Development and Validation of Air-to-Water Heat Pump Model for Greenhouse Heating

: This study proposes a building energy simulation (BES) model of an air-to-water heat pump (AWHP) system integrated with a multi-span greenhouse using the TRNSYS-18 program. The proposed BES model was validated using an experimental AWHP and a multi-span greenhouse installed in Kyungpook National University, Daegu, South Korea (latitude 35.53 ◦ N, longitude 128.36 ◦ E, elevation 48 m). Three AWHPs and a water storage tank were used to fulﬁll the heat energy requirement of the three-span greenhouse with 391.6 m 2 of ﬂoor area. The model was validated by comparing the following experimental and simulated results, namely, the internal greenhouse temperature, the heating load of the greenhouse, heat supply from the water storage tank to the greenhouse, heat pumps’ output water temperature, power used by the heat pumps, coefﬁcient of performance (COP) of the heat pump, and water storage tank temperature. The BES model’s performance was evaluated by calculating the root mean square error (RMSE) and the Nash–Sutcliffe efﬁciency (NSE) coefﬁcient of validation results. The overall results correlated well with the experimental and simulated results and encouraged adopting the BES model. The average calculated COP of the AWHP was 2.2 when the outside temperature was as low as − 13 ◦ C. The proposed model was designed simply, and detailed information of each step is provided to make it easy to use for engineers, researchers, and consultants.


Introduction
Recently, greenhouse farming has increased rapidly in many countries, including South Korea. The primary objective of greenhouse farming is to obtain a year-round and high crop production in areas with severe climatic conditions unfavorable for crop production, where farming is possible only by maintaining the optimum greenhouse microclimate throughout the production period. Different heating and cooling systems are used to provide a favorable environment to crops inside the greenhouse. Therefore, fossil fuels are being used for heating and cooling, which not only increases production costs [1] but also causes CO 2 emissions and environmental pollution [2]. In South Korea, greenhouse heating costs have increased to 45% of the total production cost [3] because of continuous oil price increases. To reduce CO 2 emissions and cope with the continuously increasing oil price, the South Korean government has promoted the use of new and renewable energy (NRE) sources for different purposes, including in the agriculture sector. The NRE law was executed in 2004, enforcing the installation of the NRE systems [4].
Heat pumps are widely used in commercial as well as residential buildings [5], airsource heat pumps (ASHPs) are the most widespread heating source in commercial and covering materials, screen materials, and environmental control inside the greenhouse under local weather conditions. The TRNSYS program is an extremely flexible graphically based environment used to simulate the behavior of transient systems. It focuses on assessing the performance of thermal energy system simulations, including building energy simulation (BES), solar thermal processes, solar applications, geothermal energy, heat pump systems, groundcoupled heat transfer, airflow modeling, wind and photovoltaic (PV) systems, power plants, and energy system calibration. The University of Wisconsin's Solar Energy Lab developed TRNSYS, which has been commercially available since 1975 for simulating thermal systems. However, it has since undergone continuous development to become a hybrid simulator [27]. It is a component-based program simulating complex energy flows in buildings. TRNSYS can be easily connected to other programs for pre-and postprocessing the model. Over the last two decades, TRNSYS has been widely used in industry and research as a reliable tool for BES [28]. Moreover, for the modeling and simulation of agricultural greenhouses, the TRNSYS program shows high flexibility, as many greenhouse models have been developed and validated [1,[29][30][31][32][33][34].
This study proposes a BES model of an AWHP system integrated with a multi-span greenhouse using the TRNSYS-18 program. The proposed model results for the internal greenhouse temperature, heat energy demand, storage tank temperature, and temperature supply and return from the heat pump and greenhouse, were validated using experimental data on the heating mode of the AWHP. Furthermore, the feasibility of the AWHP to fulfill the heat energy demand of the greenhouse was investigated. The proposed model considered all time-varying control factors of the greenhouse, including thermal screen control, ventilation control, internal temperature control, and temperature control on/off heat pump system. All physical factors of the greenhouse, including the design, covering, and screen materials and the specification of the heat pumps, storage tank, and water circulation pumps, were also considered. The study considered dynamic simulation of specific AWHP and multi-span greenhouse with fully controlled conditions under local weather conditions of Deagu, South Korea. The proposed BES model of the AWHP integrated with multi-span greenhouse was designed simply, and detailed information on each step is provided to make it easy to use for engineers, researchers, and consultants. Researcher can use this model for the dynamic thermal simulations of their specific greenhouse design and control requirements according to the local weather conditions. Moreover, AWHP analysis can help to find the feasible solution to increase the COP.

