Documentation for the creation and usage of the heat pump library (hplib)

This documentation covers the database preperation, validation and usage for simulation. If you're only interested in using hplib for simulation purpose, you should have a look into chapter 4. How to simulate.

1. Definitions
2. Database preperation
3. Work with database
   1. Load database
   2. Load specific model
   3. Load generic model
4. How to simulate
   1. Simulate on time step
   2. Simulate a time series
5. Example heat pump
6. Validation
   1. Air/Water | on/off
   2. Brine/Water | on/off
   3. Water/Water | on/off
   4. Air/Water | regulated
      1. Heating
      2. Cooling
   5. Brine/Water | regulated
7. Conclusion

import hplib as hpl
import hplib_database as db
import pandas as pd
import matplotlib.pyplot as plt
import warnings
import numpy
warnings.filterwarnings("ignore")

%%html
<style>
table {float:left}
</style>

1. Definitions

Abbreviations

Abbreviation	Meaning
P_th	Thermal output power in W
P_el	Electical input Power in W
COP	Coefficient of performance
EER	Energy Efficiency Ratio
T_in	Input temperature in °C at primary side of the heat pump
T_out	Output temperature in °C at secondary side of the heat pump
T_amb	Ambient temperature in °C
P_th_h_ref	Thermal heating output power in W at T_in = -7 °C and T_out = 52 °C
P_el_h_ref	Elecrical input power (heating) in W at T_in = -7 °C and T_out = 52 °C
COP_ref	Coefficient of performance at T_in = -7 °C and T_out = 52 °C
P_th_c_ref	Thermal cooling output power in W at T_in = 35 °C and T_out = 7 °C
P_el_c_ref	Elecrical input power (cooling) in W at T_in = 35 °C and T_out = 7 °C
p1-p4	Fit-Parameters for Fit-Function

Group IDs

Group ID	Type	Subtype
1	Outdoor Air / Water	Regulated
2	Brine / Water	Regulated
3	Water / Water	Regulated
4	Outdoor Air / Water	On-Off
5	Brine / Water	On-Off
6	Water / Water	On-Off

2. Database preparation

This section is only for documentation regarding the development of the final database. It's not neccesary to run this code again.

1. we downloaded all manufacturer data from https://keymark.eu/en/products/heatpumps/certified-products .
2. then we unzipped the files to the input folder and used the bash-skript pdf2txt.sh to convert pdf into txt.
3. afterwards we used the following functions to create and extent the heatpump keymark database.

# Import keymark data and save to csv database
db.import_heating_data()
# -> this creates /output/database_heating.csv
db.import_cooling_data()
# -> this creates /output/database_cooling.csv

# Reduce to climate measurement series with average climate, delete redundant entries and save to csv sub-database
db.reduce_heating_data('database_heating.csv','average')
# -> this creates /output/database_heating_average.csv

Process heating database

# Normalize electrical and thermal power from the keymark database to values from setpoint T_in = -7 °C and T_out = 52 °C
db.normalize_heating_data('database_heating_average.csv')
# -> this creates /output/database_heating_average_normalized.csv

# Identify subtypes like on-off or regulated heat pumps and assign group id depending on its type and subtype
db.identify_subtypes('database_heating_average_normalized.csv')
# -> this creates /output/database_heating_average_normalized_subtypes.csv

# Calculate parameters p1-p4 for P_th, P_el and COP with a least-square fit approach
# based on  K. Schwamberger: „Modellbildung und Regelung von Gebäudeheizungsanlagen 
# mit Wärmepumpen“, VDI Verlag, Düsseldorf, Fortschrittsberichte VDI Reihe 6 Nr. 263, 1991.
db.calculate_heating_parameters('database_heating_average_normalized_subtypes.csv')
# -> this creates /output/hplib_database_heating.csv
# -> this creates /src/hplib_database.csv

