This script shows how to use the existing code in opengrid

to create (a) a timeseries plot and (b) a load curve of gas, water or elektricity usage.

Todo:

Change numeric "chosen_type" to a textual choice, with lookupvalue of UtilityType in Utilitytypes.



In [ ]:

    
import os, sys
import inspect
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.dates import HourLocator, DateFormatter, AutoDateLocator
import datetime as dt
import pytz
import pandas as pd
import pdb

script_dir = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe())))
# add the path to opengrid to sys.path
sys.path.append(os.path.join(script_dir, os.pardir, os.pardir))
from opengrid.library import config

c = config.Config()

sys.path.append(c.get('tmpo', 'folder'))
import tmpo
try:
    if os.path.exists(c.get('tmpo', 'data')):
        path_to_tmpo_data = c.get('tmpo', 'data')
except:
    path_to_tmpo_data = None

from opengrid.library.houseprint import houseprint

%matplotlib inline
plt.rcParams['figure.figsize']=14,8

Script settings



In [ ]:

    
hp = houseprint.load_houseprint_from_file('new_houseprint.pkl')
hp.init_tmpo(path_to_tmpo_data=path_to_tmpo_data)

Fill in here (chosen type [0-2]) what type of data you'd like to plot:



In [ ]:

    
chosen_type = 0
#  0 =water, 1 = gas, 2 = electricity

UtilityTypes = ['water', 'gas','electricity'] # {'water','gas','electricity'} 
utility =  UtilityTypes[chosen_type] # here 'electricity'

#default values:
FL_units = ['l/day', 'm^3/day ~ 10 kWh/day','Ws/day'] #TODO, to be checked!!
Base_Units = ['l/min', 'kW','kW']
Base_Corr = [1/24.0/60.0, 1/100.0/24.0/3.600 , 3.600/1000.0/24 ] #TODO,check validity of conversions!! # water => (l/day) to (l/hr), gas: (l/day) to (kW), elektr Ws/d to kW

tInt_Units = ['l', 'kWh','kWh'] #units after integration
tInt_Corr = [1/60, 3600/60, 3600/60] #TODO, to be checked!! # water => (l/hr) to (l_cumul/min), gas: kW to (kWh/min)

# units for this utility type
bUnit = Base_Units[chosen_type]
bCorr = Base_Corr[chosen_type]
fl_unit = FL_units[chosen_type]
tiUnit = tInt_Units[chosen_type]
tiCorr = tInt_Corr[chosen_type]

Available data is loaded in one big dataframe, the columns are the sensors of chosen type.

also, it is rescaled to more "managable" units (to be verified!)



In [ ]:

    
#load data
print 'Loading', utility ,'-data and converting from ',fl_unit ,' to ',bUnit,':'
df = hp.get_data(sensortype=utility)
df = df.diff() #data is cumulative, we need to take the derivative
df = df[df>0] #filter out negative values

# conversion dependent on type of utility (to be checked!!) 
df = df*bCorr



In [ ]:

    
# plot timeseries and load duration for each retained sensor

for sensor in df.columns:
    FL = hp.find_sensor(sensor).device.key
    plt.figure()
    ax1=plt.subplot(121)
    plt.plot_date(df.index, df[sensor], '-', label="{}".format(FL))
    plt.ylabel("{}-usage [{}]".format(utility,bUnit) )
    plt.legend()
    
    ax2=plt.subplot(122)
    plt.plot(np.sort(df[sensor])[::-1], label=sensor)
    plt.ylabel("{}-load curve [{}]".format(utility,bUnit) )
    plt.legend()



In [ ]:

    
#Date/Time library
from arrow import Arrow

#Prepare NVD3.js dependencies
from IPython import display as d
import nvd3
nvd3.ipynb.initialize_javascript(use_remote=True)

#Filter sensors and period
sensorlist = ['b28509eb97137e723995838c393d49df', '2923b75daf93e539e37ce5177c0008c5', 'a926bc966f178fc5d507a569a5bfc3d7']
df_water= df[sensorlist][Arrow(2015, 4, 1).datetime:Arrow(2015, 4, 2).datetime].dropna()

#Prepare chart name and timescale in epoch
chart_name = "{}-usage [{}]".format(utility,bUnit)
df_water["epoch"] = [(Arrow.fromdatetime(o) - Arrow(1970, 1, 1)).total_seconds()*1000 for o in df_water.index]

#Create NVD3 chart
water_chart = nvd3.lineChart(x_is_date=True,name=chart_name,height=450,width=800)
for sensor in sensorlist: # df.columns:
    series_name = name="{}".format(hp.find_sensor(sensor).device.key)
    water_chart.add_serie(name=series_name, x=list(df_water["epoch"]), y=list(df_water[sensor]))

water_chart

Tests with the tmpo-based approach



In [ ]:

    
start = pd.Timestamp('20150201')
end = pd.Timestamp('20150301')



In [ ]:

    
dfcum = hp.get_data(sensortype='electricity', head= start, tail = end)



In [ ]:

    
dfcum.shape



In [ ]:

    
dfcum.columns



In [ ]:

    
dfcum.tail()



In [ ]:

    
dfi = dfcum.resample(rule='900s', how='max')
dfi = dfi.interpolate(method='time')
dfi=dfi.diff()*3600/900
dfi.plot()
#dfi.ix['20150701'].plot()



In [ ]:

    
# This works, but is a bad idea if you have multiple sensors for a FLM: you obtain identical column names.
# df.rename(columns = hp.get_flukso_from_sensor, inplace=True)



In [ ]:

    
# Getting a single sensor
dfi['1a1dac9c2ac155f95c58bf1d4f4b7d01'].plot()



In [ ]: