Demand-Based Control Design for Efficient Heat Pump Operation of Electric Vehicles

Thermal management systems of passenger vehicles are fundamental to provide adequate cabin thermal comfort. However, for battery electric vehicles they can use a significant amount of battery energy and thus reduce the real driving range. Indeed, when heating or cooling the vehicle cabin the thermal management system can consume up to 84% of the battery capacity. This study proposes a model-based approach to design an energy-efficient control strategy for heating electric vehicles, considering the entire climate control system at different ambient conditions. Specifically, the study aims at reducing the energy demand of the compressor and water pumps when operating in heat pump mode. At this scope, the climate control system of the reference vehicle is modelled and validated, enabling a system efficiency analysis in different operating points. Based on the system performance assessment, the optimized operating strategy for the compressor and the water pumps is elaborated and the results show that the demand-based control achieves up to 34% energy reduction when compared to the standard control.


Background and Problem
In recent years, electric mobility is becoming more popular and this is underlined by the fact that all big companies in the automotive industry have electric vehicles in their portfolio [1]. However, one big challenge of battery electric vehicles (BEVs) is represented by the driving range limitation due to the battery capacity constraints [2,3]. During the past years there have been many scientific studies and publications dealing with driving range anxiety [4,5], trying to overcome these limitations.

Previous Research in the Field
Thermal management is essential for ensuring adequate thermal comfort in passenger vehicles. Shete and Farrington et al. (respectively, in [6,7]) studied the negative impact of the air conditioning systems for internal combustion engine (ICE) vehicles on fuel consumption and CO 2 emissions. One major disadvantage related to the climate control system, also called heating, ventilation, and air conditioning (HVAC) system later in the paper, of BEVs compared to conventional ICE vehicles is the unavailability of waste heat from the engine for heating purposes. Instead, BEVs must drain the energy from the battery for heating, thus impacting negatively the vehicle range [8]. Indeed, the importance of thermal management optimization is highlighted by the fact that heating in cold winter conditions or cooling in hot summer conditions can increase the energy consumption from the HVAC system up to 84%, resulting in a significant reduction of the maximum driving range of the vehicle up to

Current Study Novelties and Focus
This paper follows a model-based approach to optimize the operating strategy of the compressor and the water pumps of a heat pump system for EVs. The improvements are achieved by a demand-based design of the control strategy where the components are optimally operated in part load conditions. The abovementioned studies from Kitanoski and Hofer [25], De Nunzio et al. [27], and Hosoz and Direk [28] deal with similar topics like this study. However, this study differs from the existing studies in two key aspects. On the one hand, the paper investigates the optimal demand-based control of compressor and water pumps for various operating points, deriving the optimal strategy at different ambient conditions. On the other hand, the optimal operating points are analyzed using the developed simulation models which include not only the refrigerant cycle but also the entire HVAC system Energies 2020, 13, 5440 3 of 18 components. The optimal strategy is obtained considering the performance of the entire system by varying the target air outlet temperature of the HVAC system and the pump speeds.
To validate the proposed approach, the study compares in terms of energy consumption the standard control strategy scenario with three incremental improvement scenarios based on three control mechanisms and quantifies the energy-saving potential at different ambient conditions.

Methodology
The optimal demand-based control has been developed and tested in a virtual environment based on a parametric variation. Indeed, compared to the real-life testing the model-based approach reduces the development time for optimizing the controls of the proposed components.
The system model used for the optimization process includes the HVAC model, the 1D thermal cabin model, and the control system as shown in Figure 1. The details of the sub-models are given in Sections 2.1 and 2.2. The models are implemented in Dymola/Modelica using models from the Modelica Standard Library [32] and from the TIL Suite [33]. The parameters for each of the sub-models are derived either from the available measurement data (see Section 2.3) or from technical datasheets. The model validation results are provided in Section 3.1.
Energies 2020, 13, x FOR PEER REVIEW 3 of 18 system components. The optimal strategy is obtained considering the performance of the entire system by varying the target air outlet temperature of the HVAC system and the pump speeds.
To validate the proposed approach, the study compares in terms of energy consumption the standard control strategy scenario with three incremental improvement scenarios based on three control mechanisms and quantifies the energy-saving potential at different ambient conditions.

