Experimental Analysis of Bi-Directional Heat Trading Operation Integrated with Heat Prosumers in Thermal Networks

: District cooling and heating methods that can utilize highly efﬁcient heat pumps and various unused new and renewable types of energy are required to achieve low carbon emissions and zero energy usage in buildings and community units. The technical requirements for the implementation of decentralized thermal networks and heat trading are increasing, both for thermal networks in new buildings and for those remodeled based on existing centralized thermal networks. In this study, a conventional centralized thermal network was implemented as a decentralized thermal network and the possibility of heat prosumers feeding thermal networks was demonstrated experimentally. A real-scale plant was constructed by employing unused thermal energy facilities as prosumers in a school and childcare center based on the existing small-scale block heating and cooling thermal network. The decentralized thermal network and heat prosumer concepts were proven through operation experiments performed on the constructed system in summer and winter. An economic beneﬁt can be achieved by increasing the peak power cost. The experimental results also showed that the proposed bi-directional heat trading reduced carbon emissions by 12.7% compared with conventional centralized thermal systems.


Introduction
With the increasing demand for the use of environmentally friendly energy and the expansion of distributed energy systems, interest in thermal energy trading and sharing is increasing in fourth-generation district heating and fifth-generation district cooling and heating in the thermal energy sector, as well as in the power sector [1][2][3]. In particular, thermal energy trade between buildings can be performed not only to utilize the surplus thermal energy existing in a building, but also to achieve economic efficiency or environmental friendliness of the heat source equipment. In other words, heat source facilities installed according to the maximum load of each building are selectively operated under various conditions, and the produced thermal energy is shared among multiple buildings. Such thermal energy trading or sharing can increase the efficiency of distributed energy systems and the utilization of renewable thermal energy.
Furthermore, interest in carbon neutrality and energy supply to buildings and communities based on renewable energy, such as RE100, has recently increased [4]. Among the many types of renewable energy, low-cost solar energy, whose equipment is easy to install, is being actively applied to buildings and in communities [5]. With the increasing level of solar power generation, however, linking surplus power to the grid is emerging as a problem, and the burden of linking solar power to the grid due to intermittent power generation is increasing. Consequently, interest in self-consumption in buildings and communities is growing [6] and research on the conversion of surplus power into heat or the installation of solar collectors instead of solar power generation facilities and the utilization of the generated power in buildings and communities is drawing attention [7].
The eco-friendly energy town of Jincheon was constructed as a water quality restoration center near public buildings located in Jincheon-gun in Chungbuk Innovation City. The objective is to achieve 100% self-sufficiency in terms of the annual power and thermal energy demands by installing renewable and thermal energy systems as well as supplying eco-friendly energy to nearby public buildings. The energy produced is supplied to a library, daycare center, healthcare center, youth center, and high school, as well as five other buildings, including an integrated management center that is used as a machine and public relations room. More detailed physical information of those buildings can be seen in the previous research [29]. A childcare center was recently added to the town, and a total of seven buildings were integrated into the thermal network.
Jincheon was constructed by the end of 2016 to supply thermal energy for heating and domestic hot water throughout the year based on the 1600 m 2 solar thermal system and 4000 m 3 seasonal thermal energy storage ( Figure 1). More detailed information on the solar thermal system and seasonal thermal energy storage can be seen in previous research [29]. Three 175 kW heat pumps, including geothermal and sewage source heat pumps, were also installed as auxiliary heat source systems. A heat pump for low-temperature surplus heat was installed in the seasonal storage tank as an evaporation heat source. The 200 m 3 capacity of nighttime (i.e., off-peak time) thermal energy storage (NTES) is located in the management center and connected with heat pumps as a buffer tank. Furthermore, 850 kW photovoltaic systems are installed in public buildings, parking lot roofs, and unused sites. The generated power is sold, and the grid power is consumed to realize a net-zero-energy town [29].
The 350-kW-capacity geothermal and sewage source heat pumps used as auxiliary heat source systems can produce cooling water in summer to public buildings, except for a school with large air-conditioning loads. In the school, a separate, approximately 1000 kW gas combustion absorption heat pump (AHP) has been installed and used as a cooling system. Therefore, the mechanical room in the school has a heat exchanger capable of receiving the supplied heat and channels it to supply the cooling water produced by the AHP to the interior of the school.

