Analysis of the Operation Characteristics of a Hybrid Heat Pump in an Existing Multifamily House Based on Field Test Data and Simulation

Abstract

Unrenovated multifamily houses in Germany are mostly heated by fossil heat generators; therefore, measures are required for CO₂ emission reduction. The use of air–water heat pumps is restricted by high required flow temperatures and heat output but can be mitigated by hybrid heat pumps. To limit additional operation costs by the heat pump, a new operation strategy is introduced in this study, which allows to maintain a target CO₂ emission. The operation strategy is applied in a field trial in a small unrenovated multifamily house built in 1964. A thermohydraulic simulation model is verified and is used in full-year simulations to apply improvement measures and compare the new control strategy with existing optimizing strategies. The results show that the control onto target emissions is possible and limits additional costs but can also result in higher CO₂ mitigation costs, making it less favorable. The hybrid heat pump reduces the direct fossil CO₂ emissions by 61% (in total by 22%); thus, it is a relevant solution for multifamily houses, especially within a further decarbonized electrical grid.

1. Introduction

1.1. Motivation

The German federal government is aiming for a 68% CO₂ emission reduction in the building sector by 2030, referring to 1990, in order to fulfill the targets of the Federal Climate Change Act (Bundes-Klimaschutzgesetz) [1]. With a share of 84% of the final energy consumption for the residential sector in 2017, space heating (SH) and domestic hot water (DHW) preparation are the main consumers and 71% are provided by fossil energy sources such as gas or oil [2]. To enforce the use of renewable energy, the German government strives for a 65% share of renewable heat in the case where a heating system is newly installed by 1 January 2024 [3]. Since this also applies in case of a replacement, the installation of new gas boilers as a single heat source would effectively be prevented. On a European level, a ban of gas boilers as the sole heat generator is proposed for 2029 [4].

Heat pumps, on the other hand, can reduce fossil energy consumption in residential buildings and those systems see an increasing number of installations each year. A majority of 82% of the newly installed heat pumps in Germany in 2021 are air–water heat pumps, while only 18% use soil or groundwater [5]. A similar distribution of heat sources can be observed on a European level [6]. While heat pumps are the predominant heat generators in newly built houses, the share of heat pumps gets smaller with increasing building age and is especially low in existing multifamily houses [7]. Considering that more than every second inhabited flat in Germany is situated in multifamily houses, this poses a significant potential for the reduction of fossil energy usage and shows a gap to the intended ban of gas boilers. As multifamily houses are prevalent in consolidated urban areas, major limitations are space restrictions for using environmental heat due to noise protection or earthworks of geothermal probes [8]. The area of tension, due to higher total heat demand of multifamily houses and the restricted space in cities, can be partially mitigated by a reduction in the required heat pump output in a hybrid heat pump system [9]. With respect to higher investment costs for heat pumps, compared to boiler systems, the combination of both technologies also offers potential for a reduction in the investment costs [10]. Another barrier for heat pumps in existing buildings originates from the often required high supply temperatures, which drastically reduce the efficiency of heat pumps [11]. Although the SH flow temperature can be lowered through refurbishment or the application of low-temperature emission systems, DHW preparation remains a limiting factor for the usage of heat pumps. Due to hygiene requirements, high water temperatures need to be maintained in storage water heaters and the distribution network, which leads to low efficiencies of heat pumps [12].

Hybrid heat pumps are a possibility to introduce heat pumps to existing multifamily houses. The integrated control of a hybrid heat pump offers various operation strategies, such as bivalent point control [13,14,15], optimized operation for minimal cost, CO₂ emission or primary energy consumption at that point in time [16,17], or model predictive control [18,19,20]. The emission optimization of hybrid heat pumps can achieve considerable emission reductions with a smaller increase in operation costs (as of 2020) compared to monoenergetic heat pumps [21].

1.2. Objectives

While several studies investigate hybrid heat pumps in single-family houses with a focus on space heating operation, only few address multifamily houses [21], and no research covers the performance in unrenovated building classes. Furthermore, the findings emphasize a significant operation cost increase due to the application of emission-optimized controls, which can lead to a conflict of interests (e.g., between tenants and owners in multifamily houses). The analysis is conducted with energy prices before the energy crisis of 2022, which at least temporarily alters the obtained operation costs.

To mitigate this conflict, a control strategy is developed in this study to maintain a target-specific CO₂ emission, which can reach a balance between the state-of-the-art optimization of costs or emissions at a point in time. The control strategy, the overall energetic performance, and operation characteristics are demonstrated by the evaluation of field test data of a hybrid heat pump in an unrenovated multifamily house, built in 1964, in Section 3. The thermohydraulic simulation model of the building energy system is verified with the field test data in Section 4.1. The verified model allows to investigate the effect of different improvement measures, such as a flow temperature reduction and the fix of a malfunctioning nonreturn valve, in full-year simulations in Section 4. The comparison of the new control strategy with a cost- and emission-optimized control is conducted with the optimized system and two different electricity tariffs. The results demonstrate not only the different performance figures of the control strategies but also point out the maximal achievable emission reduction of a hybrid heat pump under unfavorable conditions.

2. Material and Methods

2.1. Building and Location

The field test was conducted in an unrenovated detached multifamily house with 5 apartments, which was constructed in 1964. During the field trial period, four of the apartments were inhabited. This type of building is characteristic of approximately 12.8% of the inhabited flats in Germany, which are situated in small multifamily houses of 3 to 6 flats and were built between 1949 and 1978, before the German thermal insulation ordinance became effective [7]. The building is located in the postcode area 08626 in Germany with the nearby German meteorological service (DWD) weather stations Elster, Bad-Sohl (ID 01207), and Hof (ID: 02261) and is depicted in Figure 1.

According to the national annex of the DIN EN 12831, a design outdoor temperature of −14.7 $°$C is applicable [22]. The outdoor temperature frequency of the location from 28 June 2020 until 27 June 2021 is depicted in Figure 2 for days with a mean temperature below the heating limit of 17 $°$C.

The degree days for the same period at the location account for 4509 Kd with a room temperature of 20 $°$C and the defined heating limit. The climate factor of the postcode area from 1 May 2020 to 30 April 2021, provided by the DWD, belongs to 6% of the coldest locations for this period in Germany.

Relevant data of the field trial building are provided in

Table 1. Building data of the field trial.

Number of floors	3
Number of flats (inhabited)	5 (4)
Base area	151 m2
Window area	62 m2
Useable building area	292 m2
U-value exterior wall (W/(m2K))	1.40
U-value roof (W/(m2K))	1.40
U-value top floor (W/(m2K))	1.00
U-value ground floor (W/(m2K))	0.56
U-value windows (W/(m2K))	1.90
Heat load, transmission	24.2 kW
Heat load, ventilation	2.7 kW
Total heat load	27.0 kW

.

The calculated transmission and ventilation heat losses of the field trial are based on the above stated design room and outdoor temperature, and an air exchange rate of 0.34 h${}^{-1}$ with the simplified approach of the DIN EN 12831 [24].