# Many heat pump models have redundant entries because of different controller or storage configurations.
# Reduce to unique heat pump models.
db.reduce_to_unique()
# -> this overwrites the /output/hplib_database_heating.csv

# Calculate the relative error (RE) for each set point of every heat pump
db.validation_relative_error_heating()
# -> this creates /output/database_heating_average_normalized_subtypes_validation.csv

# Calculate the mean absolute percentage error (MAPE) for every heat pump
db.validation_mape_heating()
# -> this overwrites the /output/hplib_database_heating.csv

Process cooling database

Overall there are not so many unique Keymark heat pumps for cooling (34 models) in comparison to heating (505 models).

Out of the 34 models only 4 heat pumps had set points at different outflow temperature. With our fit method it is not possible to fit only over one outflow temperature. For that reason we added another set point at 18°C output temperature based on the heat pumps we had with this condition in Keymark. For that purpose, we identified multiplication factors for eletrical power and eer between 7°C and 18°C secondary output temperature. The mean value of that are used to calculate the electrical power and EER at 18°C for other heat pumps:

P_el at 18°C = P_el at 7°C * multiplication factor 
EER at 18°C = EER at 7°C * multiplication factor

Outside Tempertature	Multiplication factor for P_el	Multiplication factor for EER
35	0.85	1.21
30	0.82	1.21
25	0.77	1.20
20	0.63	0.95

#Only use heatpumps which are unique and also in the heating library
db.reduce_cooling_data()
#this generates /output/database_cooling_reduced.csv

# Normalize electrical and thermal power from the keymark database to values from setpoint T_outside = 35 °C and T_out = 7 °C
db.normalize_and_add_cooling_data()
#-> this creates /output/database_cooling_reduced_normalized.csv

# Used the same fit method like in heating
# except: for P_el the point 20/7 is ignored
db.calculate_cooling_parameters()
# -> this overwrites /src/hplib_database.csv

# Calculate the relative error (RE) for each set point of every heat pump
db.validation_relative_error_cooling()
# this creates /output/database_cooling_reduced_normalized_subtypes_validation.csv

# Calculate the mean absolute percentage error (MAPE) for every heat pump
db.validation_mape_cooling()
# -> this overwrites the /src/hplib_database.csv

Create generic heat pump models

# Calculate generic heat pump models for each group id
# for cooling: there is only a generic heat pump of type air/water | regulated available
db.add_generic()
# -> this overwrites the /src/hplib_database.csv

Hint: The csv files in the output folder are for documentation and validation purpose. The code hplib.py and database hplib_database files, which are meant to be used for simulations, are located in the src folder.

3. Work with database

3.1 Load database

Simply execute the command without arguments and you will get a DataFrame with the complete list of manufacturers and models. Now you are able to view, filter or sort the database.