Experimental Greenhouse
The experimental greenhouse was a three-span rectangular, north-south (N-S) oriented, Venlo-roofed greenhouse, in which, the roof was covered with horticulture glass (HG, 4 mm) and the side walls were covered with polycarbonate (PC, 16 mm Figure 2 shows the locations of the sensors and dimensions of the greenhouse. Weather data were recorded inside and outside the greenhouse from 1 January 2021 to 31 March 2021, during the heating period. The weather data recorded inside the greenhouse were air temperature, relative humidity, and solar radiation to validate the BES model. The weather data recorded outside the greenhouse were air temperature, relative humidity, solar radiation, and wind speed and direction. The ambient pressure data were obtained from the Korean Meteorological Administration (KMA). The outside weather data were obtained to use as input for the  Table 1 shows all recorded weather variables and their characteristics. Figure 3 shows the mean outside air temperature and solar radiation.

AWHP
Three AWHP units with a water storage tank were analyzed for their heating performance to fulfill the heat energy demand of the three-span greenhouse (detailed above). Figure 4a,b shows the experimental AWHP and the water storage tank, respectively, installed in Kyungpook National University, Daegu, South Korea. Figure 5 shows a schematic of the entire process of the AWHP in heating mode and the location of the sensors (Table 1 details the characteristics). The water storage tank stores hot water from the AWHP and supplies it to the greenhouse when heating had to reach the setpoint internal air temperature. Inside the greenhouse, heating pipes and two fan coil systems were installed to exchange heat in the greenhouse (GH3), and two fan coil systems each were used in the other two compartments (GH1 and GH2). Table 2 details the specifications of the AWHP system. We monitored the water flow rate and temperature and various locations mentioned in Figure 5 to calculate the heat energy supply from the AWHP to the water storage tank, and from the water storage tank to the greenhouse using Equation (1) where Q is the amount of heat transfer or heating capacity of the AWHP (kJ),ṁ is the mass flow rate (kg·s −1 ), cp is the specific heat capacity of water (kJ·kg −1 · • C −1 ), and ∆T is the temperature change (°C). Further, the COP of the heat pump was calculated using Equation (2) CP = Q P HP (2) where P HP is the power usage of AWHP in kJ.