Thermal Network Retrofitting for Heat Trading
The management center normally produces cooling and heating water utilizing various heat sources, such as solar thermal systems integrated with seasonal thermal energy Energies 2021, 14, 5881 4 of 18 storage (STES) and multi-source heat pumps. The high school consumes the heating energy supplied by the management center or produced by the AHP installed in the school. The childcare center can consume heating and cooling energy produced by the geothermal source heat pump by itself or by the other two heat prosumer buildings (the high school and management center). The high school and childcare center are equipped with their own heat production facilities, and the other buildings use the heating and cooling energy supplied by the management center (Figure 2a). The childcare center has a ground-source heat pump installed with 30 RT capacity. Due to the limited capacity of the heat pump in the childcare center, the thermal energy produced by the heat pump is served to the public health center and daycare center. Figure 2 shows the operating method of the three heat prosumer buildings for trading cooling and heating energy. receiving the supplied heat and channels it to supply the cooling water produced by the AHP to the interior of the school.  [29]; (c) hybrid renewable energy system in town [29].

Thermal Network Retrofitting for Heat Trading
The management center normally produces cooling and heating water utilizing various heat sources, such as solar thermal systems integrated with seasonal thermal energy storage (STES) and multi-source heat pumps. The high school consumes the heating energy supplied by the management center or produced by the AHP installed in the school. The childcare center can consume heating and cooling energy produced by the geothermal source heat pump by itself or by the other two heat prosumer buildings (the high school and management center). The high school and childcare center are equipped with their own heat production facilities, and the other buildings use the heating and cooling energy supplied by the management center ( Figure 2a). The childcare center has a groundsource heat pump installed with 30 RT capacity. Due to the limited capacity of the heat pump in the childcare center, the thermal energy produced by the heat pump is served to the public health center and daycare center. Figure 2 shows the operating method of the three heat prosumer buildings for trading cooling and heating energy.
During the demonstrated operations, it was determined whether the intended heat supply was properly performed while controlling each operation method according to an arbitrary schedule. Figure 2b shows the conventional centralized thermal network before retrofitting. Figure 2c depicts a typical case of the valve and pump control methods based on the prosumer operation method. As illustrated in Figure 2c, a bi-directional thermal network is enabled by installing valves and pumps. Valves VH1 and VH2 and the connecting pipes have already been installed to convert the high school into a prosumer. To convert the daycare center into a prosumer, VC1 and VC2 were installed in the management center, and the valves and connected pipes such as VC3 and VC4 were additionally installed in the daycare center. Figure 2d represents the pipeline of the high school for prosumer operation, and a control method when the heat energy produced by the AHP in the high school is shared with the buildings. Figure 2e depicts the pipeline of the childcare center for the prosumer. The childcare center can deliver or receive heating and cooling energy from the ground-source heat pump itself, the management center, or the high school.  [29]; (c) hybrid renewable energy system in town [29].