database = hpl.load_database()
database

	Manufacturer	Model	Date	Type	Subtype	Group	Refrigerant	Mass of Refrigerant [kg]	SPL indoor [dBA]	SPL outdoor [dBA]	...	p2_P_el_c [1/°C]	p3_P_el_c [-]	p4_P_el_c [1/°C]	p1_EER [-]	p2_EER [-]	p3_EER [-]	p4_EER [-]	MAPE_P_el_cooling	MAPE_EER	MAPE_Pdc
0	Advantix	i-SHWAK V4 06	2020-05-26	Outdoor Air/Water	Regulated	1.0	R410a	2.68	35.0	64.0	...	-0.010769	-0.763316	-70.588606	37.392760	0.046170	6.295339	-37.488343	12.221419	6.840097	9.194424
1	Advantix	i-SHWAK V4 08	2020-05-26	Outdoor Air/Water	Regulated	1.0	R410a	2.20	35.0	64.0	...	-0.010295	-1.008835	-70.584924	53.317794	0.045359	6.373991	-53.417576	8.557697	5.543797	6.180972
2	Advantix	i-SHWAK V4 10	2020-05-26	Outdoor Air/Water	Regulated	1.0	R410a	3.45	39.0	64.0	...	-0.009974	-1.070117	-71.138737	-38.305956	0.055509	10.401025	38.096062	12.129224	4.233020	12.245195
3	Advantix	i-SHWAK V4 12	2020-05-26	Outdoor Air/Water	Regulated	1.0	R410a	3.45	39.0	65.0	...	-0.009670	-1.199827	-79.730152	-106.597525	0.055461	10.417088	106.386968	10.736620	4.251786	11.085364
4	Advantix	i-SHWAK V4 14T	2020-05-26	Outdoor Air/Water	Regulated	1.0	R410a	4.40	40.0	68.0	...	-0.009060	-1.464498	-70.578508	-106.597525	0.055461	10.417088	106.386968	6.954977	4.251786	7.852487
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
506	Generic	Generic	NaN	Brine/Water	Regulated	2.0	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
507	Generic	Generic	NaN	Water/Water	Regulated	3.0	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
508	Generic	Generic	NaN	Outdoor Air/Water	On-Off	4.0	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
509	Generic	Generic	NaN	Brine/Water	On-Off	5.0	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
510	Generic	Generic	NaN	Water/Water	On-Off	6.0	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

511 rows × 47 columns

3.2 Load specific model

To get the parameters of a specific heat pump model, use the get_parameters() method with a specific Model name from the database. You will get a DataFrame with all parameters including the mean absolute percentage errors (MAPE) for this model.

parameters = hpl.get_parameters('i-SHWAK V4 06')
parameters

	Manufacturer	Model	MAPE_COP	MAPE_P_el	MAPE_P_th	P_th_h_ref [W]	P_el_h_ref [W]	COP_ref	Group	p1_P_th [1/°C]	...	p3_Pdc [-]	p4_Pdc [1/°C]	p1_P_el_c [1/°C]	p2_P_el_c [1/°C]	p3_P_el_c [-]	p4_P_el_c [1/°C]	p1_EER [-]	p2_EER [-]	p3_EER [-]	p4_EER [-]
0	Advantix	i-SHWAK V4 06	9.244852	22.607658	24.897666	4500.0	2866.0	1.57	1.0	1.756924	...	0.068769	-70.599887	70.638879	-0.010769	-0.763316	-70.588606	37.39276	0.04617	6.295339	-37.488343

1 rows × 35 columns

3.3 Load generic model

To get the parameters of a generic heat pump model, use the get_parameters() method with the following keyword arguments of a free choosen set point

• model='Generic'
• group_id: 1,2,4,5,6
• t_in: choose primary input temperature in °C
• t_out: choose secondary output temperature in °C
• p_th: choose thermal output power in W

You will get a DataFrame with all parameters for this generic model. For every group id the parameter set is based on the average parameters of all heat pumps of its group with an MAPE of less than 25%.

parameters = hpl.get_parameters(model='Generic', group_id=1, t_in=-7, t_out=52, p_th=10000)
parameters

	Manufacturer	Model	MAPE_COP	MAPE_P_el	MAPE_P_th	P_th_h_ref [W]	P_el_h_ref [W]	COP_ref	Group	p1_P_th [1/°C]	...	p4_Pdc [1/°C]	p1_P_el_c [1/°C]	p2_P_el_c [1/°C]	p3_P_el_c [-]	p4_P_el_c [1/°C]	p1_EER [-]	p2_EER [-]	p3_EER [-]	p4_EER [-]	EER_ref
0	Generic	Generic	NaN	NaN	NaN	10168.785754	7193.433182	1.413621	1.0	69.969927	...	-71.578974	69.044359	-0.009208	-1.368639	-68.976493	-11.722768	0.06484	11.52005	11.474317	3.278166

1 rows × 36 columns

4. How to simulate

With the Fit-Parameters p1-p4 for P_th, P_el and COP it is possible to calculate the results with the following methods:

1. P_th and P_el with Fit-Functions and COP = P_th / P_el or
2. P_th and COP with Fit-Functions and P_el = P_th / COP or
3. P_el and COP with Fut-Functions and P_th = P_el * COP

While the model by Schwarmberger [1] uses the first method, our validation showed, that the third method leads to better results. Therefore we decided to implement this in the simulate definition.