BES Modeling and Simulation
Designing the proposed BES model was divided into two steps. First, the greenhouse model was developed, and secondly, the heat pump system model was designed. The BES model of the AWHP system integrated, with the multi-span greenhouse with all physical and technical parameters the same as the experimental setup, was designed using the TRNSYS-18 program. Figure 6 shows the simulation studio (the main interface of the TRNSYS program) connecting all model components. Table 3 shows all components (types) and their complete descriptions, as used in the simulation studio of the TRNSYS program, to design the proposed model. First, the 3D model of the multi-span experimental greenhouse was designed in Transys3d, an add-on of Google SketchUp TM , and imported as a .idf file (readable by TRNSYS) into the TRNBuild (a building interface of the TRNSYS program, TYPE 56). The greenhouse covering and screen material properties were introduced into the Lawrence Berkeley National Laboratory Windows 7.7 program, creating the DOE-2 file of the materials readable by TRNBuild. Table 4 shows the covering material and screens material's properties, while in Table 5, steel (greenhouse structure) and ground properties used in the simulation are shown. TRNBuild managed the thermal model of the building to account for natural ventilation into the greenhouse thermal model TRNSFLOW (a ventilation module of the TRNSYS program), coupling the airflow network with the thermal model to simulate the effect of natural ventilation on the greenhouse's thermal environment. Furthermore, in the simulation studio, the weather data and weather data processors were linked to the TRNBuild to simulate the effect of an ambient environment. Table 6 shows the details of the referenced greenhouse, including physical and control strategies used in the BES model simulation.
After modeling the greenhouse, the AWHP system model was prepared. Three Type 941 from the TRNSYS Tess library were used to model the AWHP. This type uses performance data files for heating and cooling provided by the manufacturer and ambient air temperature and relative humidity as input. The heating performance data file used in the simulations, and datasheet of the heat pump provided by the manufacturer can be found in Appendix A. Table 2 shows the rated heating capacity and power consumption values of the heat pump. The heat pumps are controlled with an on/off signal, as they would be in a real system. In Type 649, a water-mixing valve is used to collect hot water from all three heat pumps and deliver it to the water storage tank (Type 4). Type 114 (circulation pump) delivers cold water to the heat pump at a constant speed. From the storage tank to the fan coil unit (Type 928), hot water is provided using Type 3 (variable speed circulation pump) with a Type 22 proportional integral derivative controller to control the mass flow rate with feedback control. The fan coil unit provides hot air to the greenhouse. Moreover, Type 709 (pipe) is used inside the greenhouse to provide heat. Hot water runs into the pipes, and heat loss from the pipes provides heat inside the greenhouse. Figure 7 shows the flow chart diagram of the proposed model, explaining the detail information of the pre-processing, simulation and output. Figure 6. TRNSYS simulation studio showing proposed model's components for air-to-water heat pump system integrated with multi-span greenhouse. Yellow box: heat pump, green box: storage tank, black box: fan coil units, purple box: multi-span experimental greenhouse modeling, TRNBuild, red box: weather data reading and processing.

Statistical Analysis of BES Model
Statistical analyses were performed to predict the BES model's performance using the Nash-Sutcliffe efficiency (NSE) coefficient and compare the experimentally measured data with the BES model's output. This coefficient quantitatively describes the accuracy of the model results, it indicates how well the plot of observed versus simulated data fits the 1:1 ratio. Its value ranges from −∞ to 1, and values closer to 1 indicate better predictive power of the model. The NSE is mathematically expressed using Equation (3). The performance ratings for NSE values are as follows: NSE > 0.9 = very good, 0.8-0.9 = good, 0.65-0.80 = acceptable, and <0.6 = unsatisfactory [35]. Furthermore, Equation (4) for the RMSE was used to quantify the error for units of the variables. The equations are mathematically defined as follows: where T

