 0.1% per cent between successive iteration cycles.

The continuity and momentum equations for each fluid phase can be written in generic form as follows:

where ρ is the density, τ is the viscous stress tensor, _{g}_{m}_{g}_{g}

The energy equation for both fluid and solid regions can be written as follows:_{p}T_{p}

The PCM considered is Calcium Chloride Hexahydrate, and this is represented in the CFD model by employing a linear phase-change equation, where the evolution of latent heat is expressed as a linear function of temperature based on an effective specific heat capacity corresponding to the estimated solid fraction. Therefore, the specific enthalpy

where ^{6} which controls the degree of velocity damping and

The final solution domain for the CFD simulations consists of one complete panel at the center and two half panels on either side. The mesh used for the domain consists of 30 cells in the

The simulation settings used for the final CFD simulations in PHOENICS are summarized in _{PCM}) and the temperature of the outlet air (T_{Air}) are chosen to calculate the errors. The values imply good grid independence given that the GCI is less than 9% for coarse grids and less than 3% for fine grid given that the refinement ratio is in the range 1.5 to 2 [_{air} are much more sensitive than the values of T_{PCM} which is clearly reflected in their respective GCI. For the temporal convergence a constant refinement ratio of 2 is considered and the observed order of convergence is calculated to be 1.23. The GCI here, is also calculated to be within permissible limits. All the values of GCI are calculated with a conservative factor of safety of 3. It is noted that the simulation is sensitive to large fluctuations in temperature and velocity (seen in

Parameters | Inputs |
---|---|

Turbulence model | Chen-Kim k- |

Material PCM | Calcium Chloride Hexahydrate |

Size PCM-filling (m) | 0.57 × 0.12 × 0.05 |

Material and casing size | Glass (w = 0.001 m) |

Number of inner iterations | From 1000 to 200, after 2 timesteps |

Computational time/Simulated time/time steps | 2.5 h/24 h/288 time steps |

Number of cells ( |
30 × 36 × 20 = 21600 |

Grid case | Cell count along X (N_{x}) |
Cell count along Y (N_{y}) |
Cell count along Z (N_{z}) |
Total cells | Grid convergence index (GCI_{Coarse} %) |
Grid convergence index (GCI_{Fine} %) |
||
---|---|---|---|---|---|---|---|---|

T_{PCM} |
T_{Air} |
T_{PCM} |
T_{Air} |
|||||

Case 1 | 24 | 32 | 16 | 12,288 | 0.92 | 4.17 | - | - |

Case 2 | 30 | 36 | 20 | 21,600 | 0.08 | 3.25 | 0.52 | 2.37 |

Case 3 | 38 | 49 | 25 | 46,550 | 0.15 | 6.05 | 0.04 | 1.51 |

Case 4 | 40 | 56 | 30 | 67,200 | - | - | 0.10 | 4.19 |

Time step case | Number of time steps (N_{time}) |
Grid convergence index (GCI_{Coarse} %) |
Grid convergence index (GCI_{Fine} %) |
||
---|---|---|---|---|---|

T_{PCM} |
T_{Air} |
T_{PCM} |
T_{Air} |
||

Case 1 | 144 | 0.15 | 0.60 | - | - |

Case 2 | 288 | 0.35 | 2.89 | 0.06 | 0.25 |

Case 3 | 576 | - | - | 0.15 | 1.22 |

The CFD simulations can be time-consuming for performing dynamic tasks. Therefore a flexible, dynamic model was implemented in MATLAB for integration into optimization algorithms used for developing control strategies. Such modeling essentially reduces the conservation

The control volume, which is here defined as the battery volume, is discretized into three subsections in the flow direction

where

where

Note that in contrast to

The system of equations above, i.e.,

The results produced by the CFD and MATLAB models are compared with those of experiments. The analysis shows that with an assumed constant high enthalpy in the phase change (20°C–23°C) period, a good prediction of the loading/unloading time can be made. In the following paragraphs, the results are discussed.