Thermal Network Retrofitting for Heat Trading
The management center normally produces cooling and heating water utilizing various heat sources, such as solar thermal systems integrated with seasonal thermal energy storage (STES) and multi-source heat pumps. The high school consumes the heating energy supplied by the management center or produced by the AHP installed in the school. The childcare center can consume heating and cooling energy produced by the geothermal source heat pump by itself or by the other two heat prosumer buildings (the high school and management center). The high school and childcare center are equipped with their own heat production facilities, and the other buildings use the heating and cooling energy supplied by the management center ( Figure 2a). The childcare center has a groundsource heat pump installed with 30 RT capacity. Due to the limited capacity of the heat pump in the childcare center, the thermal energy produced by the heat pump is served to the public health center and daycare center. Figure 2 shows the operating method of the three heat prosumer buildings for trading cooling and heating energy.
(a)  During the demonstrated operations, it was determined whether the intended heat supply was properly performed while controlling each operation method according to an arbitrary schedule. Figure 2b shows the conventional centralized thermal network before retrofitting. Figure 2c depicts a typical case of the valve and pump control methods based on the prosumer operation method. As illustrated in Figure 2c, a bi-directional thermal network is enabled by installing valves and pumps. Valves VH1 and VH2 and the connecting pipes have already been installed to convert the high school into a prosumer. To convert the daycare center into a prosumer, VC1 and VC2 were installed in the management center, and the valves and connected pipes such as VC3 and VC4 were additionally installed in the daycare center. Figure 2d represents the pipeline of the high school for prosumer operation, and a control method when the heat energy produced by the AHP As depicted in Figure 2d, additional experimental facilities were constructed in the school building, and Table 1 summarizes the valve controls according to various modes. Regarding the 3-way valve operation, open and close operation of middle port pathway is presented. Based on this operation scheme, three main operation modes can be identified: Consumer operation (heating mode), during which heat is received from the main mechanical room during the heating season; consumer operation (cooling mode), which takes charge of the self-cooling and heating loads through AHP operation; and prosumer operation (prosumer mode), during which heat is supplied to the block heating network while taking charge of the self-cooling and heating loads through AHP operation. As shown in Figure 2d, the centralized thermal network method typically operates as follows. Heat is supplied by the supply header in the main mechanical room to public buildings, including the school, and is recovered to the return header. Then, the supply water temperature setpoint is maintained through a heat exchanger. To implement prosumers while maintaining the conventional thermal network pipeline, heat at a higher temperature than the supply water temperature setpoint must be produced in the school and supplied to the return pipeline again. The heat that returns to the main mechanical room must be mixed with the return water from the other buildings to maintain the supply water temperature. For this purpose, the prosumer building must produce a temperature higher than the supply water temperature setpoint, and the return water temperature of the school building must remain at the supply water temperature setpoint after being mixed with the return water from the public building. However, this method is difficult to implement. Therefore, we redesigned the thermal network pipeline in this study, as illustrated in Figure 2d, so that when the school is operated as a prosumer, the valves (VH1 and VH2) are controlled such that the return water from the public building enters the heat exchanger supply side of the school, and the temperature of the hot or cooling water leaving the heat exchanger return side through AHP control will maintain the supply water temperature setpoint of the public building. Further, the heat passing through the heat exchanger is designed to return to the main mechanical room and to be supplied through the header to the public building. Table 1 lists the detailed operation characteristics of each valve.
As shown in Figure 2e, valves VC1-VC4 are adjusted to supply heat from the childcare center to the thermal network through prosumer operation, and an independent decentralized thermal network is configured by an additional pump operation. In addition, the childcare center mechanical room is designed to support consumer mode operation, which uses heat obtained from the centralized thermal network, by adjusting valves Vy1-Vy5 and prosumer mode operation, which can supply heat to the public health center and daycare center, by using part of the centralized thermal network. Table 2 summarizes the valve operation characteristics for each mode. Regarding the 3-way valve operation, open and close operation of middle port pathway is presented.

Overview of Experiments and Monitoring Systems
In this study, we analyzed the cooling energy supply operational data between 6 July and 31 August 2020. The operation experiment results of the centralized thermal network method were compared with those of the bi-directional operation experiment based on 20 days of data. The operation experiment of the centralized thermal network method was performed from 6 July to 31 July, and the bi-directional operation experiment was performed from 3 August to 31 August. During operation, optimization such as the reduction of the operation cost and peak load compared to the centralized thermal network method was not performed. Hence, the operation cost and greenhouse gas emission reduction effects that can be obtained when operating a distributed system were compared through simulation.
As shown in Table 3, various measurement sensors, such as temperature, flow, solar radiation, and electric power sensors, were installed to measure and monitor the system performance. In the experimental system, which was within the area range of this study, five pressure gauges, five flow meters, seventeen temperature sensors (RTDs), and six watt-hour meters were installed. The pressure gauges and flow meters measured the pressure and flow of the operating fluid flowing through the main pipe, respectively. The watt-hour meters measured the power consumption of the circulation and heat pumps. The measured data were recorded in 30 s intervals.

Energy Balance Equations
Using the data measured by the temperature sensors and flowmeters installed in the heat-producing facilities of the centralized thermal network and each building, the thermal energy balance of this town was estimated. The cooling load ( . Q C,total ) of all the buildings can be estimated by Equation (1) (5) and (6), respectively. The cooling supply from the ground-source heat pump in the childcare center ( . Q GSHP,C,child ) can be calculated using Equation (7). . . . . . . where: T cws is the chilled water supply temperature from the centralized mechanical room [ • C]. T cwr is the chilled water return temperature from the centralized mechanical room . V L is the volumetric water flow rate [m 3 /s].