Results and Discussion
To validate the proposed AWHP system integrated with the multi-span greenhouse BES model, the computed internal air temperature of the greenhouse, heating load, heat pump output temperature, and storage tank temperature were compared with those obtained experimentally using the same physical and operating conditions. Validation was conducted during the winter from 1 January to 31 March 2021. The normal operation of the heat pump in heating mode was repeated daily; therefore, only some days' analysis results are presented here. Figure 8 shows the greenhouse's internal experimental and simulated temperature along with the ambient temperature from 1-21 January 2021. The validation results are for all three compartments of the greenhouse, where the heating setpoint was different for each compartment. The heating setpoints were 16, 18, and 17 • C for compartments 1, 2, and 3, respectively, whereas the ventilation setpoint was 26 • C for all compartments. The simulated greenhouse internal temperature results correlated well with those of the experimentally measured temperatures. The RMSE values for the validation results were 1.9, 1.8, and 2.0 • C, indicating the maximum temperature difference between the predicted and experimental results. The NSE values were 0.71, 0.70, and 0.65, respectively, which are acceptable. Compartment 3 shows a slightly lower value because, for some days, the experimental temperature was not well controlled to 17 • C. The overall performance analysis results show that the proposed BES model is accurate enough to predict the internal greenhouse temperature.   Figure 3 shows the ambient temperature for the entire analysis period, showing 8 January 2021, with the lowest temperature. The middle compartment showed the least heating load, even when the heating setpoint was higher than that of the other two compartments because of the reduction in heat loss. Both sidewalls were adjacent to the other compartments, and less area was exposed to the ambient environment than in the other compartments. Furthermore, compartment 3 had less heating load than that of compartment 1 because its two sidewalls were exposed to the sun and received more solar heat during the day than compartment 1, of which only one sidewall was exposed to the sun. 30,000, 25,300, and 26,400 Kcal·h −1 on 08 January 2021, when the outside temperature was the lowest for the season (−13 °C). Figure 3 shows the ambient temperature for the entire analysis period, showing 08 January 2021, with the lowest temperature. The middle compartment showed the least heating load, even when the heating setpoint was higher than that of the other two compartments because of the reduction in heat loss. Both sidewalls were adjacent to the other compartments, and less area was exposed to the ambient environment than in the other compartments. Furthermore, compartment 3 had less heating load than that of compartment 1 because its two sidewalls were exposed to the sun and received more solar heat during the day than compartment 1, of which only one sidewall was exposed to the sun. Furthermore, the model's performance for the results shown in Figure 9 was analyzed using NSE, RMSE, and a scatter plot against a 1:1 line to visually inspect the results. Figure 10a-c shows the scatter plot for the experimental vs. simulated heating loads of the greenhouse against a 1:1 line and the NSE value. The higher performance can Furthermore, the model's performance for the results shown in Figure 9 was analyzed using NSE, RMSE, and a scatter plot against a 1:1 line to visually inspect the results. Figure 10a-c shows the scatter plot for the experimental vs. simulated heating loads of the greenhouse against a 1:1 line and the NSE value. The higher performance can result in a scatter plot closer to the 1:1 line. The NSE values of 0.73, 0.81, and 0.67 for greenhouse compartments 1, 2, and 3, respectively, were acceptable. Moreover, to quantify the error for heating load units, the RMSE values of greenhouse compartments 1, 2, and 3 were 5140, 3674, and 5141 Kcal·h −1 . The maximum difference between the experimental and simulated results occurred on 4 January 2021, because the experimentally calculated heat load was not in a steady state, resulting in a sudden rise and fall of hot water supply to the greenhouse. Figure 9 shows that the experimentally calculated heat supply was higher on 4 January 2021, even though the outside temperature was higher than on 8 January 2021. However, the simulated results were in a steady state and showed an acceptable heating load trend with the outside temperature.   Figure 11 shows the results for the experimentally calculated vs. simulated heating supplied from the water storage tank to the greenhouse along with ambient air temperature from 1 January to 15 March 2021. The experimental values were calculated per unit area of the greenhouse using Equation (1) (Section 2.2) using the measured water flow rate and temperature difference between the supply and return water temperature. The results indicate that the maximum heating supplied was on 8 January 2021, as the ambient temperature was at its lowest value of −13 • C. The results shown in Table 7 are the heating demand of the experimental greenhouse estimated by TRNSYS without using a particular heating system, and they confirm the maximum supplied load under the same weather conditions. The experimental results show high energy supplied to the greenhouse from 1-12 February 2021, even when the ambient temperature was high, which only occurred for one hour because of the unsteady water flow, whereas simulated results showed linear interpolated results. The overall trend and results correlated well with the validation results. Figure 12 shows that the simulated value is close enough to the experimental values, as the scattered values follow the 1:1 line. Moreover, the statistical analysis of the results shows an acceptable NSE value of 0.70 and RMSE value of 20 Kcal·h −1 ·m −2 . Figure 11. Experimental vs. simulated heating supply from water storage tank to greenhouse.  After validating the greenhouse's internal air temperature and heating load, the experiments were further extended to validate the AWHP results. The output (hot water temperature) of the experimental heat pumps was compared with the simulated hot water temperature. Because the heat pump's function was the same throughout the season, two days' results were compared. The simulated heat pump used ambient air temperature and relative humidity as an input, and a 10 min time step, the same as the time of the data logger, was used for the simulations. The results correlated well with the experimental and simulated results (Figure 13), with a small RMSE of 0.4 • C. Three heat pumps were used in this study to heat the water and store it in the water storage tank. The storage tank temperature controlled the heat pumps' ON/OFF setting. Figure 14a-c shows the seven-day data of the experimental and simulated heat pumps, 1, 2, and 3, respectively, and the electrical consumption shows the switching ON/OFF of the heat pumps according to the need. Furthermore, the validation of both experimental and simulated results correlated well.  Figure 15 shows the COP of the heat pumps with greenhouse internal and amb temperatures. The heat pumps' heating performance was evaluated during the maxim heating requirement period. The results shown are from 07-08 January 2021, when ambient temperature was at its lowest (−13 °C). The results show that when the amb temperature starts reducing to −13 °C from 1 to 5 a.m., the average COP of the heat pu reduced from 2.0 to 1.7 in heating mode. The calculated average COP value of 2.2 sho  Figure 15 shows the COP of the heat pumps with greenhouse internal and ambient temperatures. The heat pumps' heating performance was evaluated during the maximum heating requirement period. The results shown are from 7-8 January 2021, when the ambient temperature was at its lowest (−13 • C). The results show that when the ambient temperature starts reducing to −13 • C from 1 to 5 a.m., the average COP of the heat pump reduced from 2.0 to 1.7 in heating mode. The calculated average COP value of 2.2 shows the same value as the manufacturer's recommended COP for this type of AWHP. Like other analyses, the normal operation of the water storage tank's charging and discharging was the same; therefore, only some days' results were shown for validation analysis (1-21 January 2021). Figure 16 shows the experimental and simulated storage tank temperatures. The validation results correlated well with an RMSE of 0.5 • C. The validation results show that the storage tank model is fair enough to be adapted. Only one water temperature sensor was installed on the top of the tank; therefore, the average water temperature results are not shown in the study.