2.2. Heat Generation System and Settings

The previously installed 40-kW gas boiler and a 200 l domestic hot water storage were replaced by a hybrid heat pump system, as displayed in Figure 3.

The system consists of an 8.4-kW (A-7/W35) air–water heat pump with a COP of 2.96, a gas condensing boiler with a thermal output of 4.0 to 27.5 kW, and a 277 l domestic hot water storage with an internal coil heat exchanger. The heat pump is located upstream of the boiler and is connected to the boiler hydraulic circuit with a bypass. The hydraulic is a serial connection of both heat generators, in which the heat pump increases the temperature of the boiler return temperature. The boiler is undersized w.r.t. to the combined SH and DHW demand due to a planned renovation of the building before the next heating season. The pipe to the heat pump is 7.3 m in single length inside the building envelope and 1.3 m outside. The COP is calculated according to Equation (1) and considers thermal losses when the heat pump is off, used thermal and electrical energy during defrost cycles, and the electrical consumption of supply pump P1.

$$COP=\frac{Q_{\mathrm{th},\mathrm{hp}}}{W_{\mathrm{el},\mathrm{hp}}+W_{\mathrm{el},\mathrm{P}1}}$$

(1)

The radiator heat emission system, consisting of panel radiators and the domestic hot water pipework with circulation line, were not changed.

Due to the unrenovated state of the building, a weather compensation control (WCC) with 76 $°$C flow temperature at −15 $°$C outdoor temperature was applied, which is similar to the observed temperatures of the prior system. The room temperature set point is 20 $°$C with a night setback to 18 $°$C during the time from 23:00 to 5:00. The flow temperature set point, the building heat load, and the thermal output of the heat pump at the maximum compressor frequency are displayed in Figure 4. The capacity of the heat pump is calculated with the simulation model in Section 2.4.

Based on the maximum thermal capacity of the heat pump, a theoretical bivalence point of 4 $°$C can be obtained. The performance share for central heating ${\xi }_{\mathrm{sh}}$ with the design heat load ${\mathsf{\Phi }}_{\mathrm{design}}$ is 19%, calculated by Equation (2). As the required flow temperature at design conditions is outside the heat pump’s operation envelope, the maximum flow temperature of the heat pump is used for the calculation leading to the operation point A-14.7/W55 with 5-kW thermal output. The funding scheme for hybrid heat pumps in Germany requires a performance share of 25% at the design outdoor temperature and 35 $°$C flow temperature [25]. The lower required flow temperature of the funding scheme leads to an overestimated performance share of 24%.

$${\xi }_{\mathrm{sh}}=\frac{{\dot{Q}}_{\mathrm{th},\mathrm{hp},\max }}{{\mathsf{\Phi }}_{\mathrm{design}}}=\frac{5.0\mathrm{kW}}{27.0\mathrm{kW}}=0.19$$

(2)

A DHW storage charge is started if the temperature at the storage sensor drops below 60 $°$C and ends if the sensor temperature reaches 65 $°$C. This setting ensures the fulfillment of the German regulation DVGW 551 [26], which requires for large DHW systems a minimal storage outlet temperature of 60 $°$C to maintain DHW hygiene. The DHW distribution system is assumed to exceed a water content of 3 l; therefore, it is considered to be “large” according to the German regulation for potable water [27].

Due to the high set point, the heat pump cannot be operated for storage charges, since the maximum condensation temperature of 65 $°$C would be exceeded at any point of the charging. Therefore, the storage is charged by the boiler with a constant flow temperature set point of 85 $°$C.

Due to hygiene requirements, the circulation pump is operated constantly to avoid temperature drops in the distribution pipework.

2.3. Data Acquisition

The displayed quantities in Figure 3 are recorded with a 30-s resolution. For thermal powers, heat meters (Amtron Sonic D and WDV Molline S44) with ultra sonic flow metering and Pt500 are used, which are in accordance with DIN EN 1434-1 [28]. The electricity meters (KDK-Dornscheid) are similar to the described equipment in Günther et al. [11], in which a maximum error of 5% was obtained for the COP.

The gas consumption of the boiler is assessed via a magnetic rotary disc of the gas meter (BK-4 MT), which is in accordance with DIN EN 1359 [29]. Since the transition flow of 0.6 $\frac{{\mathrm{m}}^3}{\mathrm{h}}$ is exceeded during a large portion of the operation range, a maximum deviation of 3% can be assumed, leading to an error of less than 5% for the boiler efficiency.

An additional influence arises from the calculation of provided heat based on the logged data in the 30 s interval. This is necessary to deduct extracted heat during defrosts and to consider influences of thermal capacity, since the heat meter only accumulates positive values.

Data of the device controls, e.g., actuator states, energy monitoring data, and NTC temperature sensors, are evaluated.

Since the flow temperature sensor of the boiler is wrongly installed, as shown in Figure 3, during SH operation the sensor T_flow,sh is used and during DHW operation the boiler flow temperature is used from the internal controls.

2.4. Modeling

The simulation is conducted in a proprietary MATLAB Simulink library of Bosch Thermotechnology GmbH, which incorporates weather data, user behavior, and building properties to simulate the thermohydraulic behavior of the heat generation system with its corresponding controls. An overview of the general model structure is provided in [30,31] and is displayed for the hybrid heat pump system in Figure 5.

2.4.1. Weather

The weather data input contains temperatures of the air, sky, water, and soil as well as the air humidity and velocity in combination with radiation intensity and angle. The respective data points are imported from the surrounding DWD weather stations of Bad Elster (ID: 1207) and Hof (ID: 2261). The latter one provides more required properties but results in a higher deviation compared to the locally measured temperature.

As the sky temperature is not recorded at nearby meteorological stations, the “Einfachstmethode” is applied to calculate values [32]. All weather data are obtained from the DWD open data server [33].

2.4.2. User

The user model provides profiles of occupancy, ventilation, electrical device usage, domestic hot water tapping, and room temperature set points for 4 actively heated zones and one unheated zone of the building. The four actively heated zones combine all room types of the five flats in the field trial building. The heating set points for the zones are provided in

Table 2. Thermal zones of the building model with respective usage type and heating set point.

Zone Number	Zone Description	Heating Set Point ($°$C)
1	Living room	20
2	Kitchen/bath	20
3	Office	20
4	Bedroom	19
5	Cellar/Attic/Stairway	-

.

The occupancy and window opening profile for zones 1 and 2 are displayed in Figure 6 and for zones 3 and 4 in Figure 7. The heat emission of a person is assumed with 95 $\mathrm{W}$ and a time-variable electrical device usage with a weekly consumption of 133 kWh.

The DHW tapping profile is generated by the program DHWcalc [34] to distribute the average weekly consumption of 810 l tapped water. The water tapping takes place in zone 2 and leads to additional internal gains.

2.4.3. Building Model and Heat Transfer System

The building model of the TRNSYS Type 56 is used and the zonal distribution is displayed in Figure 8.

The unheated zone 5 contains the attic, stairway, and basement and receives the air temperature $T_{\mathrm{air},\mathrm{zone}}$ as an input to calculate the losses of the pipework, the boiler, and storage, which are fed back as gain to the respective zone. The operative temperature $T_{\mathrm{op},\mathrm{zone}}$ as mean value from the air and wall surface temperature of the heated zones is used by the radiator models to calculate the convective ${\dot{Q}}_{\mathrm{conv}}$ and radiative power ${\dot{Q}}_{\mathrm{rad}}$, which are fed back to the zones. The thermostatic radiator valves control the operative temperature onto the user set point $T_{\mathrm{trv},\mathrm{set}}$ for each zone, according to Table 2.

For each zone, five parallel radiators are simulated that have identical thermohydraulic properties but receive an altered set point to reflect different user behavior and avoid simultaneity of the thermostat operation. The radiator set point $T_{\mathrm{trv},\mathrm{set},\mathrm{i}}$ for each radiator i is scaled equidistantly around the zone set point with offsets of –1, –0.5, 0, 0.5, and 1 K, which are estimated to fit the simulation behavior to the measured volume flow of the heat transfer system.

2.4.4. Heat Generator

• Boiler

The boiler is modeled as a one-dimensional thermal capacity. The temperature distribution is calculated by Equation (3) with z thermal nodes and the capacity $C_i$ as the zth part of the boiler block capacity.

$$C_i·\Delta T_i={\dot{Q}}_{\mathrm{in},i}+{\dot{Q}}_{\mathrm{amb},i}+{\dot{Q}}_{\mathrm{con},i}+{\dot{m}}_i·c_{\mathrm{p}}·(T_i-T_{i-1})$$

(3)

A graphical representation of this model is provided in Figure 9.

Each node exchanges heat with the ambiance ${\dot{Q}}_{\mathrm{amb},i}$ driven by the temperature difference and in dependency of a heat exchange coefficient. Furthermore, the nodes exchange heat with neighboring nodes via a conductive and convective term ${\dot{Q}}_{\mathrm{con},i}$ and can receive heat from an input ${\dot{Q}}_{\mathrm{in},i}$, which, in the case of the boiler, is the conversion of the gas energy to heat by Equation (4) with the conversion efficiency ${\eta }_{\mathrm{b}}$.

$${\dot{Q}}_{\mathrm{in},i}=\frac{{\dot{Q}}_{\mathrm{gas}}·{\eta }_{\mathrm{b}}}{z_{\mathrm{nodes}}}$$

(4)

The efficiency is modeled by an empirical polynomial, which, inter alia, reflects the condensation influence similar to Seifert [35] and considers the flow $T_{\mathrm{flow},\mathrm{b}}$ and return temperature $T_{\mathrm{ret},\mathrm{b}}$, as well as the used gas power ${\dot{Q}}_{\mathrm{gas}}$.

The hydraulic pressure loss of the boiler is modeled by using the simplified approach of a straight pipe with a length $l_{\mathrm{hyd}}$ of 1 m, for which the Colebrook equation is applied and iteratively solved. The hydraulic diameter $d_{\mathrm{hyd}}$ and friction factor f are optimized to reproduce catalog pressure drop data of the boiler. The pressure drop of the system and the pump head are represented in a simplified generic approach by the coefficients $p_1$, $p_2$, and $p_3$ in Equation (5).

$$\Delta p=f·\frac{l_{\mathrm{hyd}}}{d_{\mathrm{hyd}}}·\frac{\rho }{2}{v}^2=p_1+p_2·\dot{m}+p_3·{\dot{m}}^2$$

(5)

The coefficients of all hydraulic components are used to iteratively solve the hydraulic network, as described in Hähnlein [36] by using matrices for the mass flow balance at all nodes and the pressure equality in intermeshed circuits.

• Heat pump

The heat pump model couples the thermal capacity model and the hydraulic pressure loss model for the sink and source side of the refrigerant cycle with a polynomial curve based model, as shown in Figure 10.

The curve-based model uses the approach of Afjei et al. [37] and adds the influence of the relative frequency of the compressor for the calculation of the COP and electrical power consumption of inverter-driven heat pumps. After the start up of the compressor, a first-order reduction in the COP is applied, due to the inertia of the refrigerant cycle.

The air outlet temperature is an input of the polynomial model, as this allows the consideration of icing and condensation influences on the heat pump efficiency and, therefore, the thermal output. An increasing ice layer results in a higher air side resistance due to the reduction in the free hydraulic area. An increased air pressure drop leads to a reduction in the air mass flow and, consequently, the air outlet temperature drops. As this also leads to a reduction in the evaporation temperature of the refrigerant, the temperature difference between air inlet temperature and evaporation temperature is used to start a defrost by the heat pump controller. During a defrost, the refrigerant cycle is reversed by a four-way valve, which leads to an intake of heat from the water side to the air side to melt the ice layer.

The applied model allows a more detailed simulation compared to performance table models, which typically do not consider thermal capacities, cycle losses, and defrosts [13,16]. An alternative to the applied curve-based model is a simulation of refrigerant properties with component models [10,38].

2.5. Energy Prices and Specific Emissions

The gas and electricity price [39] and their specific CO₂ emissions for 2020 are provided in

Table 3. Daily heat load parameter.

	Gas	Electricity
Emission $\left[\frac{{\mathrm{g}}_{{\mathrm{CO}}_{2,eq}}}{\mathrm{kWh}}\right]$	222.9	404.7
Price $\left[\frac{\mathrm{ct}}{\mathrm{kWh}}\right]$	6.1	29.3/22.5

. The stated price ratio of gas to electricity was representative for a longer period until the European energy crisis in 2022, in which an increase in the gas price can be observed. The CO₂ equivalent emissions of electricity at the low voltage transfer grid and of gas w.r.t. the lower calorific value at the building level with a power-to-gas share are obtained from GEMIS version 5.0 [40]. A heat pump tariff requires a separate electrical metering and has a price of 22.5 $\frac{\mathrm{ct}}{{\mathrm{kWh}}_{\mathrm{el}}}$.

2.6. Control Strategy for Specific CO₂ Emission

A main influence on the choice of the preferred heat generator is the spark spread, which reflects the ratio of the used final energies’ properties. If the spark spread for both optimization targets varies highly, the minimization of one value can lead to a deterioration of the other.

The given emission factors lead to a spark spread of only 2.0, while the cost ratio is 4.8 for a household tariff and 3.7 for a heat pump tariff. Based on the spark spread and the efficiency of the boiler, the $CO{P}_{\mathrm{limit}}$ of the heat pump can be calculated at which a parity of the final energy usage is reached for the respective property according to Equation (6).

$$CO{P}_{\mathrm{limit}}=\frac{x_{\mathrm{hp}}}{x_{\mathrm{b}}}·{\eta }_{\mathrm{b}}$$

(6)

Considering a condensing boiler efficiency of 86% w.r.t. the upper calorific value [41], the $CO{P}_{\mathrm{limit}}$ for CO₂ emissions is 1.7, for operation costs with household tariff 4.2, and with heat pump tariff 3.2. If the actual COP of the heat pump is above 1.7, an emission reduction compared to the boiler operation can be achieved, but leads to higher operation costs, as those can only be reduced when the heat pumps efficiency is above 4.2.