The accuracy of the CFD model is evaluated by considering the PCM temperatures obtained numerically and experimentally. Such an analysis is based on a time horizon of 24 h, each hour denoting the average of 12 measurements carried out every 5 min. During this time interval, the temperature and velocity of the air entering the PCM battery (as shown in

The PCM is modeled with a linear phase change, as described by the Scheil-Gulliver equation, between 20°C to 23°C. Whereas the phase change in the PCM is non-linear with the maximum phase change occurring at 22°C (see

Although the temperature sensors are isolated with PU foam, some influence of the air temperatures on the panel temperature sensors is expected. This leads to the temperature sensors measuring a marginally different temperature than its actual value.

Finally, the simulation average is calculated, including the inner temperatures of the PCM, while the measurements only consider the surface temperatures of the panel.

Another noticeable deviation of the CFD results from the measurements can be seen in the time interval between 02:00 and 10:00; where the CFD predicts a uniform trend in the PCM temperature, whereas the measurements show a marginal increase during this period. It is to be noted that the measured temperatures rise despite the decreasing inlet air temperature during the period mentioned above (see

The performance of the PCM battery is evaluated by using a steady inflow rate of air at three different uniform inlet velocities: 0, 8, 1.5, and 2.3 m/s, and a constant inlet temperature of 15°C. However, a fixed inlet temperature cannot be realized in the battery. This is why the validation is derived from simulations of a “real-life” working pattern, as in

Air flow condition | ^{2}K) |
(un)load time (s) | Calculated maximum heat transfer (kW) at a |
Calculated average heat transfer (kW) | Latent heating/cooling per panel (Wh) | Total latent heating/cooling (kWh) | |
---|---|---|---|---|---|---|---|

10,800 m^{3}/h, 2.3 m/s, 15°C |
20.4 | 18,000 s (5 h) | 21 | 23 | 10.3 | 155 | 181 |

7,200 m^{3}/h, 1.5 m/s, 15°C |
14.5 | 36,000 s (10 h) | 16 | 16.4 | 5.2 | ||

3,600 m^{3}/h, 0.8 m/s, 15°C |
8.8 | 64,800 s (18 h) | 8 | 9.9 | 2.8 |

Note: The h_{c} value is based on the calculation with

The study shows that initially, the heating and cooling capacity of the PCM battery is higher than the nominal power of the PCM battery (stated as 7.2 kW) and then becomes lower with time, depending on the inlet air velocity and temperature. Another observation from this study is that the buoyancy effect used in the CFD model of the PCM shows a minimal influence on the temperature profile of the PCM panel. This highlights that even at low or zero flow rate, the effects of natural convection in the air channel, on the temperature profile of the PCM panels can be ignored. Finally, it can be seen from

Therefore, in theory, the electrical power of a fan for an air velocity of 1 m/s will be, theoretically, 2^{3 }= 8 times lower than for a velocity of 2 m/s. However, in practice, the difference will be much smaller in practice due to energy losses.

Finally, temperature contours at a central plane of the PCM panel are shown in ^{th} of August using air conditions for a peak summer day in Delft. The yellow zone identifies temperatures above 20°C, so one can see the phase-change front (i.e., the yellow zone) getting cooler with time. The buoyancy effect can also be seen in the contours with a larger quantity of warmer PCM mass at the top of the panel, and a colder mass at the bottom. Consequently, the rate of solidification inside the PCM is higher at the bottom, while the melting rate is higher at the top of the panel.