Methodology
The optimal demand-based control has been developed and tested in a virtual environment based on a parametric variation. Indeed, compared to the real-life testing the model-based approach reduces the development time for optimizing the controls of the proposed components.
The system model used for the optimization process includes the HVAC model, the 1D thermal cabin model, and the control system as shown in Figure 1. The details of the sub-models are given in sections 2.1 and 2.2. The models are implemented in Dymola/Modelica using models from the Modelica Standard Library [32] and from the TIL Suite [33]. The parameters for each of the submodels are derived either from the available measurement data (see section 2.3) or from technical datasheets. The model validation results are provided in section 3.1. Reference scenario (Control Base),  1st improvement step: load controlled-compressor scenario with demand-dependent air-outlettemperature setpoint of the HVAC system (Control Step 1),  2nd improvement step: load controlled-water-pumps scenario with demand-dependent speed of the water pumps (Control Step 2),  3rd improvement step: combination of steps 1 and 2 (Control Step 3).
The proposed methodology, represented in Figure 2, develops an enhanced and efficient operating strategy for the compressor and the two water pumps in a virtual environment. Thereby, it offers a reduced development time when compared to real-life testing and it proposes how to determine the optimum load setpoints of the proposed components.
The methodology includes the following steps: 1. System identification and modelling in an iterative process: models and measured data from the real-world system have been used for the parametrization; 2. HVAC-compressor and water-pumps controller analysis and synthesis: identification of the dynamic characteristics of the system model and definition of an appropriate controller; 3. Ambient temperature variation: simulation allows the investigation under various conditions to find the optimal setpoints; 4. Determination and extrapolation of the demand-based optimal operating points of the HVAC compressor and the water pumps at various boundary conditions; Reference scenario (Control Base), • 1st improvement step: load controlled-compressor scenario with demand-dependent air-outlet-temperature setpoint of the HVAC system (Control Step 1), • 2nd improvement step: load controlled-water-pumps scenario with demand-dependent speed of the water pumps (Control Step 2), • 3rd improvement step: combination of steps 1 and 2 (Control Step 3).
The proposed methodology, represented in Figure 2, develops an enhanced and efficient operating strategy for the compressor and the two water pumps in a virtual environment. Thereby, it offers a reduced development time when compared to real-life testing and it proposes how to determine the optimum load setpoints of the proposed components.
The methodology includes the following steps: 1. System identification and modelling in an iterative process: models and measured data from the real-world system have been used for the parametrization; 2.
HVAC-compressor and water-pumps controller analysis and synthesis: identification of the dynamic characteristics of the system model and definition of an appropriate controller; 3.
Ambient temperature variation: simulation allows the investigation under various conditions to find the optimal setpoints; 4.
Determination and extrapolation of the demand-based optimal operating points of the HVAC compressor and the water pumps at various boundary conditions;  To find the optimal setpoints at different ambient conditions, a parameter variation is performed. An overview of the boundaries is given in Table 1. Although the climate control system is capable of operating in both heating and cooling mode, the study focuses on heat pump mode, and the assessment is performed at an ambient temperature of -10 °C to 10 °C in steps of 5 °C with the condenser temperature setpoint varied continuously between 22 °C and 70 °C and the speed of both water pumps varied continuously between 5 Hz and 100 Hz. [-10, -5, 0, 5, 10]

2.1.D Thermal Cabin Model
A schematic overview of the cabin model is provided in Figure 3. The 1D thermal cabin model uses an object-oriented approach and includes heat exchange mechanism of conduction and convection to the ambient. Under steady state conditions, the energy equations for car cabin thermal equilibrium can be written as: Qlosses = Qheat = Qtrans + Qven + Qrad + Qmet (1) where Qlosses are the overall heat losses between the cabin and ambient and are balanced with the energy provided by the climate control system (Qheat). The overall heat losses of the cabin can be divided into: , transmission losses through the body of the cabin; To find the optimal setpoints at different ambient conditions, a parameter variation is performed. An overview of the boundaries is given in Table 1. Although the climate control system is capable of operating in both heating and cooling mode, the study focuses on heat pump mode, and the assessment is performed at an ambient temperature of −10 • C to 10 • C in steps of 5 • C with the condenser temperature setpoint varied continuously between 22 • C and 70 • C and the speed of both water pumps varied continuously between 5 Hz and 100 Hz.