To ensure a balanced operation between the established cost or CO₂-optimized operation of hybrid heat pump systems, a new control strategy is introduced, which aims to not exceed a predefined specific CO₂ emission per provided kWh of heat according to Equation (7).

$$C{O}_{2,\mathrm{spec}}=\frac{x_{{\mathrm{CO}}_2,\mathrm{el}}\left(W_{\mathrm{el},\mathrm{hp}}+W_{\mathrm{el},\mathrm{b}}+W_{\mathrm{el},\mathrm{pump},\mathrm{hp}}\right)+x_{{\mathrm{CO}}_2,\mathrm{gas}}·Q_{\mathrm{gas},\mathrm{b}}}{Q_{\mathrm{th},\mathrm{hp}}+Q_{\mathrm{th},\mathrm{b}}}$$

(7)

For this purpose, the electrical consumption of the heat pump $W_{\mathrm{el},\mathrm{hp}}$ and its auxiliary pump $W_{\mathrm{el},\mathrm{pump},\mathrm{hp}}$; the gas consumption of the boiler $Q_{\mathrm{gas},\mathrm{b}}$; and the heat output of both generators $Q_{\mathrm{th},\mathrm{hp}}$, $Q_{\mathrm{th},\mathrm{b}}$ are estimated by the control unit of the system. This is achieved based on characteristic curves, which use the relative actuator power and temperatures. Auxiliary consumers such as DHW circulation pump, SH pump, and storage charge pump are not considered, since the used pump setup is not known by the system controller and the electric energy consumption is expected to be low.

The final energy consumption is converted to a CO₂ emission with the respective specific emission coefficients $x_{{\mathrm{CO}}_{2,\mathrm{el}}}$ and $x_{{\mathrm{CO}}_{2,\mathrm{gas}}}$, which can either be provided statically by a customer input—as in this investigation—or dynamically via an internet connection. The latter one becomes increasingly important, as the usage of fluctuating renewable energy sources, such as wind or solar, leads to significant variation of cost or emission parameters throughout the day or the year [42].

The current specific CO₂ emission is compared with a set point $C{O}_{2,\mathrm{spec},\mathrm{set}}$ according to Figure 11.

Based on the emission factor of natural gas and a condensing boiler efficiency of 86%, a specific CO₂ emission of 233 $\frac{{\mathrm{g}}_{{\mathrm{CO}}_2}}{{\mathrm{kWh}}_{\mathrm{th}}}$ can be expected for a boiler system as a typical reference. For the field trial, a set point of 190 $\frac{{\mathrm{g}}_{{\mathrm{CO}}_2}}{{\mathrm{kWh}}_{\mathrm{th}}}$ is set, which would equal a reduction of 18.5%. This value is expected to be above the maximum possible CO₂ emission reduction and shall show the switching operation of the heat sources.

If the emission target is exceeded, the heat pump is activated, in the case that a potential exists to reduce the specific CO₂ emission. The efficiency of boiler and heat pump are calculated based on the simulation model approach in Section 2.4. If the heat pump is running and surpasses the minimal required COP to reduce the system’s CO₂ emission, it is switched off and the boiler is used. Conversely, the heat pump is used if its operation leads to a cost reduction. During the heat pump operation, the boiler can be additionally activated regardless of the emission target, if the space heating flow temperature set point is surpassed for a time of 20 min or instantly if the required flow temperature is above the maximum flow temperature of the heat pump.

3. Results of Field Data Evaluation

As hybrid heat pump systems have a variety of different possible operation strategies, the field data are evaluated both with a focus on energetic performance as well as the detailed operation characteristic to identify typical features. These features are used to verify the simulation model in Section 4.1.

3.1. Energetic Performance

The control algorithm was initially applied from 18 December 2020 and the provided heat in SH and DHW operation mode is displayed in 1 K temperature classes for both heat generators in Figure 12 for the period until 27 June 2021.

3.1.1. Space Heating

The calculated design heat loads and the mean measured heat loads for space heating show a high discrepancy for each bin. This is partially caused by the vacancy of one flat during the measurement period. Furthermore, the design heat load does not consider additional thermal gains by user activity and solar radiation as well as the thermal inertia of the heated zones. Based on the measured mean heat load at design conditions of 17.5 kW in Figure 12, an actual performance share ${\xi }_{\mathrm{sh}}$ of 29% is calculated.

A maximum of heat is provided around 0 $°$C as, based on the outdoor temperature distribution in Figure 2, a maximum of hours occurs in this temperature range. If a temperature of –8 $°$C is surpassed, the heat pump is not operated due to the high corresponding flow temperatures of at least 64 $°$C and the low achievable COP. Furthermore, the heat pump can cover the heat demand above an ambient temperature of 5 $°$C nearly without boiler support and reaches a coverage ${\alpha }_{sh,hp}$ for SH of 57%. This coverage is below the estimated value of 87% of the VDI 4650 [43] or 80% in Annex 45 [44] at the corrected performance share of 29%. The lower reached share is caused by only partial monitoring of the heating season without favorable hours in the autumn. The high required flow temperatures lower the thermal output of the heat pump and already 9% of the heat is required below −8 $°$C and, therefore, cannot be provided by the heat pump.

The specific space heating consumption of 155 $\mathrm{kWh}/{\mathrm{m}}^2\mathrm{a}$ from 28 June 2020 to 27 June 2021 is calculated based on the measurement data and an extrapolation of the uncovered period based on the heating degree days and the linear heat load fit of the building in Figure 12. With regard to a typical heating consumption of at least 138 $\mathrm{kWh}/{\mathrm{m}}^2\mathrm{a}$ for buildings with 5 to 7 flats, which are supplied by a gas or oil boiler, the obtained consumption is considered plausible [45].

3.1.2. Domestic Hot Water Preparation

In DHW mode, the heat pump cannot provide heat; thus, this demand is completely covered by the boiler, which accounts for 26% of the total provided heat of the system. At 45%, nearly half of the provided heat during DHW preparation is not transferred to the storage and heats up the thermal capacity of the heat generator system and the building; a defective nonreturn valve in the SH circuit led to a reversed flow in the heat distribution system when the DHW charge pump was activated.

Of the transferred heat to the storage, 79% is transmitted to the building via losses of the circulation system and the storage. The circulation losses are mainly influenced by the high storage temperature and the constant operation of the circulation pump. The storage losses, obtained as a residual of extracted heat from the storage and provided heat via the internal coil, account for 21%, which results in only 21% of the energy used for tapping.

Based on an assumed tapping temperature of 40 $°$C, the equivalent tapped water volume can be calculated, which accounts for 810 l per week. Considering 5 tenants, 23.1 l per day and person are tapped, which is at the lower range of the VDI 2067 part 12 [46].

