Example of usage of data for further modelling

In this demo we will show you how to load and use data in preheat.

# Load packages from preheat and time management
from preheat_open import set_logging_level, test
set_logging_level("WARNING") # prevent annoying info logs from printing
from preheat_open.building import Building
from datetime import datetime, timedelta
import pandas as pd
pd.options.plotting.backend = "plotly"

2022-03-18 11:59:28,031  INFO      api                               Loading config from /home/ask/.preheat/config.json                                                                                                                .(api.py:126)

Let us consider a test-location

building = Building(2371)

To access energy data we query for a list of heating units (there’s only one)

heating = building.qu("heating")
print("Heating unit: ", heating)

Heating unit:  heating(heating_7252)

Let us consider the components of the heating unit

print("The indoor climate unit has the following components: ", heating.components)

The indoor climate unit has the following components:  [Component(primaryReturnT, HEAT_RETURN_TEMP (S5))]

Unfortunately, the heating unit only has the return temperature mapped not any energy meter. We must infer the energy from the difference between main and hot water. (We could also traverse down the tree of the heating unit and check if the secondaries that belong to it have energy meters, however, we refrain from that in this tutorial.)

main = building.qu("main")
print("Main:", main)
hot_water = building.qu("hotWater")
print("Hot water:", hot_water)

Main: [main(Hus 11 Main), main(MAIN [10+11])]
Hot water: hotWater(Hus 11 Varmt brugsvand)

It appears that this location shares its main unit with another location (the MAIN [10+11]) however, the unit Hus 11 Main provides us with information for only this location.

main = main[0]

Let us check that these have the desired meters

for u in [main, hot_water]:
    print(f"Components of {u.unit_type}:", u.components)

Components of main: [Component(supplyT, FLOW_TEMPERATURE), Component(returnT, RETURN_TEMPERATURE), Component(flow, VOLUME_FLOW), Component(volume, VOLUME), Component(energy, ENERGY), Component(power, POWER)]
Components of hotWater: [Component(primarySupplyT, FLOW_TEMPERATURE), Component(primaryReturnT, RETURN_TEMPERATURE), Component(primaryFlow, VOLUME_FLOW), Component(volume, VOLUME), Component(energy, ENERGY), Component(power, POWER)]

Luckily, they both provide power readings. Let us load the daily average power consumption for the past three months and approximate the consumption as the difference between the two. To avoid wasting effort on loading all available data, we restrict to the component “power” (you can leave components unspecified to load all data!)

end = datetime.now()
start = end - timedelta(days=90)
resolution = "hour"
df = pd.DataFrame()
for u in [main, hot_water]:
    u.load_data(start,end,resolution, components=["power"])
    df[u.unit_type] = u.data["power"]
df["heating"] = df[main.unit_type] - df[hot_water.unit_type]
df.describe().transpose()

	count	mean	std	min	25%	50%	75%	max
main	2157.0	46.093011	13.453564	7.266667	39.225000	48.218182	54.916667	83.036364
hotWater	2158.0	5.766640	3.574115	1.600000	2.955492	4.713636	7.448864	25.760000
heating	2157.0	40.324532	12.962233	3.541667	34.383333	43.036364	48.063636	79.927273

While we are at it, let us load the weather information for the same period and append the data to our dataframe. Note that as a rule of thumb, the weather components begin with a capital letter. That allows us to distinguish the indoor temperature (a component of the indoor climate units) from the outdoor Temperature etc. We load the ambient temperature and the incoming sun power.

weather = building.weather
print("Weather components:", weather.components)
weather_components = ["Temperature", "DirectSunPowerVertical", "DiffuseSunPower"]
weather.load_data(start,end,resolution, components=weather_components)
df[weather_components] = weather.data[weather_components]
df.describe().transpose()

Weather components: [Component(Temperature, None), Component(Humidity, None), Component(WindDirection, None), Component(WindSpeed, None), Component(Pressure, None), Component(LowClouds, None), Component(MediumClouds, None), Component(HighClouds, None), Component(Fog, None), Component(WindGust, None), Component(DewPointTemperature, None), Component(Cloudiness, None), Component(Precipitation, None), Component(DirectSunPower, None), Component(DiffuseSunPower, None), Component(SunAltitude, None), Component(SunAzimuth, None), Component(DirectSunPowerVertical, None)]

	count	mean	std	min	25%	50%	75%	max
main	2157.0	46.093011	13.453564	7.266667	39.225000	48.218182	54.916667	83.036364
hotWater	2158.0	5.766640	3.574115	1.600000	2.955492	4.713636	7.448864	25.760000
heating	2157.0	40.324532	12.962233	3.541667	34.383333	43.036364	48.063636	79.927273
Temperature	2159.0	3.425938	2.605860	-5.500000	1.700000	3.300000	5.500000	9.600000
DirectSunPowerVertical	2159.0	0.118278	0.207390	0.000000	0.000000	0.000000	0.156606	0.741441
DiffuseSunPower	2159.0	0.014667	0.027488	0.000000	0.000000	0.000000	0.016838	0.153348

We can use the acquired data to construct a simple linear model for the heating energy demand

import statsmodels.formula.api as smf
res = smf.ols("heating ~ 1 + Temperature + DirectSunPowerVertical + DiffuseSunPower", df).fit()
res.summary()

OLS Regression Results

Dep. Variable:	heating	R-squared:	0.479
Model:	OLS	Adj. R-squared:	0.479
Method:	Least Squares	F-statistic:	660.5
Date:	Fri, 18 Mar 2022	Prob (F-statistic):	2.25e-304
Time:	11:59:29	Log-Likelihood:	-7882.7
No. Observations:	2157	AIC:	1.577e+04
Df Residuals:	2153	BIC:	1.580e+04
Df Model:	3
Covariance Type:	nonrobust

	coef	std err	t	P>|t|	[0.025	0.975]
Intercept	49.4827	0.341	145.220	0.000	48.814	50.151
Temperature	-1.3682	0.079	-17.352	0.000	-1.523	-1.214
DirectSunPowerVertical	-27.9668	1.548	-18.061	0.000	-31.003	-24.930
DiffuseSunPower	-79.4236	11.723	-6.775	0.000	-102.413	-56.434

Omnibus:	9.255	Durbin-Watson:	0.777
Prob(Omnibus):	0.010	Jarque-Bera (JB):	10.786
Skew:	-0.082	Prob(JB):	0.00455
Kurtosis:	3.305	Cond. No.	256.

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.

Let us investigate the quality of the fit