4.1 Simulate one timestep

Please define a primary input temperature (t_in_primary), secondary input temperature (t_in_secondary), ambient / outdoor temperature (t_amb) in °C and the parameters from the previous step. The t_in_secondary is supposed to be heated up by 5 K which then results in output temperature.

Important:

• for air / water heat pumps t_in_primary and t_amb are supposed to be the same values.
• for brine / water or water / water heat pumps t_in_primary and t_amb are independent values.

# Load parameters
parameters = hpl.get_parameters(model='Generic', group_id=1, t_in=-7, t_out=52, p_th=10000)
# Create heat pump object with parameters
heatpump = hpl.HeatPump(parameters)
# Simulate with values
# whereas mode = 1 is for heating and mode = 2 is for cooling
results = heatpump.simulate(t_in_primary=-7, t_in_secondary=47, t_amb=-7, mode=1)
print(pd.DataFrame([results]))

   T_in  T_out  T_amb       COP  EER         P_el     P_th    m_dot
0    -7     52     -7  1.413621    0  7074.033573  10000.0  0.47619

4.2 Simulate a timeseries

The simulation approach could be to do a for loop around the method of simulating one timestep. But it will be faster to use the previous method with arrays as inputs. For usage you are allowed to define pandas.Series or arrays as input values. It is also possible to combine single values and pandas.Series / arrays.

The next example uses a measured timeseries for ambient / outdoor temperature and secondary input temperature of a real heating system to demonstrate the simulation approach. The data represents on year and has a temporal resolution of 1 minute

# Read input file with temperatures
df = pd.read_csv('../input/TestYear.csv')
df['T_amb'] = df['T_in_primary'] # air/water heat pump -> T_amb = T_in_primary
# Simulate with values
# Load parameters
parameters = hpl.get_parameters(model='Generic', group_id=1, t_in=-7, t_out=52, p_th=10000)
# Create heat pump object with parameters
heatpump = hpl.HeatPump(parameters)
# whereas mode = 1 is for heating and mode = 2 is for cooling
results = heatpump.simulate(t_in_primary=df['T_in_primary'].values, t_in_secondary=df['T_in_secondary'].values, t_amb=df['T_amb'].values, mode=1)
# Plot / Print some results
# example for distribution of COPs
results=pd.DataFrame.from_dict(results)
results['COP'].plot.hist(bins=50, title='Distribution of COP values') 
# Calclulate seasonal performance factor (SPF)
SPF = results['P_th'].mean() / results['P_el'].mean() 
print('The seasonal performance factor (SPF) for one year is = '+str(round(SPF,1)))

The seasonal performance factor (SPF) for one year is = 3.9

5. Example heat pump

To get a overview over the different operation conditions, this section plots the electrical and thermal power as well as the COP for all possible primary and secondary input temperatures

HEATING: Schematic plot of COP, P_el and P_th for an generic air/water heat pump: subtype = on/off

# Define Temperatures
T_in_primary = range(-20,31)
T_in_secondary = range(20,56)
T_in = numpy.array([])
T_out = numpy.array([])

# Load parameters of generic air/water | regulated
parameters = hpl.get_parameters('Generic', group_id=4, t_in=-7, t_out=52, p_th = 10000)
heatpump=hpl.HeatPump(parameters)
results=pd.DataFrame()
# Create input series
for t1 in T_in_primary:
    for t2 in T_in_secondary:
        T_in=numpy.append(T_in,t1)
        T_out=numpy.append(T_out,t2)
results=heatpump.simulate(t_in_primary=T_in, t_in_secondary=T_out, t_amb=T_in, mode=1)
results=pd.DataFrame.from_dict(results)