3.1.3. Energy Efficiency

The boiler operation results in a mean efficiency of 88.1% w.r.t. to the upper calorific value, and the mean COP of the heat pump is 2.65. The dependency of both efficiency values from the daily outdoor temperature is displayed in Figure 13, in which the boiler efficiency is scaled with the ratio of the CO₂ emission factors of both used final energies.

As nearly all reached COP of the heat pump are above the CO₂ scaled boiler efficiency, the control algorithm successfully avoided operation conditions in which the heat pump operation would not ensure an emission reduction. The decrease in boiler efficiency at higher outdoor temperatures results from additional heat losses due to an intermittent operation in the summer time, as described in [41].

The reduction in the heat pump efficiency at higher outdoor temperatures is influenced by the same effect. The total heat losses at the outdoor unit are assessed in Section 4.2.

The mean flow temperature ${\overline{T}}_{\mathrm{flow},\mathrm{hp}}$, weighted with the provided heat of the heat pump, is 46.1 $°$C with a mean outdoor temperature ${\overline{T}}_{\mathrm{outdoor}}$ of 1.8 $°$C.

The COP of the field trial is at the lower end of the spectrum of COPs observed in buildings erected until 1979 in the project “WPSmart im Bestand” from 2.5 to 3.6 [11]. This can mainly be traced back to the especially low outdoor temperatures at the location and the high specific heat losses of the building.

Due to the small performance share ${\xi }_{\mathrm{sh}}$ of 29%, the heat pump is frequently operated at high relative compressor frequencies, which deteriorate the efficiency compared to systems with a higher relative heat pump capacity.

An additional influence arises from the sound protection hood of the heat pump, which directs the originally horizontal air flow of the in- and outlet to a downward-oriented vertical flow. With an increasing air flow, a recirculation occurs, which reduces the mean air inlet temperature of the heat pump by 1.6 K compared to the nearby external outdoor temperature measurement.

3.2. Control Strategy Behavior for Specific CO₂ Emissions

To investigate the impact of the newly developed control strategy, the cumulative provided heat and used final energy from measurement and internal energy monitoring of the controller are compared in Figure 14 as they are the basis for the control at a specific CO₂ emission.

It can be seen that the provided heat of the internal calculation slightly exceeds the measured one at the end of the measurement period. This results mainly from an overestimation of the provided heat by the boiler, according to the overview in

Table 4. Measured and internally calculated provided and used energy with respective deviation.

Property		Measurement (kWh)	Calculation (kWh)	Deviation (%)
Heat	Boiler	17,978	18,181	1.3
	Heat pump	17,183	17,210	0.2
Final energy	Boiler (gas)	20,415	19,979	−2.1
	Boiler (electricity)	63	51	−18.5
	Heat pump	6441	6991	8.5
	Supply pump P1	44	45	5.7
	SH pump	21	-	-
	DHW charge pump	19	-	-
	DHW circulation pump	21	-	-

.

The energy monitoring of the boiler cannot distinguish if the provided heat is actually used as the volume flow is not measured by the device. Therefore, it overestimates the provided heat especially during the summer time, when the intermittent operation leads to higher heat losses, as displayed in Figure 13. This effect is reflected in an increasing deviation of the provided heat during the summer time in Figure 14. Additionally, the gas consumption is underestimated by 2% by the internal energy monitoring, which could be caused by the burner settings during the commissioning.

The main deviation occurs at the electrical consumption, which is overestimated by 7.3% and is dominated by the electrical consumption of the heat pump with a deviation of 8.5%. A minor absolute deviation occurs at the supply pump of the heat pump P1 and the boiler fan and controller and the unconsidered auxiliary pump consumption, as explained in Section 2.6.

Due to the deviations of the calculated energy of the internal energy monitoring, the obtained specific CO₂ emission of the controller of 194.3 $\frac{{\mathrm{g}}_{{\mathrm{CO}}_2}}{{\mathrm{kWh}}_{\mathrm{th}}}$ exceeds the set point of 190 $\frac{{\mathrm{g}}_{{\mathrm{CO}}_2}}{{\mathrm{kWh}}_{\mathrm{th}}}$. Meanwhile, the corresponding specific CO₂ emission of the measurement reaches a value of 192.7 $\frac{{\mathrm{g}}_{{\mathrm{CO}}_2}}{{\mathrm{kWh}}_{\mathrm{th}}}$ due to the lower actual electrical consumption of the heat pump. The auxiliary pumps lead to an unconsidered CO₂ emission of 0.7 $\frac{{\mathrm{g}}_{{\mathrm{CO}}_2}}{{\mathrm{kWh}}_{\mathrm{th}}}$ and are, therefore, rated as neglectable.

Since the algorithm starts in the middle of December, the switching between the sources to sustain a specific CO₂ emission is only observed in December, when the heat pump can still provide a high share of the building heat load and can operate at favorable flow and outdoor temperatures. Thus, a reduction in the operation costs compared with a CO₂-optimized control algorithm can only be achieved during a few days, as the heat pump is operated in the remaining period whenever a CO₂ emission reduction is possible.

During February, the outdoor temperature reached −20 $°$C with a weekly mean temperature of −6 $°$C in two consecutive weeks. In this period, the heat pump is only operated during noon time, as the low outdoor temperatures at the remaining time of the day cannot ensure an emission reduction through the heat pump operation. The corresponding increase in the specific CO₂ emission of the system is visible in Figure 14.

Despite the high deviation of the electrical consumption of the heat pump, the internal energy monitoring values represent the measured properties with only small deviations; therefore, they can be successfully used to control the hybrid heat pump onto a target specific CO₂ emission.

The resulting performance of the system during the whole heating period and with different set points for the specific CO₂ emission is investigated in a building simulation in Section 4.

3.3. Flow Temperature Control Quality and Cycling Behavior

As no room temperature measurements are available at the field trial, the deviation between WCC set point and actual flow temperature downstream of the boiler is evaluated by Equation (8) and is used in Section 4.1 to verify the simulation model.

$$\Delta T_{\mathrm{flow},\mathrm{sh}}=min\left(\sum _{t=0}^{t=900\mathrm{s}}{T}_{\mathrm{flow},\mathrm{sh}}-T_{\mathrm{flow},\mathrm{sh},\mathrm{set}},0\right)$$

(8)

All negative temperature differences in a 15 min period are displayed in dependency of the outdoor temperature and time of the day in Figure 15.

It can be seen that the deviations increase during the day time when surpassing an outdoor temperature of 6 $°$C and sharply decline when reaching −3 $°$C. The same effect can be seen during the night with a shift towards lower outdoor temperatures. This effect is caused by both a time dependency of the defrost distribution and the building’s heat demand. The defrost cycles of the heat pump have a maximum of 0 $°$C and show a lower total number of defrosts during the daytime as the relative humidity decreases.

The building heat exhibits a maximum during the afternoon and a minimum at night time. As the heat load during the night at the same temperature is lower due to the night reduction in the WCC, the bivalent point moves to lower outdoor temperatures. The bivalent point as intersection of the mean building heat load and the maximum thermal output of the heat pump is displayed in Figure 15 for the normal and reduced time period of the WCC. The flow temperature deviations already increases before the bivalent point is reached, due to a range of building heat loads. If the outdoor temperatures falls further, the building heat load increases until the boiler can be continuously operated and causes a reduction in the flow temperature deviation.

