def setUpClass(cls):
filename = join(data_dir(), 'random.h5')
cls.datastore = HDFDataStore(filename)
cls.keys = ['/building1/elec/meter{:d}'.format(i) for i in range(1,6)]
def setUpClass(cls):
filename = join(data_dir(), 'co_test.h5')
cls.dataset = DataSet(filename)
def setUpClass(cls):
filename = join(data_dir(), 'co_test.h5')
cls.dataset = DataSet(filename)
def test_LatentBayesianMelding_correctness(self):
"""
Created on Friday January 1 20:27:27 2016
@author: Mingjun Zhong, School of Informatics, University of Edinburgh
Email: [email protected]
Requirements:
1) the NILMTK: Non-Intrusive Load Monitoring Toolkit,
https://github.com/nilmtk/nilmtk;
2) the (free academic) MOSEK license:
https://www.mosek.com/resources/academic-license
References:
[1] Mingjun Zhong, Nigel Goddard and Charles Sutton.
Latent Bayesian melding for integrating individual and population models.
In Advances in Neural Information Processing Systems 28, 2015.
[2] Mingjun Zhong, Nigel Goddard and Charles Sutton.
Signal Aggregate Constraints in Additive Factorial HMMs,
with Application to Energy Disaggregation.
In Advances in Neural Information Processing Systems 27, 2014.
Note:
1) The optimization problems described in the papers were transformed to
a second order conic programming (SOCP) which suits the MOSEK solver;
I should write a technical report for this.
2) Any questions please drop me an email.
"""
############# some global variables ##########################
# Sampling time was 2 minutes
sample_seconds = 120
####### Select a house: this is the house 2 in UKDALE #########
# This is a demo for applying Latent Bayesian Melding to energy disaggregation.
# The HES data were used for training the model parameters, and the population
# models. This demo shows how to use the LBM to disaggregate the mains readings
# of a day in UKDALE.
# Note that since HES data were read every 2 minutes, so UKDALE data were
# resampled to 2 minutes.
# Note that all the readings were tranferred to the unit: deciwatt-hour,
# which is identical to the unit of HES data.
################################################################
# Note that this demo is only for a chunk (=a day).
# This should be easy to amend for all the data you want to disaggregate.
#########################################################################
## This is the building information: dataset/building_number/date
building_information = 'ukdale/building2/2013-06-08'
# Read the data from h5 file.
# meterdata is a DataFrame and its columns are: 1) appliance readings;
# 2) the mains readings; 3) the synthetic mains readings which are the sum
# of all the appliance readings.
meterdata = self.meterdata_ukdale['meterdata']
# Map the keywords of appliances between HES and UKDALE
appliance_map = {'cooker':"('cooker', 1)",
'kettle':"('kettle', 1)",
'dishwasher':"('dish washer', 1)",
'toaster':"('toaster', 1)",
'washingmachine':"('washing machine', 1)",
'fridgefreezer':"('fridge', 1)",
'microwave':"('microwave', 1)"}
########### Appliances to be disaggregated: not all of the appliance #########
meterlist = ['cooker', 'kettle', 'dishwasher','toaster',
'washingmachine','fridgefreezer', 'microwave']
## the ground truth reading for those appliances to be disaggregated ########
appliancedata = meterdata
groundTruthApplianceReading = pd.DataFrame(index=meterdata.index)
for meter in appliance_map:
groundTruthApplianceReading[meter] = meterdata[appliance_map[meter]]
appliancedata = appliancedata.drop(appliance_map[meter],axis=1)
## the sum of other meter readings which will not be disaggregated
groundTruthApplianceReading['othermeters'] = appliancedata.sum(axis=1)
## The mains readings to be disaggregated ###
mains = meterdata['mains']
#### declare an instance for lbm ################################
fhmm = FHMM_Relaxed()
# to obtain the model parameters trained by using HES data
individual_model = fhmm.import_model(meterlist,join(data_dir(), 'appliance_model_induced_density.json'))
# use lbm to disaggregate mains readings into appliances
results = fhmm.disaggregate_chunk(mains)
# the inferred appliance readings
infApplianceReading=results['inferred appliance energy']
def setUpClass(cls):
filename_meterdata_ukdale = join(data_dir(), 'meterdata_ukdale.h5')
cls.meterdata_ukdale = HDFStore(filename_meterdata_ukdale)
def setUpClass(cls):
filename = join(data_dir(), 'random_csv')
cls.datastore = CSVDataStore(filename)
cls.keys = ['/building1/elec/meter{:d}'.format(i) for i in range(1, 6)]
def setUpClass(cls):
filename = join(data_dir(), "random_csv")
cls.datastore = CSVDataStore(filename)
cls.keys = ["/building1/elec/meter{:d}".format(i) for i in range(1, 6)]