# Plot COP
fig1, ax1 = plt.subplots()
plot = plt.tricontourf(results['T_in'], results['T_out'], results['COP'])
ax1.tricontour(results['T_in'], results['T_out'], results['COP'], colors='k')
fig1.colorbar(plot)
ax1.set_title('COP')
ax1.set_xlabel('Primary input temperature [°C]')
ax1.set_ylabel('Secondary output temperature [°C]')
fig1.show

# Plot electrical input power
fig1, ax1 = plt.subplots()
plot = plt.tricontourf(results['T_in'], results['T_out'], results['P_el'])
ax1.tricontour(results['T_in'], results['T_out'], results['P_el'], colors='k')
fig1.colorbar(plot)
ax1.set_title('Eletrical input power [W]')
ax1.set_xlabel('Primary input temperature [°C]')
ax1.set_ylabel('Secondary output temperature [°C]')
fig1.show

# Plot thermal output power
fig1, ax1 = plt.subplots()
plot = plt.tricontourf(results['T_in'], results['T_out'], results['P_th'])
ax1.tricontour(results['T_in'], results['T_out'], results['P_th'], colors='k')
fig1.colorbar(plot)
ax1.set_title('Thermal output power [W]')
ax1.set_xlabel('Primary input temperature [°C]')
ax1.set_ylabel('Secondary output temperature [°C]')
fig1.show

<bound method Figure.show of <Figure size 432x288 with 2 Axes>>

COOLING: Schematic plot of EER, P_el and P_th for an generic air/water heat pump: subtype = regulated

# Define Temperatures
T_in_primary = range(20,36)
T_in_secondary = range(10,30)
T_in = numpy.array([])
T_out = numpy.array([])

# Load parameters of generic air/water | regulated
parameters = hpl.get_parameters('Generic', group_id=1, t_in=-7, t_out=52, p_th = 10000)
heatpump=hpl.HeatPump(parameters)
results=pd.DataFrame()
# Create input series
for t1 in T_in_primary:
    for t2 in T_in_secondary:
        T_in=numpy.append(T_in,t1)
        T_out=numpy.append(T_out,t2)
results=heatpump.simulate(t_in_primary=T_in, t_in_secondary=T_out, t_amb=T_in, mode=2)
results=pd.DataFrame.from_dict(results)
# Plot EER
fig1, ax1 = plt.subplots()
plot = plt.tricontourf(results['T_in'], results['T_out'], results['EER'])
ax1.tricontour(results['T_in'], results['T_out'], results['EER'], colors='k')
fig1.colorbar(plot)
ax1.set_title('EER')
ax1.set_xlabel('Primary input temperature [°C]')
ax1.set_ylabel('Secondary output temperature [°C]')
fig1.show

# Plot electrical input power
fig1, ax1 = plt.subplots()
plot = plt.tricontourf(results['T_in'], results['T_out'], results['P_el'])
ax1.tricontour(results['T_in'], results['T_out'], results['P_el'], colors='k')
fig1.colorbar(plot)
ax1.set_title('Eletrical input power [W]')
ax1.set_xlabel('Primary input temperature [°C]')
ax1.set_ylabel('Secondary output temperature [°C]')
fig1.show

# Plot thermal output power
fig1, ax1 = plt.subplots()
plot = plt.tricontourf(results['T_in'], results['T_out'], results['P_th'])
ax1.tricontour(results['T_in'], results['T_out'], results['P_th'], colors='k')
fig1.colorbar(plot)
ax1.set_title('Thermal output power [W]')
ax1.set_xlabel('Primary input temperature [°C]')
ax1.set_ylabel('Secondary output temperature [°C]')
fig1.show