This influence can also be seen in the daily number of starts and the runtime in Figure 16.

At high outdoor temperatures, the minimum number of starts for both generators is defined by the number of storage charges, since the high temperatures force the heat pump to stop while the boiler is not required for SH operation and will stop after DHW preparation. The heat pump reaches its maximum starts at high outdoor temperatures, due to cycling operation with a reduced runtime per day. Frequent short cycling of a heat pump can reduce the lifetime [47] and potentially lower the efficiency of the heat pump [48]. The maximum daily heat pump runtime is limited to 21 h, as for the remaining time, the storage is either charged by the boiler or the thermal capacity of the distribution system is still too warm for the heat pump operation after a charge. Despite that no buffer is installed, the heat pump runs on average 65 min per start with a total number of 1373 starts in the observed 6-month period.

When the mean outdoor temperature of 6 $°$C is surpassed, the number of boiler starts increases due to the insufficient heat pump capacity until a maximum of 29 starts is reached at −3 $°$C. With further decreasing outdoor temperatures, the boiler is nearly continuously operated; meanwhile, the heat pump’s runtime declines due to the high flow temperatures.

4. Simulation Results and Discussion

The analyzed operation characteristic of the monitoring period is used to verify the thermohydraulic simulation model. This allows the investigation of various improvements to the system and different set points for the newly developed control algorithm in full year simulations from 1 June 2020 until 31 May 2021.

4.1. Model Verification

The building simulation model is verified with the heat generation and energy consumption data, as well as the operation characteristic and flow temperature control of Section 3.1.

4.1.1. Energetic Performance

The distribution of the mean hourly space heating demand for the different months of the observed period in simulation and measurement is displayed in Figure 17.

It can be seen that the demand in the morning in the winter period is overestimated by the simulation model, whereas in the transition period, the demand in the measurement is higher and more uniformly distributed throughout the day. This can be traced back to an unknown user interaction with the thermostatic radiator valves and uncertainties of the radiation gains at the location due to shading. As a result, the measured mass flow of the heating system is more uniformly distributed.

Despite these differences, the overall heat demand profile of the field trial, such as the peaks in the morning and evening, due to the night set back influence on the WCC curve, is reproduced by the simulation. The deviations between measured and simulated provided heat and consumed final energy of the actuators are displayed in

Table 5. Provided and used energy with respective deviation in measurement and simulation.

Property		Measurement (kWh)	Simulation (kWh)	Deviation (%)
Heat	Boiler	17,943	17,165	−4.3
	Heat pump	17,116	17,009	−0.6
Final energy	Boiler (gas)	20,379	19,336	−5.1
	Boiler (electricity)	62.5	48	−23.5
	Heat pump	6416	6268	−2.3
	Supply pump P1	44	48	8.2
	SH pump	21	34	63.6
	DHW charge pump	19	20	2.9
	DHW circulation pump	21	19	−7.6

.

The time series of the provided heat and used final energy is displayed in Figure 18 for both heat generators.

The heat pump provides 0.5% higher amount of heat during space heating operation but, for total heat, an underestimation of 0.6% occurs. The simulated electrical consumption of the compressor and fan shows an underestimation of 2.3%. Therefore, the mean COP of the period accounts for 2.71 and exceeds the measured one by 0.06.

The provided heat of the boiler is reproduced with a deviation of −4.3% but shows an increasing deviation in the transition period as a result of the above-described less uniform space heating demand. According to Section 3.1, as nearly half of the heat during a DHW storage charge is not transferred to the storage, the higher expected opening ratio of the radiator valves in the measurement also favors the occurrence of higher unwanted volume flows through the heating system in the measurement. Consequently, the simulated boiler provides less heat in the transition period.

The electrical consumption of the pumps is reproduced with only small deviations by the simulation model, despite a large overestimation of the space heating pump. The main reason for this deviation is the unknown pump control in the measurement due to a proprietary control strategy of the supplier, whereas in the simulation, a constant pressure control with 150 mbar is used. The absolute deviation of the auxiliary actuators is considered sufficient for further investigations.

4.1.2. Flow Temperature Control and Cycling

To verify the operation behavior of the hybrid heat pump in the simulation, the flow temperature undershoot criteria from Section 3.3 is displayed in Figure 19.

The graphic shows a similar level and distribution of the undershoot in dependency of the time and outdoor temperature compared to Figure 15. During daytime, the undershoot significantly increases when surpassing a temperature of 5 $°$C and at night below 3 $°$C. Similar to Figure 19, the undershoot reduces when surpassing −2 $°$C in the day and −6 $°$C in the night. Thus, the two main underlying causes in the form of the bivalence point of the heat pump and the defrost occurrence are reproduced by the simulation. The total number of defrosts is underestimated in the simulation by 18%.

The daily runtime and the number of starts for both simulation and measurement are displayed in Figure 20 and Figure 21. The characteristic of the runtime of both heat generators can be reproduced by the simulation, as the heat pump is mainly operated to provide the base load with a majority of days with 19 to 21 h of run time.

The smaller heat demand in the simulation during the transition period is reflected in an overestimation of the starts of the heat pump of 8% in total. With regard to Figure 16 (left), this seems plausible, as high numbers of starts of the heat pump occur, especially on days with higher ambient temperatures. The boiler is operated as a peak load heat source and shows a similar operation in the simulation with some days of higher and lower runtime. The observed deviation is directly linked to the mean space heating demand of the days. As some days have a higher heat demand in the simulation, the boiler is also operated longer on such days and vice versa.

Despite that no measured room temperatures are available, the system provides similar flow temperatures as the replaced gas boiler, for which it is assumed that the thermal comfort was achieved. Furthermore, the verification shows that the model reproduces the characteristic of the transferred heat to the space heating system and the provided flow temperatures. Therefore, the building model is used for a scenario analysis to rate the comfort impact of different setups. According to the German tenants’ association (DMB) a minimum room temperature of 20 to 22 $°$C has to be maintained from 6:00 to 23:00 with an allowed night setback to 18 $°$C in the remaining time [49]. A failure to achieve these temperatures can result in a rent deduction. According to jurisdiction, a deduction is imposed if a certain temperature tolerance band is surpassed for a nonmarginal time. This is taken into account by Equation (10), in which the total undershoot time $\sum t_{{\Delta T}_{\mathrm{sh},\mathrm{z}}}$ is evaluated if an undershoot is identified $Y_{\Delta T,\mathrm{z}}$ that lasts for more than 2 h. An undershoot is assumed if the operative temperature surpasses the minimum temperature $T_{\mathrm{op},\min ,\mathrm{z}}$ by more than 1 K.

$$Y_{\Delta T,z}=\{\begin{array}{cc}1,\hfill & \mathrm{if} T_{\mathrm{op},z}<T_{\mathrm{op},\min ,\mathrm{z}}-1\mathrm{K},\\ 0,\hfill & \mathrm{if} T_{\mathrm{op},z}>T_{\mathrm{op},\min ,\mathrm{z}}-1\mathrm{K},\end{array}$$

(9)

$$\sum t_{{\Delta T}_{\mathrm{sh},\mathrm{z}}}=\sum _{t=0}^{t=\tau }\Delta t·Y_{\Delta T,z,t}\left[\sum \Delta t>2\mathrm{h}\right]$$

(10)

The corresponding mean operative temperature of each zone during the heating season is listed in

Table 6. Heating set point and mean operative temperature with corresponding undershoot time of the simulated heated zones.