1D Thermal Cabin Model
A schematic overview of the cabin model is provided in Figure 3. The 1D thermal cabin model uses an object-oriented approach and includes heat exchange mechanism of conduction and convection to the ambient. Under steady state conditions, the energy equations for car cabin thermal equilibrium can be written as: Q losses = Q heat = Q trans + Q ven + Q rad + Q met (1) where Q losses are the overall heat losses between the cabin and ambient and are balanced with the energy provided by the climate control system (Q heat ). The overall heat losses of the cabin can be divided into: The temperature of the cabin, calculated in the heat capacity of the cabin air (cpAir), is therefore the result of the energy flow between indoor and ambient air through conduction and convection phenomena as reported graphically in Figure 3. The presented parameters have been estimated from measurement-and material data. Conduction, convection, and heat capacity models are taken from the Modelica Standard Library. Additionally, for the realistic behavior of the system, the ventilation ducts have been considered and modelled including heat losses through conduction and convection and heat capacity of the duct itself. Specifically, the air path (orange) starts at the inlet and leads through the ducts to the cabin-air volume and finally to the outlet. The thermal connections of the model are represented by the red lines. From the inside wall surfaces of the ducts, thermal energy is conducted through the wall to a heat capacity (thermal point mass). From there, the thermal energy is also dissipated to the ambient via convective heat transfer. The air within the cabin is convectively transferred to the heat capacity The components of the energy equations (1) can be written as: Q ven = m in c pAir (T vent − T amb ) where: The temperature of the cabin, calculated in the heat capacity of the cabin air (c pAir ), is therefore the result of the energy flow between indoor and ambient air through conduction and convection phenomena as reported graphically in Figure 3. The presented parameters have been estimated from measurement-and material data. Conduction, convection, and heat capacity models are taken from the Modelica Standard Library. Additionally, for the realistic behavior of the system, the ventilation ducts have been considered and modelled including heat losses through conduction and convection and heat capacity of the duct itself.
Specifically, the air path (orange) starts at the inlet and leads through the ducts to the cabin-air volume and finally to the outlet. The thermal connections of the model are represented by the red lines. From the inside wall surfaces of the ducts, thermal energy is conducted through the wall to a heat Energies 2020, 13, 5440 6 of 18 capacity (thermal point mass). From there, the thermal energy is also dissipated to the ambient via convective heat transfer. The air within the cabin is convectively transferred to the heat capacity of the interior parts, such as the seats or the dashboard. In addition, thermal energy is transferred through the chassis to the ambient. Thereby, the heat flow is determined by convective heat transfer between the cabin inside air and the chassis, thermal conduction through the chassis and finally convective heat transfer between the outside of the chassis and the ambient.

HVAC System and Control System
A schematic overview for the implemented propane-based (R290) HVAC system model is depicted for both cooling mode and heat pump mode in Figure 4. The models are implemented in Dymola/Modelica using models from the TIL Suite. Details of the mathematical models of the single components are reported in [33].
Energies 2020, 13, x FOR PEER REVIEW 6 of 18 of the interior parts, such as the seats or the dashboard. In addition, thermal energy is transferred through the chassis to the ambient. Thereby, the heat flow is determined by convective heat transfer between the cabin inside air and the chassis, thermal conduction through the chassis and finally convective heat transfer between the outside of the chassis and the ambient.

HVAC System and Control System
A schematic overview for the implemented propane-based (R290) HVAC system model is depicted for both cooling mode and heat pump mode in Figure 4. The models are implemented in Dymola/Modelica using models from the TIL Suite. Details of the mathematical models of the single components are reported in [33]. The model is divided into three parts: refrigerant (in green), coolant (in blue), and air cycles (in orange). The refrigerant cycle is based on R290 and considers the compressor, condenser, internal heat exchanger, expansion valve, and evaporator. The coolant cycle is based on water-glycol at a mass-fraction of 50% each and consists of the water side of the condenser, evaporator and front heat exchanger (main radiator), pumps and valves. By switching the coolant cycle valves, the refrigerant cycle can be either used in cooling or in heat pump mode, cp. Figure 4 left and right, respectively. The model is divided into three parts: refrigerant (in green), coolant (in blue), and air cycles (in orange). The refrigerant cycle is based on R290 and considers the compressor, condenser, internal heat exchanger, expansion valve, and evaporator. The coolant cycle is based on water-glycol at a mass-fraction of 50% each and consists of the water side of the condenser, evaporator and front heat exchanger (main radiator), pumps and valves. By switching the coolant cycle valves, the refrigerant cycle can be either used in cooling or in heat pump mode, cp. Figure 4 left and right, respectively. The Energies 2020, 13, 5440 7 of 18 air cycle considers the front heat exchanger, the cabin heat exchanger, the main radiator fan, and cabin fan and a cabin model.
For the evaluation of the performance of the system the following key performance indicators (KPIs) have been considered when comparing the scenarios: • Electrical power consumption P electricalSys [W], as the electrical energy demand of the entire HVAC system (consisting of compressor, water pumps and cabin fan); • Average cabin temperature; • Coefficient of performance (COP and COP sys ), related to the heating mode and determined as: COP sys = Q condenser /P electricalSys (5) with: , thermal energy output of the condenser on the refrigerant side; , the electrical energy of the compressor.
The base control algorithms (Control Base) for the five main components are described in the following section and depicted in Figure 5.  The base control algorithms (Control Base) for the five main components are described in the following section and depicted in Figure 5.

Models Parametrization
In the first step, each single component of the HVAC system, for which measurement data were available, has been parameterized separately. Figure 6 shows the workflow from the monitoring data up to the system model. First, monitoring data were used to analyze the components and to choose the appropriate ones from the available libraries. Second, a comparison between results and monitoring data was used to extract component parameters, with the target of reducing simulation

Models Parametrization
In the first step, each single component of the HVAC system, for which measurement data were available, has been parameterized separately. Figure 6 shows the workflow from the monitoring data up to the system model. First, monitoring data were used to analyze the components and to choose the appropriate ones from the available libraries. Second, a comparison between results and monitoring

Demand-Based Control Design
Although the implemented HVAC system model is capable of simulating both cooling and heat pump mode, this study focuses only on heat pump mode. At lower ambient temperatures, the HVAC system must provide higher heating power to compensate the difference between ambient temperature and target cabin air temperature. Therefore, the heating demand of the heat pump system increases with a decrease in ambient temperature. This, in turn, means that the design of the demand-based control for the HVAC system needs to consider the dependency of the ambient temperature on the heating demand.
The developed system model has been used to investigate the behavior of the HVAC system in further operating points which have not been measured. The main advantages of the proposed model-based approach are therefore related to the development-time efficiency (simulations are largely faster than measurements) and the safety aspects (hardware and software modifications are applied to models which, when working outside their operating range, cannot get damaged). The analysis for the demand-based control design has been performed for three different components (see Figure 4) in three consecutive steps: The base control strategy of the main components of the HVAC system, which has been described in chapter 2.2, serves as the reference scenario and is summarized in Table 2. Thereby, "PI" means that the quantity is controlled by a PI controller and "f(x)" expresses a dependency upon the variable "x". As before the optimization the system performance is not known entirely, TinTarget was chosen high enough to guarantee appropriate heat-up and the pumps run at higher speeds than necessary, to be on the safe side. In order to determine the demand-dependent optimum control of the components, TinTarget and the speed of the pump on the condenser side (npumpCond) and on the evaporator side (npumpEvap) have been varied using ramp profiles at different ambient temperatures. The simulation is terminated either after the ramps have reached their final values at 10000 s or when the target cabin temperature of 22 °C can no longer be assured (i.e. when Tcab falls below 22 °C in steady state). The simulations have been repeated for different ambient temperatures, ranging from -10 °C to 10 °C .

Component
Quantity Control Base

Control
Step 1

Control
Step 2

Demand-Based Control Design
Although the implemented HVAC system model is capable of simulating both cooling and heat pump mode, this study focuses only on heat pump mode. At lower ambient temperatures, the HVAC system must provide higher heating power to compensate the difference between ambient temperature and target cabin air temperature. Therefore, the heating demand of the heat pump system increases with a decrease in ambient temperature. This, in turn, means that the design of the demand-based control for the HVAC system needs to consider the dependency of the ambient temperature on the heating demand.
The developed system model has been used to investigate the behavior of the HVAC system in further operating points which have not been measured. The main advantages of the proposed model-based approach are therefore related to the development-time efficiency (simulations are largely faster than measurements) and the safety aspects (hardware and software modifications are applied to models which, when working outside their operating range, cannot get damaged). The analysis for the demand-based control design has been performed for three different components (see Figure 4) in three consecutive steps: The base control strategy of the main components of the HVAC system, which has been described in chapter 2.2, serves as the reference scenario and is summarized in Table 2. Thereby, "PI" means that the quantity is controlled by a PI controller and "f(x)" expresses a dependency upon the variable "x". As before the optimization the system performance is not known entirely, T inTarget was chosen high enough to guarantee appropriate heat-up and the pumps run at higher speeds than necessary, to be on the safe side. In order to determine the demand-dependent optimum control of the components, T inTarget and the speed of the pump on the condenser side (n pumpCond ) and on the evaporator side (n pumpEvap ) have been varied using ramp profiles at different ambient temperatures. The simulation is terminated either after the ramps have reached their final values at 10000 s or when the target cabin temperature of 22 • C can no longer be assured (i.e. when T cab falls below 22 • C in steady state). The simulations have been repeated for different ambient temperatures, ranging from −10 • C to 10 • C. Speed Opening Speed As the compressor controls T inTarget , the operation of the compressor can be optimized by adapting this value. Therefore, in Control Step 1 of the demand-based control design, a ramp starting from 70 • C and decreasing to 22 • C within 10,000 s has been used to find the optimum value for T inTarget at different ambient temperatures. In Control Step 2, n pumpCond and n pumpEvap are varied, respectively, using a ramp profile changing from 100 Hz to 5 Hz within 10,000 s. The optimum T inTarget , n pumpCond , and n pumpEvap for each of the analyzed ambient temperatures have been stored in a lookup table. Control Step 3 combines the results of Control Steps 1 and 2. Table 3 summarizes the ramp parameters that have been used for parameter variation. The target of the variation study is to find the operating point for the three components at different T amb where the target cabin temperature of 22 • C can still be kept and where the electrical power consumption of the HVAC system is at a minimum. Table 2 summarizes the component control strategies and dependencies of the main components for the different steps of the demand-based system analysis. The results and analysis of the variation simulation is reported in chapter 3.2.