<bound method Figure.show of <Figure size 432x288 with 2 Axes>>

6. Validation

The following plots will give you a detailed view on the differences between simulation and measurement from heat pump keymark. Therefore, all set points for all heat pumps are loaded from the file database_heating_average_normalized_subtypes_validation.csv.

df = pd.read_csv('../output/database_heating_average_normalized_subtypes_validation.csv')

6.1 Air/Water | on/off

• The mean absolute percentage error (MAPE) over all heat pumps is
  • 5.4 % for COP
  • 2.5 % for P_el
  • 5.8 % for P_th
• The errors come from deviation mostly at low ambient / outside temperatures

# Plot relative error for all heat pumps of type air/water | on-off
Group = 4
data = df.loc[(df['Group']==Group)]
data = data[['T_amb [°C]','RE_COP', 'RE_P_el', 'RE_P_th']]
ax = data.boxplot(by='T_amb [°C]', layout=(1,3), figsize=(10,5), showfliers=False)
ax[0].set_ylim(-60,60)
ax[0].set_ylabel('relative error in %')
data.abs().mean()[1:4] # mean absolute percentage error (MAPE)

RE_COP     5.404139
RE_P_el    2.525073
RE_P_th    5.806641
dtype: float64

# Plot absolute values all heat pumps of type air/water | on-off as scatter plot
Group = 4
fig, ax = plt.subplots(1,3)
data = df.loc[(df['Group']==Group)]
data.plot.scatter(ax=ax[0], x='COP',y='COP_sim', alpha=0.3, figsize=(12,5), grid=True, title='Coefficient of performance')
data.plot.scatter(ax=ax[1], x='P_el [W]',y='P_el_sim', alpha=0.3, figsize=(15,5), grid=True, title='Electrical input power [W]')
data.plot.scatter(ax=ax[2], x='P_th [W]',y='P_th_sim', alpha=0.3, figsize=(15,5), grid=True, title='Thermal output power [W]')
ax[0].set_xlim(0,10)
ax[0].set_ylim(0,10)
ax[1].set_xlim(0,20000)
ax[1].set_ylim(0,20000)
ax[2].set_xlim(0,50000)
ax[2].set_ylim(0,50000)

(0.0, 50000.0)

6.2 Brine/Water | on/off

• The mean absolute percentage error (MAPE) over all heat pumps is
  • 3.9 % for COP
  • 1.7 % for P_el
  • 2.7 % for P_th
• Important: Validation data is only available at primary input temperature of 0 °C.

# Plot relative error for all heat pumps of type brine/water | on-off
Group = 5
data = df.loc[(df['Group']==Group)]
data = data[['T_amb [°C]','RE_COP', 'RE_P_el', 'RE_P_th']]
ax = data.boxplot(by='T_amb [°C]', layout=(1,3), figsize=(10,5), showfliers=False)
ax[0].set_ylim(-60,60)
ax[0].set_ylabel('relative error in %')
data.abs().mean()[1:4] # mean absolute percentage error (MAPE)

RE_COP     3.897897
RE_P_el    1.678427
RE_P_th    2.668440
dtype: float64

# Plot absolute values all heat pumps of type brine/water | on-off as scatter plot
Group = 5
fig, ax = plt.subplots(1,3)
data = df.loc[(df['Group']==Group)]
data.plot.scatter(ax=ax[0], x='COP',y='COP_sim', alpha=0.3, figsize=(12,5), grid=True, title='Coefficient of performance')
data.plot.scatter(ax=ax[1], x='P_el [W]',y='P_el_sim', alpha=0.3, figsize=(15,5), grid=True, title='Electrical input power [W]')
data.plot.scatter(ax=ax[2], x='P_th [W]',y='P_th_sim', alpha=0.3, figsize=(15,5), grid=True, title='Thermal output power [W]')
ax[0].set_xlim(0,10)
ax[0].set_ylim(0,10)
ax[1].set_xlim(0,20000)
ax[1].set_ylim(0,20000)
ax[2].set_xlim(0,50000)
ax[2].set_ylim(0,50000)

(0.0, 50000.0)

6.3 Water/Water | on/off

• The mean absolute percentage error (MAPE) over all heat pumps is
  • 1.6 % for COP
  • 1.6 % for P_el
  • 2.4 % for P_th
• Important: Validation data is only available at primary input temperature of 10 °C.