Zone Number	Heating Set Point ($°$C)	Mean Operative Temperature ($°$C)	$\sum {\mathit{t}}_{{\mathbf{\Delta }\mathbf{T}}_{\mathbf{sh},\mathbf{z}}}$ (h)
1	20	20.1	0
2	20	20.4	0
3	20	20.2	0
4	19	19.5	0

with the resulting comfort criteria. The total undershoot time $\sum t_{{\Delta T}_{\mathrm{sh},\mathrm{z}}}$ of all zones is 0 h as no temperature undershoot exceeds 1 h in the simulation, and therefore, the comfort criteria is fulfilled.

4.2. Investigation of Improvement Measures

An overview of the investigated scenarios is given in

Table 7. Performance indicators of investigated scenarios with different setup changes.

Scenario	Nonreturn	Air	Design Flow	${\mathit{\alpha }}_{\mathbf{sh},\mathbf{hp}}$	COP	${\mathit{\eta }}_{\mathbf{b}}$	${\mathit{CO}}_{2,\mathbf{spec}}$	Cost	$\sum {\mathit{t}}_{{\mathbf{\Delta }\mathbf{T}}_{\mathbf{sh},\mathbf{z}}}$
	Valve	Recirculation	Temperature ($°$C)	(%)	(-)	(%)	(g/kWhth)	(ct/kWhth)	(h)
1	defect	yes	76	46	2.72	89.7	196.5	8.50	0
2	normal	yes	76	44	2.69	91.0	196.3	8.42	0
3	normal	no	76	41	2.76	91.3	196.1	8.21	0
Investigation of WCC set point
4	normal	no	74	40	2.80	91.4	196.1	8.11	0
5	normal	no	72	39	2.84	91.6	196.1	8.00	7
6	normal	no	70	38	2.88	91.7	196.0	7.92	52
7	normal	no	68	37	2.92	91.8	196.0	7.85	296
8	normal	no	66	36	2.96	91.9	196.0	7.76	673
Heat pump electricity tariff
9	normal	no	70	36	3.05	91.7	194.9	7.01	37

.

Other settings are in accordance with the described field test setup in Section 2.

Scenario 1 is used as a baseline, onto which different changes are applied. Those contain an intact nonreturn valve in the space heating circuit in scenario 2 and, additionally, a normally routed air flow at the outdoor unit in scenario 3. In scenarios 4 to 8, reduced WCC design flow temperatures are applied to investigate the draw off between increased efficiency of the heat pump system and a possible undersupply of the zones. A WCC set point is selected that provides a compromise between comfort and operation efficiency and is simulated again in scenario 9 with a heat pump electricity tariff.

An overview of the obtained performance indicators is given in Table 7 and is presented graphically in Figure 22. In the baseline scenario, the heat pump provides nearly half of the total supplied heat for space heating with a COP similar to the validation period. Heat losses at the heat pump circuit add up to 4.1 $\mathrm{kWh}/{\mathrm{m}}^2\mathrm{a}$, which are around 2% of the total provided heat. The boiler completely covers the DHW demand and achieves an efficiency of 89.7%.

An intact nonreturn valve in scenario 2 results in a significant reduction in the provided heat during the DHW pump operation of 33%, as the heat is then solely used to heat up the DHW storage without volume flow through the building’s distribution system. This increases the boiler efficiency by 1.3%, as the heat for space heating is provided at lower flow and return temperatures. The increased boiler efficiency leads to a reduction in the operation costs of the system of 1% and a smaller required heat pump share to not exceed the target CO₂ emission, since the boiler operation causes lower emissions compared to the baseline scenario.

The removal of the recirculation of air at the evaporator, in scenario 3, increases the COP by 0.08. With an increasing efficiency of the heat pump, it needs to provide less heat so as to not exceed the system’s CO₂ emission target. This leads to a reduced share of the heat pump and increases the boiler efficiency, as the heat pump operation increases the boiler return temperature in the applied serial hydraulic. The increased heat pump and boiler efficiency lower the operation cost by 2.5% compared with scenario 2.

The reduction in the WCC design flow temperature in 2 K steps in scenarios 4 to 8 increases the efficiency of the heat pump and boiler due to lower flow and return temperatures of the system. Consequently, the total required heat pump operation, to result in the same specific CO₂ emission, is reduced from 41% in scenario 3 with 76 $°$C to 36% in scenario 8 with 66 $°$C. The specific costs are reduced by 5.5% but longer temperature undershoots of the operative temperature also occur, as shown in Figure 23, since the heat transfer area of the radiators becomes increasingly insufficient to maintain the requested room temperature. For further investigations, scenario 6 is exemplarily selected, as it results in only few violations of the comfort limit.

In scenario 9, the lower electricity price reduces the operation costs by 11.4% and leads to a lower required heat pump share of 36%. This is mainly caused by the additional aim of the applied strategy to minimize operation costs, since the achievable COPs in the transition period are typically above the $CO{P}_{limit}$ of 3.2, as seen in Figure 13.

The results demonstrate that careful selection of the SH flow temperature, prevention of operation failures, and selection of a proper heat pump tariff can ensure a significant emission reduction with only a small increase in running costs, even in an unrenovated multifamily house.

The impact of different target-specific CO₂ emissions in comparison with a pure cost or CO₂ emission optimization algorithm is analyzed in a second step.

4.3. Investigation of Different Operation Strategies

The cost and CO₂ emission optimization algorithm is implemented by comparing the current COP of the heat pump with the $CO{P}_{\mathrm{limit}}$ for the respective optimization target derived in Section 2.6. If the current COP is above the $CO{P}_{limit}$, the heat pump is chosen as the preferred heat source. If the COP falls below the limit minus a hysteresis of 0.25, the boiler is selected. For the newly developed control strategy, the target specific CO₂ emissions of 180, 190, and 200 $\frac{{\mathrm{g}}_{{\mathrm{CO}}_2}}{{\mathrm{kWh}}_{\mathrm{gas}}}$ are applied.

An overview of the respective performance indicators for both electricity tariffs is provided in

Table 8. Performance indicators of different operation strategies with household and heat pump electricity tariff.