Simulation Overview
During the demonstration experiment period of this study, city gas work had not been completed in the school building. Thus, LPG gas was temporarily used for the AHP operation and cooking. However, because city gas will be installed in the future, we calculated the demonstration experiment results and operating costs in the case of using LPG gas and analyzed the reductions in operating cost and carbon emissions in bi-directional heat trading compared to the conventional centralized thermal network method in the case in which city gas is installed and supplied to the school in the future. The analysis was conducted under the following assumptions: the average heating value (HV) of the LPG is 94.2 MJ/m 3 ; the density is 1.868 kg/m 3 , and the price of LPG is 1.47 USD/kg (0.029 USD/MJ); the HV of LNG is 43.54 MJ/m 3 , and the price is 0.036 USD/kWh (0.01 USD/MJ) and 0.056 USD/kWh (0.015 USD/MJ) during the cooling and heating seasons, respectively, for building cooling and heating. These values were obtained from the Korea City Gas Association. In this study, the exchange rate is assumed at 1100 USD/Won. The initial cost of the additional underground pipeline, pipeline, and valves in the mechanical room is not considered in this paper due to the fact that this cost is miscellaneous, and this paper does not consider the payback period of the proposed system.
We also considered the environmental impact of an energy-output-based emission factor approach for the proposed system. The general CO 2 emission was used to represent the total greenhouse gas emission to estimate the environmental impact. As shown in Equation (8), the CO 2 emission savings (∆CO 2 ) can be calculated by comparing the proposed (m PS CO 2 ) and conventional systems (m CS CO 2 ) using global CO 2 equivalent emissions (m CO 2 ). The values of m PS CO 2 and m CS CO 2 in Equation (8) were estimated using Equations (9) and (10), respectively: where: γ is the CO 2 equivalent emission factor for electricity consumption [gCO 2 /kWh el ]. δ is the CO 2 equivalent emission factor for gas consumption [gCO 2 /Nm 3 ].  Table 4. The electricity prices range from 0.052 USD/kWh to 0.12 USD/kWh during the test cooling season.  Figure 3 presents the outdoor air temperature (T oa ) and thermal load (i.e., cooling load) during the test period. The measured daily average outdoor air temperature was 22.0-33.3 • C. In this test site, the peak cooling season is July and August. As Meteonorm 7.3 software (Fabrikstrasse, Switzerland) is able to generate TMY2 climate data using the past more than 10 years metrological data, the daily average outdoor air temperature was 21.3-30.3 • C in the peak cooling season. It was proved that the test period of this paper can represent and cover the cooling season. The total cooling load of the town ranges from 1754.9 kWh to 4580.0 kWh. During the experimental period, school buildings account for 20.5-70.6% of the total cooling load. From 6 to 10 July, the school accounts for approximately 70% of the total cooling load, but this percentage decreases to 20.6-56.6% from 13 July to 31 August, which is a vacation period.