# Plot relative error for all heat pumps of type water/water | on-off
Group = 6
data = df.loc[(df['Group']==Group)]
data = data[['T_amb [°C]','RE_COP', 'RE_P_el', 'RE_P_th']]
ax = data.boxplot(by='T_amb [°C]', layout=(1,3), figsize=(10,5), showfliers=False)
ax[0].set_ylim(-60,60)
ax[0].set_ylabel('relative error in %')
data.abs().mean()[1:4] # mean absolute percentage error (MAPE)

RE_COP     1.563460
RE_P_el    1.582038
RE_P_th    2.415430
dtype: float64

# Plot absolute values all heat pumps of type water/water | on-off as scatter plot
Group = 6
fig, ax = plt.subplots(1,3)
data = df.loc[(df['Group']==Group)]
data.plot.scatter(ax=ax[0], x='COP',y='COP_sim', alpha=0.3, figsize=(12,5), grid=True, title='Coefficient of performance')
data.plot.scatter(ax=ax[1], x='P_el [W]',y='P_el_sim', alpha=0.3, figsize=(15,5), grid=True, title='Electrical input power [W]')
data.plot.scatter(ax=ax[2], x='P_th [W]',y='P_th_sim', alpha=0.3, figsize=(15,5), grid=True, title='Thermal output power [W]')
ax[0].set_xlim(0,10)
ax[0].set_ylim(0,10)
ax[1].set_xlim(0,20000)
ax[1].set_ylim(0,20000)
ax[2].set_xlim(0,50000)
ax[2].set_ylim(0,50000)

(0.0, 50000.0)

6.4 Air/Water | regulated

Heating

• The mean absolute percentage error (MAPE) over all heat pumps is
  • 12.1 % for COP
  • 19.5 % for P_el
  • 23.5 % for P_th

Cooling

• The mean absolute percentage error (MAPE) over all heat pumps is
  • 4.8 % for EER
  • 16.5 % for P_el
  • 17.0 % for P_th

Because of different control strategies, the deviation over different heat pump models is much higher compared to on/off types.

6.4.1 Heating

# Plot relative error for all heat pumps of type air/water | regulated
Group = 1
data = df.loc[(df['Group']==Group)]
data = data[['T_amb [°C]','RE_COP', 'RE_P_el', 'RE_P_th']]
ax = data.boxplot(by='T_amb [°C]', layout=(1,3), figsize=(10,5), showfliers=False)
ax[0].set_ylim(-60,60)
ax[0].set_ylabel('relative error in %')
data.abs().mean()[1:4] # mean absolute percentage error (MAPE)

RE_COP     12.076160
RE_P_el    19.456473
RE_P_th    23.522542
dtype: float64

# Plot absolute values all heat pumps of type air/water | regulated as scatter plot
Group = 1
fig, ax = plt.subplots(1,3)
data = df.loc[(df['Group']==Group)]
data.plot.scatter(ax=ax[0], x='COP',y='COP_sim', alpha=0.3, figsize=(12,5), grid=True, title='Coefficient of performance')
data.plot.scatter(ax=ax[1], x='P_el [W]',y='P_el_sim', alpha=0.1, figsize=(15,5), grid=True, title='Electrical input power [W]')
data.plot.scatter(ax=ax[2], x='P_th [W]',y='P_th_sim', alpha=0.1, figsize=(15,5), grid=True, title='Thermal output power [W]')
ax[0].set_xlim(0,10)
ax[0].set_ylim(0,10)
ax[1].set_xlim(0,20000)
ax[1].set_ylim(0,20000)
ax[2].set_xlim(0,50000)
ax[2].set_ylim(0,50000)