Results
This section describes the results of the study and consists of three parts. First, the validation of the developed simulation models is reported. Second, the results for the demand-based control strategy, which has been elaborated using the above-mentioned methodology, is described. Finally, the performance of a representative HVAC system using the developed control strategies (Control Steps 1, 2 and 3) is analyzed and compared to the reference case using a non-optimized control strategy (Control Base).

Models Validation
The 1D thermal cabin model, developed within the framework of the QUIET project [34], has been validated against measurement data in heating mode operation. The available measurement dataset corresponds to the heat-up phase with the maximum power at an ambient temperature of 20 • C. The cabin fan was at maximum level and the provided air temperature at the cabin heat exchanger was 55 • C. Figure 7 shows the comparison of the measurements (dashed) and the simulation (solid). The results illustrate that the 1D thermal cabin model can accurately reproduce the air temperature within the cabin. Indeed, the cabin temperature prediction at steady state conditions (after 4000 s) has an error of 0.44 • C or 1%. Moreover, the transient behavior of the simulated cabin temperature is predicted well with a standard deviation of 1.55 • C or 5.3%. Subsequently, the total refrigerant cycle has been validated against measurements for one representative operating point, e.g., at an ambient temperature of -10 °C . The validation results of the refrigerant-cycle model on the pressure-enthalpy (p,h) diagram can be seen in Figure 8 where the grey line represents the saturation line of propane, the dashed line the measurements, and the solid line the corresponding simulation results. In the investigated operating point of the validation case, the compressor model has a deviation in the efficiency, resulting in a maximum enthalpy error of about 19 kJ/kg or 2.4% compared to the measurements. However, they represent only minor deviations when considering the complexity of the HVAC system, i.e., the simulation results are in good accordance with the measurements for the operational point under consideration.

Demand-Based Control Strategy
The power consumption of the HVAC system at different ambient temperatures and Tin is represented in Figure 9. The crosses mark the operating points with the lowest energy consumption in each scenario. At high cabin inlet temperatures, the compressor needs to provide high heating power. Starting from the highest temperature (70 °C ), the trends of the lines show that the power Subsequently, the total refrigerant cycle has been validated against measurements for one representative operating point, e.g., at an ambient temperature of −10 • C. The validation results of the refrigerant-cycle model on the pressure-enthalpy (p,h) diagram can be seen in Figure 8 where the grey line represents the saturation line of propane, the dashed line the measurements, and the solid line the corresponding simulation results. In the investigated operating point of the validation case, the compressor model has a deviation in the efficiency, resulting in a maximum enthalpy error of about 19 kJ/kg or 2.4% compared to the measurements. However, they represent only minor deviations when considering the complexity of the HVAC system, i.e., the simulation results are in good accordance with the measurements for the operational point under consideration. Subsequently, the total refrigerant cycle has been validated against measurements for one representative operating point, e.g., at an ambient temperature of -10 °C . The validation results of the refrigerant-cycle model on the pressure-enthalpy (p,h) diagram can be seen in Figure 8 where the grey line represents the saturation line of propane, the dashed line the measurements, and the solid line the corresponding simulation results. In the investigated operating point of the validation case, the compressor model has a deviation in the efficiency, resulting in a maximum enthalpy error of about 19 kJ/kg or 2.4% compared to the measurements. However, they represent only minor deviations when considering the complexity of the HVAC system, i.e., the simulation results are in good accordance with the measurements for the operational point under consideration.