Conclusions
This study validates the proposed BES model of an AWHP system integrated with a multi-span greenhouse using the TRNSYS-18 program. The model was validated by comparing experimental and simulated results, namely, internal air temperature of greenhouse, the heating load of the greenhouse, heat supply from the water storage tank to the greenhouse, heat pumps' output water temperature, power used by heat pumps, COP of heat pumps, and water storage tank temperature. The BES model's performance was evaluated by calculating the RMSE and NSE coefficient of validation results. The specific statistical analyses of all validation results are as follows: • The RMSE values for the internal greenhouse air temperatures were 1. The maximum heating energy demand for the studied greenhouse was found to be 250 kcal·h −1 ·m −2 .
The overall results correlate well with experimental and simulated results and encourage adopting the BES model. The model implemented in the TRNSYS-18 program can run year-round simulations. The proposed BES model of an AWHP integrated with a multispan greenhouse was designed simply, and detailed information on each step is provided to make it easy to use for engineers, researchers, and consultants. The presented model is developed for being used as decision-making tool for the dynamic thermal simulations of specific greenhouse design and control requirements according to the local weather conditions. All the validation results put in evidence that the proposed model is capable of evaluating such a system. Moreover, AWHP analysis can help to find a feasible solution to increase the COP. Future work will consider validating the model in the cooling mode. The proposed model can be used to optimize the control strategies and improvements to systems.

Conflicts of Interest:
There is no conflict of interest regarding the publication of this research.

Appendix A
The AWHP model component in the TRNSYS requires a heating performance data file as an input. Table A1 shows the heating performance data used in the study, as provided by the manufacturer. Table A2 presents the data sheet provided by the manufacturer and the testing conditions of the heat pump. Capacity calculation standard for heating: inlet/outlet water temperature: 40 • C/45 • C, ambient temperature: 7 • C. Nominal data are shown with a gray background.