Scenario	${\mathit{\alpha }}_{\mathit{sh},\mathit{hp}}$ (%)	COP (-)	${\mathit{\eta }}_{\mathit{b}}$ (%)	${\mathit{CO}}_{2,\mathit{spec}}$ (g/kWh)	Cost (ct/kWh)
Household electricity tariff
Boiler	-	-	93.2	218.1	6.71
Cost opt.	2	2.72	93.2	216.9	6.78
200 g/kWh	21	2.82	92.5	206.2	7.43
190 g/kWh	38	2.88	91.7	196.0	7.92
180 g/kWh	53	2.94	90.6	185.9	8.33
CO2 opt.	74	3.05	87.7	170.0	8.74
Heat pump electricity tariff
Cost opt.	11	4.13	92.8	207.7	6.62
200 g/kWh	18	3.40	92.6	204.5	6.74
190 g/kWh	36	3.05	91.7	194.9	7.01
180 g/kWh	52	3.00	90.6	185.0	7.20
CO2 opt.	74	3.05	87.7	170.0	7.34

with a calculated boiler system and depicted in Figure 24.

The cost-optimized strategy with household tariff results in a heat pump share of only 2% of space heating, since the spark spread of 4.8, as described in Section 2.6, allows only a few operation hours at which cost-saving through the heat pump operation is possible. Despite the cost-optimized operation, only a COP of 2.72 is achieved, which can be traced back to the impact of heat losses and intermittent operation, which is depicted in Figure 13, and lowers the original generation efficiency of 4.2. Due to only small CO₂ emission reductions of 0.5% and additional costs due to the heat pump and respective auxiliary pump operation, the emission mitigation costs reach 59 $\mathrm{ct}/\mathrm{kg}$, as shown in Figure 25. The CO₂ mitigation costs are calculated by Equation (11).

$$Cos{t}_{{\mathrm{CO}}_2,\mathrm{mitigation}}=\frac{Cos{t}_{\mathrm{hybrid}}-Cos{t}_{\mathrm{b}}}{C{O}_{2,\mathrm{b}}-C{O}_{2,\mathrm{hybrid}}}$$

(11)

On the other end of the range, the CO₂-optimized strategy with household tariff leads to a total heat pump share in SH operation of 74% with a CO₂ emission reduction of 22% and mitigation costs of 42 $\mathrm{ct}/\mathrm{kg}$ with a COP of 3.05. Due to the unrenovated state of the building, the achieved emission reduction and COP are below the obtained values in Bongs et al. [21], in which the same emission factors are applied for a bivalent heat pump in a renovated German multifamily house. The heat pump tariff leads to a cost increase of only 9.1% compared to 30.3% with a household tariff. As this is traced back to the high spark spread, which persist for the past decades, the significant difference of heat pump share between a cost- and emission-optimized control is similar to the results in Form et al. [16].

The direct CO₂ emissions by the usage of fossil fuel are reduced to only 85.4 $\frac{{\mathrm{g}}_{{\mathrm{CO}}_2}}{{\mathrm{kWh}}_{\mathrm{gas}}}$ with an emission-optimized control, which is 39% of the emissions that a new boiler system would have. In an energy system with zero emissions of electrical generation, the heat pump can drastically lower the CO₂ emissions by 61%. With respect to a required emission neutrality by the middle of the century, this is still insufficient and further measures are required. The results outline the necessity to combine a heating system that incorporates renewable energies with a building renovation, as shown in the calculation of Acatech [50]. Not only will the space heating demand drop significantly due to a renovation, but also the heat pump efficiency is expected to increase due to lower required SH flow temperatures. As the building is due to be renovated before the heating season 2021/22, the impact on the energetic performance will be evaluated. Nevertheless, the high DHW temperatures still require the boiler to ensure DHW comfort and hygiene with the applied storage. A DHW storage with two sensor positions could lead to a significant DHW share of the heat pump by preheating the water, as demonstrated in Kropp et al. [12], since the boiler would only need to be activated if the storage top temperature falls below 60 $°$C.

The newly developed control strategy allows to reach an emission target between the two conventional optimizing strategies, leading to a smaller operation cost increase. As a consequence of the boiler compensation, the heat pump operates more often at the coldest time of the year, at which the high building heat load requires a frequent boiler operation and CO₂ “debts” are accumulated. This is reflected in lower COPs at high absolute emission targets, especially when a price ratio spread prevents heat pump operation in the transition period, which would allow an operation cost reduction. In case of the household tariff, this leads to higher mitigation costs compared to the emission-optimized strategy, in which the heat pump is operated frequently in the transition period. Due to the historically low spark spread of energy prices in Germany during the European Energy crisis in 2022, a sensitivity analysis has to be conducted to investigate the impact of a wider range of energy price ratios on the control strategy. To improve the practical relevance, further research is also required to extend the heat pump operation at favorable conditions, even if no cost reduction is possible. Furthermore, in the presented study, constant emission factors are applied, which do not represent the actual time dependency due to the extensive use of renewable energies [42]. Thus, further investigations will focus on the impact of variable emission factors on the control of bivalent heat pump systems.

5. Conclusions

In this study, the field test of a hybrid heat pump system, consisting of an air source heat pump with a gas condensing boiler in an existing multifamily house, is investigated. The measurement data are used to verify a thermohydraulic model of the building energy system, with respect to the energetic performance, flow temperature control, and the cycling of the heat generators. The model is applied in full-year simulations to investigate a newly developed operation strategy aiming to not exceed specific CO₂ emissions and to apply different setup optimizations, such as a flow temperature set point reduction and correction of a malfunctioning nonreturn valve. The main goal of the study was, on one hand, to demonstrate the performance figures that can be expected by applying a heat pump system in an unrenovated multifamily house and, on the other hand, to investigate the possibilities to limit additional costs through the heat pump.

Optimizing the flow temperature settings and eliminating faults of components results in a 6.8% cost reduction and significantly reduces the DHW heating demand of the boiler, while tenant thermal comfort is maintained.

Applying a cost-optimized strategy to the improved system shows an only marginal cost reduction potential of the hybrid compared to a boiler system, even with a heat pump tariff, due to the low achieved COP.

Meanwhile, the emission-optimized strategy increases the operation costs by 9% with an emission reduction of 22% compared to a new condensing boiler system, if a proper electricity tariff is selected. Despite the unfavorable operation conditions for heat pumps in existing buildings, the system could reduce the total emissions in a future carbon neutral electrical grid by 61% even in unrenovated multifamily houses in cold areas of Germany. The building represents a significant share of flats in Germany, as 12.8% of those are situated in buildings with a similar size, which were erected before thermal insulation ordinances became effective.

The newly developed algorithm successfully enables a scaling of the costs and CO₂ emissions between the two optimization strategies. The algorithm can result in higher mitigation costs due to operating at lower source temperatures compared to a CO₂-optimized strategy.

Considering a possible future ban of gas and oil boilers as single heat source, hybrid heat pumps are an immediate solution that substantially reduce CO₂ emissions, while maintaining the thermal comfort of the user with the flexibility to adjust the operation based on varying energy prices. Nevertheless, further measures such as building renovations are necessary to increase heat pump efficiency and lower operation costs as well as to increase the share of the heat pump, which can then be adapted to the improved operating conditions.