Demand-Based Control Strategy
The power consumption of the HVAC system at different ambient temperatures and Tin is represented in Figure 9. The crosses mark the operating points with the lowest energy consumption in each scenario. At high cabin inlet temperatures, the compressor needs to provide high heating power. Starting from the highest temperature (70 °C ), the trends of the lines show that the power

Demand-Based Control Strategy
The power consumption of the HVAC system at different ambient temperatures and T in is represented in Figure 9. The crosses mark the operating points with the lowest energy consumption in each scenario. At high cabin inlet temperatures, the compressor needs to provide high heating power. Starting from the highest temperature (70 • C), the trends of the lines show that the power consumption first decreases with reduced cabin inlet temperature because the compressor needs to provide high heating capacity. When T in is further decreased, at a specific point the power consumption starts to rise again. This can be explained by the fact that the controller of the cabin fan tries to keep the cabin temperature at the setpoint, which is 22 • C. When T in decreases, the fan must provide a higher air mass flow, in order to compensate the lower air temperature and therefore the power consumption starts to increase. Additionally, as the system operates in fresh-air mode, the amount of air that is blown out of the vehicle is increased with higher fan levels, which in turn leads to higher ventilation losses and therefore power consumption.
Energies 2020, 13, x FOR PEER REVIEW 11 of 18 consumption first decreases with reduced cabin inlet temperature because the compressor needs to provide high heating capacity. When Tin is further decreased, at a specific point the power consumption starts to rise again. This can be explained by the fact that the controller of the cabin fan tries to keep the cabin temperature at the setpoint, which is 22 °C . When Tin decreases, the fan must provide a higher air mass flow, in order to compensate the lower air temperature and therefore the power consumption starts to increase. Additionally, as the system operates in fresh-air mode, the amount of air that is blown out of the vehicle is increased with higher fan levels, which in turn leads to higher ventilation losses and therefore power consumption.  Figure 10 shows the condenser water pump speed optimization with a similar trend comparable to the compressor. For high pump speeds, the power consumption of the entire HVAC system is high. Toward lower pump speeds the power consumption decreases until it reaches an optimum operating point. Then, when further decreasing the pump speed, the overall power consumption rises again. At very low pump speeds, the slope of the power consumption significantly increases. This can be explained by the fact that the mass flow in the water cycle gets too low to transmit the thermal energy from the refrigerant cycle to the cabin heat exchanger and thus the target inlet air temperature cannot be reached. Consequently, the PI controller of the compressor increases the compressor speed. At the same time, the PI controller of the cabin fan increases the fan speed to compensate the reduced temperature at the air inlet, thus increasing also the ventilation losses.  Figure 10 shows the condenser water pump speed optimization with a similar trend comparable to the compressor. For high pump speeds, the power consumption of the entire HVAC system is high. Toward lower pump speeds the power consumption decreases until it reaches an optimum operating point. Then, when further decreasing the pump speed, the overall power consumption rises again. At very low pump speeds, the slope of the power consumption significantly increases. This can be explained by the fact that the mass flow in the water cycle gets too low to transmit the thermal energy from the refrigerant cycle to the cabin heat exchanger and thus the target inlet air temperature cannot be reached. Consequently, the PI controller of the compressor increases the compressor speed. At the same time, the PI controller of the cabin fan increases the fan speed to compensate the reduced temperature at the air inlet, thus increasing also the ventilation losses. Energies 2020, 13, x FOR PEER REVIEW 12 of 18 Figure 10. Optimum condenser pump speed at different ambient temperatures. Figure 11 compares the power consumption of the HVAC system when varying the speed of the water pump on evaporator side at different ambient temperatures. The charts show a similar trend like for the water pump on condenser side but with slightly flatter slope, especially at lower pump speeds.  Table 4 summarizes the results of the study. As the study showed, the heating demand of the HVAC system depends strongly on the ambient temperature. Therefore, the demand-based control (Control Step 3) replaces the constant control values from the base control (Control Base) with values, which depend on Tamb, to achieve better efficiency. Table 4. Base control compared to demand-based control of the HVAC system.   Figure 11 compares the power consumption of the HVAC system when varying the speed of the water pump on evaporator side at different ambient temperatures. The charts show a similar trend like for the water pump on condenser side but with slightly flatter slope, especially at lower pump speeds.  Table 4 summarizes the results of the study. As the study showed, the heating demand of the HVAC system depends strongly on the ambient temperature. Therefore, the demand-based control (Control Step 3) replaces the constant control values from the base control (Control Base) with values, which depend on Tamb, to achieve better efficiency. Table 4. Base control compared to demand-based control of the HVAC system.  Table 4 summarizes the results of the study. As the study showed, the heating demand of the HVAC system depends strongly on the ambient temperature. Therefore, the demand-based control (Control Step 3) replaces the constant control values from the base control (Control Base) with values, which depend on T amb , to achieve better efficiency.  Figure 12 presents the proposed control functions for the demand-based control. As already mentioned earlier, the heating demand for the HVAC system strongly depends on the ambient temperature. Therefore, the optimum operating points, which have been found above (see crosses in Figure 12), are correlated with the ambient temperature. A clear linear dependency between all three quantities (T in , n pumpCond and n pumpEvap ) and T amb can be seen. Therefore, a linear function has been fit through the extracted operating points. The same function has been used to extrapolate the functions also down to -15 • C and up to 20 • C ambient temperature.  Figure 12 presents the proposed control functions for the demand-based control. As already mentioned earlier, the heating demand for the HVAC system strongly depends on the ambient temperature. Therefore, the optimum operating points, which have been found above (see crosses in Figure 12), are correlated with the ambient temperature. A clear linear dependency between all three quantities (Tin, npumpCond and npumpEvap) and Tamb can be seen. Therefore, a linear function has been fit through the extracted operating points. The same function has been used to extrapolate the functions also down to -15 °C and up to 20 °C ambient temperature.