Outdoor Air Conditions and Thermal Loads
As mentioned above, the heat was supplied by the centralized thermal network method from 6 to 31 July, and a decentralized bi-directional thermal network was operated from 3 to 31 August. In Figure 3, the gray and green bars represent the heat supplied to the public building by the bi-directional thermal network from the school and childcare centers, respectively. During the experimental period, the cooling energy supplied to the school and childcare center through the heat prosumers accounts for 3.7-2.0% and 4.8-18.0% of the total cooling load of the town, respectively.
As mentioned above, the heat was supplied by the centralized thermal network method from 6 to 31 July, and a decentralized bi-directional thermal network was operated from 3 to 31 August. In Figure 3, the gray and green bars represent the heat supplied to the public building by the bi-directional thermal network from the school and childcare centers, respectively. During the experimental period, the cooling energy supplied to the school and childcare center through the heat prosumers accounts for 3.7-2.0% and 4.8-18.0% of the total cooling load of the town, respectively.  Figure 4 presents an analysis of the operation profile for the representative period from 8 to 10 July during the above-mentioned period of centralized thermal network operation. Figure 4a,b shows the inner temperature of the NTES according to thermal energy charging and discharging during the operation of the centralized thermal network. The blue area represents the heat charged into the NTES through heat pump operation (Q , , ), and the orange area represents the discharging of the cooling to the thermal network from the NTES (Q , ). As mentioned above, heat is charged into the NTES through heat pump operation during the off-peak period, and when the temperature rises in the heat storage tank, cooling is charged through heat pump operation even during the peak period. The experimental results demonstrate that most heat charging occurs during the off-peak period. When the set cooling temperature cannot be maintained in the thermal network owing to a temperature increase at the bottom of the heat storage tank to more than 9 °C, heat is charged into the heat storage tank even during the peak period. Figure 4c shows the distributions of the cooling water supply temperature ( ( ) ) and cooling water return temperature ( ( ) ) inside the thermal network. The results confirm that the cooling water is supplied to the thermal network at approximately 11 °C and returned at 13 °C. The indoor temperatures of the buildings in the thermal network are maintained at appropriate levels. Supply temperature fluctuations occurred because the    Figure 4 presents an analysis of the operation profile for the representative period from 8 to 10 July during the above-mentioned period of centralized thermal network operation. Figure 4a,b shows the inner temperature of the NTES according to thermal energy charging and discharging during the operation of the centralized thermal network. The blue area represents the heat charged into the NTES through heat pump operation ( . Q cha,NTES,C ), and the orange area represents the discharging of the cooling to the thermal network from the NTES ( . Q CN,C ). As mentioned above, heat is charged into the NTES through heat pump operation during the off-peak period, and when the temperature rises in the heat storage tank, cooling is charged through heat pump operation even during the peak period. The experimental results demonstrate that most heat charging occurs during the off-peak period. When the set cooling temperature cannot be maintained in the thermal network owing to a temperature increase at the bottom of the heat storage tank to more than 9 • C, heat is charged into the heat storage tank even during the peak period. Figure 4c shows the distributions of the cooling water supply temperature (T cws(TM7) ) and cooling water return temperature (T cwr(TM11) ) inside the thermal network. The results confirm that the cooling water is supplied to the thermal network at approximately 11 • C and returned at 13 • C. The indoor temperatures of the buildings in the thermal network are maintained at appropriate levels. Supply temperature fluctuations occurred because the flow rate of the thermal network in actual operation is operated at a constant flow rate, and when a load is generated, the heat is supplied by the pump on/off of the primary heat exchanger in the thermal network. This phenomenon occurs when the load of a building is relatively small. Figure 4d depicts the temperature distributions according to the cooling water supply (T AHP,out(Ts2) ) and return (T AHP,in(Ts1) ) through the operation of the AHP in the school. The results show that the cooling energy supply temperature of the school is approximately 11 • C and the return temperature is approximately 14 • C, indicating a stable heat supply. and when a load is generated, the heat is supplied by the pump on/off of the primary heat exchanger in the thermal network. This phenomenon occurs when the load of a building is relatively small. Figure 4d depicts the temperature distributions according to the cooling water supply ( , ( ) ) and return ( , ( ) ) through the operation of the AHP in the school. The results show that the cooling energy supply temperature of the school is approximately 11 °C and the return temperature is approximately 14 °C, indicating a stable heat supply.    Figure 5 shows the operation profile for the representative period from August 12 to 14 during the decentralized thermal network operation period. During this period, heat was supplied by the centralized thermal network method and the heat prosumer method in combination. Figure 5a,b depicts the inner temperature of the NTES during decentral-  Figure 5 shows the operation profile for the representative period from 12 to 14 August during the decentralized thermal network operation period. During this period, heat was supplied by the centralized thermal network method and the heat prosumer method in combination. Figure 5a,b depicts the inner temperature of the NTES during decentralized thermal network operation. In Figure 5a, the blue area represents the heat charged into the NTES through heat pump operation ( . Q cha,NTES,C ) and the orange area represents the heat discharged from the NTES ( . Q CN,C ). In addition, the green area represents the bidirectional heat supply from the school to the thermal network according to the prosumer operation. In the conventional centralized thermal network operation, the temperature of the NTES increases to more than 9 • C ( Figure 5b); thus, the central heat pump that uses electricity must be operated even during the peak period. However, the heat prosumer of the school, which is an AHP using gas, assists in reducing the power consumption during the peak period using the non-electric thermal energy production facility. Regarding heat prosumer operation, Figure 5c reveals that there is no flow to the school in the conventional thermal network (Figure 4c), although a pipe connected to the school was operated for heat prosumer operation (i.e., FM4). During heat prosumer operation, the temperatures corresponding to TM7 and TM10 are the same, confirming the possibility of receiving cooling energy from the heat pipe of the school and supplying it to the public building. An examination of the operation status of the heat pipe of the school during the same time period confirms that, as shown in Figure 5d, the flow rate (Fs2) of the school side of the heat exchanger connected to the thermal network is generated during heat prosumer operation. It is also evident that a bi-directional thermal network is able to work because the cooling water that passes through the AHP supplies cooling to the thermal network by exchanging heat through the heat exchanger, increasing the heat exchanger outlet temperature in the school (T sch,hx,out(Ts4) ).