(0.0, 50000.0)

6.4.2 Cooling

data = pd.read_csv('../output/database_cooling_reduced_normalized_validation.csv')
data = data[['T_outside [°C]','RE_Pdc', 'RE_P_el', 'RE_EER']]
ax = data.boxplot(by='T_outside [°C]', layout=(1,3), figsize=(10,5), showfliers=False)
ax[0].set_ylim(-60,60)
ax[0].set_ylabel('relative error in %')
data.abs().mean()[1:4] # mean absolute percentage error (MAPE)

RE_Pdc     16.964163
RE_P_el    16.476100
RE_EER      4.843020
dtype: float64

data = pd.read_csv('../output/database_cooling_reduced_normalized_validation.csv')
fig, ax = plt.subplots(1,3)
data.plot.scatter(ax=ax[0], x='EER',y='EER_sim', alpha=0.3, figsize=(12,5), grid=True, title='Coefficient of performance')
data.plot.scatter(ax=ax[1], x='P_el [W]',y='P_el_sim', alpha=0.1, figsize=(15,5), grid=True, title='Electrical input power [W]')
data.plot.scatter(ax=ax[2], x='Pdc [W]',y='Pdc_sim', alpha=0.1, figsize=(15,5), grid=True, title='Thermal output power [W]')
ax[0].set_xlim(0,10)
ax[0].set_ylim(0,10)
ax[1].set_xlim(0,6000)
ax[1].set_ylim(0,6000)
ax[2].set_xlim(0,15000)
ax[2].set_ylim(0,15000)

(0.0, 15000.0)

6.5 Brine/Water | regulated

• The mean absolute percentage error (MAPE) is over all heat pumps is
  • 4.2 % for COP
  • 17.7 % for P_el
  • 19.7 % for P_th
• Important: Validation data is only available at primary input temperature of 0 °C.

# Plot relative error for all heat pumps of type brine/water | regulated
Group = 2
data = df.loc[(df['Group']==Group)]
data = data[['T_amb [°C]','RE_COP', 'RE_P_el', 'RE_P_th']]
ax = data.boxplot(by='T_amb [°C]', layout=(1,3), figsize=(10,5), showfliers=False)
ax[0].set_ylim(-60,60)
ax[0].set_ylabel('relative error in %')
data.abs().mean()[1:4] # mean absolute percentage error (MAPE)

RE_COP      4.228648
RE_P_el    17.650300
RE_P_th    19.737515
dtype: float64

# Plot absolute values all heat pumps of type brine/water | regulated as scatter plot
Group = 2
fig, ax = plt.subplots(1,3)
data = df.loc[(df['Group']==Group)]
data.plot.scatter(ax=ax[0], x='COP',y='COP_sim', alpha=0.3, figsize=(12,5), grid=True, title='Coefficient of performance')
data.plot.scatter(ax=ax[1], x='P_el [W]',y='P_el_sim', alpha=0.1, figsize=(15,5), grid=True, title='Electrical input power [W]')
data.plot.scatter(ax=ax[2], x='P_th [W]',y='P_th_sim', alpha=0.1, figsize=(15,5), grid=True, title='Thermal output power [W]')
ax[0].set_xlim(0,10)
ax[0].set_ylim(0,10)
ax[1].set_xlim(0,20000)
ax[1].set_ylim(0,20000)
ax[2].set_xlim(0,50000)
ax[2].set_ylim(0,50000)

(0.0, 50000.0)

7. Conclusion

• On/Off heat pumps can be simulated very well (mean relative error < 6 %)
• Regulated heat pumps show a good values in terms of COP and EER but relative mean errors about 15-20 % for electrical and thermal power, because of the non-linearity regarding different primary / secondary temperatures.
• Despite of that, generic heat pumps should work well, because the median shows only a small relative error