Use Case Results
The impact of the demand-based control of the HVAC system has been evaluated at different ambient temperatures. Figure 13 compares the COP of the climate control system for the various scenarios. As described in (4), the COP represents the efficiency with respect to only the compressor power consumption. In this case Figure 13 shows that, as expected, the efficiency increases with increasing ambient temperature, since the temperature lift between the source (evaporator) and the sink (condenser) is reduced. However, for a correct evaluation of the impact of the control strategies it is required to compare them related to the entire HVAC system and not only limiting the evaluation to the refrigerant cycle.

Use Case Results
The impact of the demand-based control of the HVAC system has been evaluated at different ambient temperatures. Figure 13 compares the COP of the climate control system for the various scenarios. As described in (4), the COP represents the efficiency with respect to only the compressor power consumption. In this case Figure 13 shows that, as expected, the efficiency increases with increasing ambient temperature, since the temperature lift between the source (evaporator) and the sink (condenser) is reduced. However, for a correct evaluation of the impact of the control strategies it is required to compare them related to the entire HVAC system and not only limiting the evaluation to the refrigerant cycle. Energies 2020, 13, x FOR PEER REVIEW 14 of 18 Figure 13. Evaluation of the COP of the optimization steps.
Indeed, Figure 14 presents the results for the COPsys, based on (5). As it can be seen, the total system efficiency of the base scenario (Control Base) is largely influenced by the effect of the water pumps and the cabin fan and the COPsys decreases when increasing the ambient temperature. In contrast, the total system efficiency of the optimized scenario (Control Step 3) could be significantly increased by up to 81% by applying the demand-based control, especially at mild ambient conditions.   Indeed, Figure 14 presents the results for the COP sys , based on (5). As it can be seen, the total system efficiency of the base scenario (Control Base) is largely influenced by the effect of the water pumps and the cabin fan and the COP sys decreases when increasing the ambient temperature. In contrast, the total system efficiency of the optimized scenario (Control Step 3) could be significantly increased by up to 81% by applying the demand-based control, especially at mild ambient conditions. Indeed, Figure 14 presents the results for the COPsys, based on (5). As it can be seen, the total system efficiency of the base scenario (Control Base) is largely influenced by the effect of the water pumps and the cabin fan and the COPsys decreases when increasing the ambient temperature. In contrast, the total system efficiency of the optimized scenario (Control Step 3) could be significantly increased by up to 81% by applying the demand-based control, especially at mild ambient conditions.     . highlights the climate control system savings for the various scenarios at different ambient temperatures. As for both the COPsys and the electric load, savings are more pronounced with higher ambient temperatures. Indeed, the energy savings at an ambient temperature of -10 °C are respectively for the Control Steps 1, 2, and 3 of 1.5%, 6.3%, and 7.7% and at an ambient temperature of 10 °C they are of 13.0%, 20.6%, and 33.6%. Additionally, to validate consistently the developed demand-based control design, the optimal control strategy (Control Step 3) has been tested under transient behavior. Indeed, Figure 17 shows the behavior during the first 30 minutes during the heat-up phase starting from various ambient temperatures. For both ambient temperature of -10 °C and -5 °C the time to reach the setpoint is unchanged. For temperatures above 0 °C a trade-off between heat-up time and energy efficiency must be found, as at 10 °C the heat-up time increases by 256 s, while at the same time achieving a considerable efficiency increase of 81% and energy savings up to 33.6%.  Figure 16 highlights the climate control system savings for the various scenarios at different ambient temperatures. As for both the COP sys and the electric load, savings are more pronounced with higher ambient temperatures. Indeed, the energy savings at an ambient temperature of −10 • C are respectively for the Control Steps 1, 2, and 3 of 1.5%, 6.3%, and 7.7% and at an ambient temperature of 10 • C they are of 13.0%, 20.6%, and 33.6%.  . highlights the climate control system savings for the various scenarios at different ambient temperatures. As for both the COPsys and the electric load, savings are more pronounced with higher ambient temperatures. Indeed, the energy savings at an ambient temperature of -10 °C are respectively for the Control Steps 1, 2, and 3 of 1.5%, 6.3%, and 7.7% and at an ambient temperature of 10 °C they are of 13.0%, 20.6%, and 33.6%. Additionally, to validate consistently the developed demand-based control design, the optimal control strategy (Control Step 3) has been tested under transient behavior. Indeed, Figure 17 shows the behavior during the first 30 minutes during the heat-up phase starting from various ambient temperatures. For both ambient temperature of -10 °C and -5 °C the time to reach the setpoint is unchanged. For temperatures above 0 °C a trade-off between heat-up time and energy efficiency must be found, as at 10 °C the heat-up time increases by 256 s, while at the same time achieving a considerable efficiency increase of 81% and energy savings up to 33.6%. Additionally, to validate consistently the developed demand-based control design, the optimal control strategy (Control Step 3) has been tested under transient behavior. Indeed, Figure 17 shows the behavior during the first 30 minutes during the heat-up phase starting from various ambient temperatures. For both ambient temperature of −10 • C and −5 • C the time to reach the setpoint is unchanged. For temperatures above 0 • C a trade-off between heat-up time and energy efficiency must be found, as at 10 • C the heat-up time increases by 256 s, while at the same time achieving a considerable efficiency increase of 81% and energy savings up to 33.6%. Energies 2020, 13, x FOR PEER REVIEW 16 of 18 Figure 17. Transient cabin temperature for base-and optimized control.