Operating Cost and Environmental Impact of the Bi-Directional Thermal Network
During the demonstration period, the AHP was operated from 3 August for the cooling energy supply and prosumer operation of the school. The total LPG gas consumption during operation is 1598 Nm 3 . The supply to the public building through the AHP in the bi-directional thermal network is 8301.8 kWh. The volume of LPG that is additionally consumed by the AHP for prosumer operation is 332 Nm 3 , or 20.8% of the total gas consumption. The current price of LPG gas is 2.76 USD/Nm 3 ; thus, the cost of gas consumption for prosumer operation is 914.1 USD. In contrast, assuming that the heat supplied by the absorption heat pump is supplied through the daytime operation of the heat pump in the centralized mechanical room, an additional 2004.8 kWh of electricity is used during the day, and the power cost is 229.9 USD. It is assumed that the heat pump will show a COP of 4.16, as the mean COP of the ground-source heat pumps in the centralized mechanical room.
As mentioned above, LPG gas was temporarily used to operate the absorption heat pump in the school. The cost of LPG gas (i.e., 0.03 USD/MJ) is 3 times higher per megajoule than that of general city gas (i.e., 0.01 USD/MJ). Thus, it was calculated that if the absorption heat pump is operated by city gas, its operation cost will be 301.2 USD. Therefore, even if LPG gas is changed to city gas, the operation cost will be 1.31 times higher if the peak power rate is 0.12 USD/kWh. As shown in Figure 6, if the peak power rate increases, a greater reduction in the operation cost can be obtained than is possible by using the AHP. It is predicted that this reduction is achievable if the peak rate is greater than 0.16 USD/kWh.   than that of general city gas (i.e., 0.01 USD/MJ). Thus, it was calculated that if the absorption heat pump is operated by city gas, its operation cost will be 301.2 USD. Therefore, even if LPG gas is changed to city gas, the operation cost will be 1.31 times higher if the peak power rate is 0.12 USD/kWh. As shown in Figure 6, if the peak power rate increases, a greater reduction in the operation cost can be obtained than is possible by using the AHP. It is predicted that this reduction is achievable if the peak rate is greater than 0.16 USD/kWh. The reduction rate of carbon emissions was analyzed using the demonstration operation. The results showed that the demonstration operation generated total carbon emissions of 5.75 tCO2 in the conventional centralized thermal network method, whereas the operation of the decentralized heat prosumers produced total carbon emissions of 7.15 tCO2. However, the total cooling load of the town was smaller during centralized thermal network operation in the demonstration experiment. Thus, the greenhouse gas generation related to the total cooling load (in megawatt-hours) of the town was calculated accordingly. Consequently, as shown in Figure 7, the conventional centralized thermal network method produces 2.22 tCO2/MWh of carbon emissions, whereas the decentralized heat prosumers yield 1.94 tCO2/MWh of carbon emissions. Hence, the carbon emission is reduced by 12.7%. The reduction rate of carbon emissions was analyzed using the demonstration operation. The results showed that the demonstration operation generated total carbon emissions of 5.75 tCO 2 in the conventional centralized thermal network method, whereas the operation of the decentralized heat prosumers produced total carbon emissions of 7.15 tCO 2 . However, the total cooling load of the town was smaller during centralized thermal network operation in the demonstration experiment. Thus, the greenhouse gas generation related to the total cooling load (in megawatt-hours) of the town was calculated accordingly. Consequently, as shown in Figure 7, the conventional centralized thermal network method produces 2.22 tCO 2 /MWh of carbon emissions, whereas the decentralized heat prosumers yield 1.94 tCO 2 /MWh of carbon emissions. Hence, the carbon emission is reduced by 12.7%.