The numbers of cells x-, y- and z-direction are: 10, 36, 40 (14,400) here. For the final simulations in

For this model, an additional grid-refinement evaluation has also been executed. The maximum numbers of cells in this evaluation are x = 60, y = 56 and z = 30. The effect of increasing the number of cells can be neglected, as also could be concluded by comparing the MATLAB and CFD calculations in

In

Next, we compare the temperatures obtained in MATLAB against the sensor measurements extracted from a different time horizon, which was recorded on December 02, 2021. First,

The strange peak in the CFD simulation values after 200 time-steps has a relation with the warm supplied air at that moment (see

The PCM temperatures calculated in MATLAB are displayed in

After assessing the conformity between numerical and experimental results, we now employ the MATLAB model to evaluate the PCM-battery performance over different conditions while extrapolating scenarios with full charge/discharge periods. Such conditions depend on weather conditions rarely reached in the building in winter, such as, for example, an inlet air temperature >23°C. Therefore, an initial condition where the PCM battery is fully charged is assumed. At a PCM temperature of 23°C, the discharge time depends on the airflow through the PCM and the inlet air temperature, as shown in

The dynamic behavior of a PCM battery, designed for buffering part of the energy demand, is investigated experimentally and numerically. The measurements use 18 sensors distributed over the system and provide the PCM-temperature profile over 5-min intervals. Numerical data are available from CFD (PHOENICS) and MATLAB simulations.

The following conclusions can be drawn:

A PCM-based heat exchanger can be designed for air preheating and cooling in climate systems of large office buildings.

The PCM battery, a composite of air/solid/liquid, can be simulated in CFD and MATLAB while predicting the loading and unloading times with a maximum variance of circa 10%.

The analyses show that the simulation-methods are comparable when evaluating the PCM-battery performance. The charge/discharge times are 2–50 h, depending on the inlet airflow rate (m^{3}/h) and temperature differences.

The use of MATLAB as an addition to CFD reduces the calculation time and makes it possible to find an integrated solution. There is also a MATLAB-model developed for the whole building, so in this way it is possible to find the most optimal phase change temperatures.

The latent storage of the PCM can be simulated as a material with high enthalpy and to a considerable degree of accuracy. This shows that the effect of buoyancy or natural convection is low. This, and the limited effect of the grid size, has also been concluded by other researchers [

The heat transfer coefficient and specific latent heat play a major role when comparing numerical and experimental results. Further works are suggested to improve such correlation, including developing a grey-box model using Sequential Quadratic Programming to calibrate relevant system parameters [

The measured COP for cold and heat loading is circa 27 with the current fans operating at an airflow rate of 6.480 m^{3}/h with an average ΔT of 6.5 K. For heating or cooling during occupancy time, the COP is much higher because the usage of fans is a part of the total system. The resistance of the battery with a maximum below 25 Pa is only a small percentage of the total system resistance. The fans themselves have a maximum flow already at a resistance of circa 2,500 Pa.

The following areas could become the subject of future research:

A low-pressure fan should be applied and the valve openings, and maybe the battery as well, should be designed in such a way that ventilation (and Aeolian) noise stays within an acceptable level at a much lower electrical energy consumption.

With a low-pressure ventilation system, the COP can become higher, above 100, considering the high current resistance of the current ventilation system.

Since the PCM battery is a low-pressure solution (<25 Pa), it can even be integrated into natural ventilation systems.

In the present building, a combination of PCMs with a heat pump connected to pipes in the ground (foundation pillars) will be applied soon. This will reduce the relevance of PCMs for cooling and heating, and this will also increase the general COP for heating from 2 to circa 5. Summer cooling by PCMs will reduce any unwanted warming up of the cold storage in the ground.

A main point of attention is how to load the PCM with sufficient warm air. In the current design, the availability of sufficient (passive solar) heat is very limited, especially in mid-winter.

A future option could be to load the PCM panels with heat from solar panels on the roof or to load the panels via the heat pump when there is a surplus of PV energy during the day.

More detailed simulations considering the material’s hysteresis (

The intention is to apply PCM batteries like this in other buildings. However, more insight into the added value of PCM batteries, often combined with heat pumps, is still necessary to be able to draw more detailed conclusions about the advantages and most optimal integration of such a system.

In the meantime more PCMs (1/3 of the total capacity) in the free part of the PCM battery (

The PCM-battery and installations are designed and built by van Dorp B.V. (

_{2}⋅6H

_{2}O as a phase change material for thermo-regulation and enhancing photovoltaic panels’ conversion efficiency: Experimental study and TRNSYS validation