Discussion and Conclusions
This paper shows the applicability of a workflow for using the simulation model of an entire HVAC system to design alternative demand-based operating strategies of the compressor and water pumps toward energy efficient operation. Models have been parameterized based on measurement data on a component level and have used to build up the entire HVAC system model. The validation showed that the HVAC system model can predict accurately the cabin temperature and the HVAC system performance. Finally, the HVAC system model has been used to analyze the efficiency in different operating points (various ambient temperatures) and to derive the optimal controls for operating the compressor and both water pumps with regard to the current heating demand at the maximum efficiency. The study results confirmed that the total energy consumption of the compressor and the pumps can be significantly reduced by up to 34% when adopting the proposed demand-based control strategy.
Summarizing, the proposed demand-based optimal control strategy shows how a relatively simple modification on the software-side (usage of the outdoor temperature value in the existing controllers to define the compressor and water pumps setpoints) can significantly improve the efficiency of the entire system. In that scope, the approach requires a reliable simulation system model, which will play a key role for future research as the energy consumption of the compressor and the pumps can be reduced with minor modifications and shorter development time. That would allow the proposed approach as well to be extended to existing vehicles using a comparable HVAC system set-up. In addition, the workflow could be expanded to include a broader ambient condition range and/or supplementary components to enhance furthermore the HVAC system operating strategy.

Discussion and Conclusions
This paper shows the applicability of a workflow for using the simulation model of an entire HVAC system to design alternative demand-based operating strategies of the compressor and water pumps toward energy efficient operation. Models have been parameterized based on measurement data on a component level and have used to build up the entire HVAC system model. The validation showed that the HVAC system model can predict accurately the cabin temperature and the HVAC system performance. Finally, the HVAC system model has been used to analyze the efficiency in different operating points (various ambient temperatures) and to derive the optimal controls for operating the compressor and both water pumps with regard to the current heating demand at the maximum efficiency. The study results confirmed that the total energy consumption of the compressor and the pumps can be significantly reduced by up to 34% when adopting the proposed demand-based control strategy.
Summarizing, the proposed demand-based optimal control strategy shows how a relatively simple modification on the software-side (usage of the outdoor temperature value in the existing controllers to define the compressor and water pumps setpoints) can significantly improve the efficiency of the entire system. In that scope, the approach requires a reliable simulation system model, which will play a key role for future research as the energy consumption of the compressor and the pumps can be reduced with minor modifications and shorter development time. That would allow the proposed approach as well to be extended to existing vehicles using a comparable HVAC system set-up. In addition, the workflow could be expanded to include a broader ambient condition range and/or supplementary components to enhance furthermore the HVAC system operating strategy.