Discussion and Future Work
This study experimentally proves the possibility of decentralized prosumer heat trading during the cooling season in a centralized thermal network and demonstrates the achievable operation cost and carbon emission reductions. The thermal network used in this study is a small thermal network with seven buildings, which is smaller than a general thermal network. However, this research is meaningful because it provides experimental results in a thermal network reflecting the recent concept of low-temperature heating and high-temperature cooling, such as fourth-generation district cooling and heating. In future studies, thermal load prediction and operation optimization based on the prediction are required, and the present study serves as a basic demonstration for this future work. As analyzed in this research, an operation optimization strategy can be established based on the operating cost and environmental impact in terms of CO2 emissions. Furthermore, future studies will be conducted from the perspective of heat sharing and trading for electric and thermal convergence systems for zero-carbon emission communities.
Short-term thermal energy storage (TES) for a one-day load is important for reducing the operating cost when implementing a thermal network. However, the initial cost in-

Discussion and Future Work
This study experimentally proves the possibility of decentralized prosumer heat trading during the cooling season in a centralized thermal network and demonstrates the achievable operation cost and carbon emission reductions. The thermal network used in this study is a small thermal network with seven buildings, which is smaller than a general thermal network. However, this research is meaningful because it provides experimental results in a thermal network reflecting the recent concept of low-temperature heating and high-temperature cooling, such as fourth-generation district cooling and heating. In future studies, thermal load prediction and operation optimization based on the prediction are required, and the present study serves as a basic demonstration for this future work. As analyzed in this research, an operation optimization strategy can be established based on the operating cost and environmental impact in terms of CO 2 emissions. Furthermore, future studies will be conducted from the perspective of heat sharing and trading for electric and thermal convergence systems for zero-carbon emission communities.
Short-term thermal energy storage (TES) for a one-day load is important for reducing the operating cost when implementing a thermal network. However, the initial cost increases as the amount of TES increases. If the amount of TES is small, the operating cost inevitably increases because of the additional operation of heat pumps during the daytime. In this case, if a non-electric heat supply during the daytime can be achieved using idle equipment in a building connected to a thermal network, as in this study, it will be effective for reducing the initial investment cost and operating cost. Therefore, we plan to study design optimization considering this operation method and the initial investment cost.
Regarding carbon emissions reduction, South Korea has high CO 2 emissions from power generation. It is estimated that this high emission level occurs because coal-fired power plants, which have low power production costs but high CO 2 emissions, constitute the major source of power generation [30]. If natural-gas-fired power plants are used instead, CO 2 emissions will be reduced, but the power generation cost will rise to 0.09 USD/kWh, resulting in a 28% higher power generation cost than that of coal-fired power plants, which is 0.07 USD/kWh. However, the use of natural-gas-fired power plants will continuously increase with international efforts to reduce carbon emissions, and the peak electric price is expected to increase gradually as a result. Consequently, the utilization of a non-electric thermal supply is expected to increase.

Conclusions
This study implemented heat prosumer buildings in a small thermal network and proposed a control method for a heat-trading system using heat prosumers. Heat prosumers were implemented on a real scale in a high school and childcare center. The operation results during the cooling season in terms of the heat trading of each building were also presented. The results indicated that the temperature of the thermal network ranged from 11 • C to 13 • C. It proved that similar to the centralized heat supply case, decentralized heat prosumer buildings can be used in a thermal network by supplying an appropriate level of cooling energy to the network. In this research, the operating cost was also calculated using the current gas and electricity demonstration operation. It was found that the current demonstration operation cannot achieve economic benefit. However, if the peak power cost becomes higher than 0.16 USD/kWh and general city gas (i.e., 0.01 USD/MJ) is used, the heat production and supply using AHP could operate cost effectively in the future. Further, the proposed bi-directional operation method is expected to achieve a carbon emission reduction of 12.7% compared to the conventional centralized thermal network method.