def correlate(chan_id, weather_variable, subplot_index, annotate_y):
# 25 = lighting circuit # (R^2 = 0.443)
# 8 = kitchen lights (R^2 = 0.194)
# 2 = boiler (versus radiation R^2 = 0.052,
# versus mean_temp R^2 = 0.298,
# versus max_temp R^2 = 0.432,
# versus min_temp R^2 = 0.212)
# 3 = solar (R^2 = 0.798)
# 12 = fridge vs min_temp R^2 = 0.255 (with on_power_threshold = 20)
print("Opening channel data...")
channel = Channel(DATA_DIR, chan_id)
print("Calculating...")
channel.on_power_threshold = 20
hours_on = channel.usage_per_period('D', tz_convert='UTC').hours_on
hours_on = hours_on[hours_on > ON_DURATION_THRESHOLD]
hours_on.description = 'hours on'
print("Got {} days of data from usage_per_period.".format(hours_on.size))
print("Plotting...")
x_aligned, y_aligned = pda.stats.align(weather_variable, hours_on)
print(x_aligned.description)
slope, intercept, r_value, p_value, std_err = linregress(x_aligned.values,
y_aligned.values)
ax = fig.add_subplot(2,2,subplot_index)
ax = spfl.format_axes(ax)
ax = pda.stats.plot_regression_line(ax, x_aligned, y_aligned, slope,
intercept, r_value,
annotate_y=annotate_y)
print("R^2={:.3f}".format(r_value**2))
ax.set_title('Correlation between ' + channel.get_long_name() + ' and ' +
metoffice.get_long_name(weather_variable.name))
def test(data_input='random'):
# First, we will specify the prior. We will then generate some fake data
# from the prior specification. We will then perform inference. Then
# we'll plot some things.
def hazard_func(r):
return constant_hazard(r, _lambda=200)
if data_input == 'random':
# generate test data
N = 100 # how many data points to generate?
x, changepoints = generate_test_data(N, hazard_func)
elif data_input == 'ones':
x = np.ones(N)
changepoints = []
elif data_input == 'signature':
from pda.channel import Channel
from os import path
DATA_DIR = '/data/mine/domesticPowerData/BellendenRd/wattsUp'
#SIG_DATA_FILENAME = 'breadmaker1.csv'
SIG_DATA_FILENAME = 'washingmachine1.csv'
chan = Channel()
chan.load_wattsup(path.join(DATA_DIR, SIG_DATA_FILENAME))
x = chan.series.values[142:1647]
N = x.size
# plot
fig = plt.figure()
ax = fig.add_subplot(2, 1, 1)
ax.plot(x)
ylim = ax.get_ylim()
for cp in changepoints:
ax.plot([cp, cp], ylim, color='k')
# do inference
beliefs, maxes = inference(x, hazard_func)
# plot beliefs
beliefs = beliefs.astype(np.float32)
# print(beliefs)
ax2 = fig.add_subplot(2, 1, 2, sharex=ax)
ax2.imshow(-np.log(beliefs),
interpolation='none',
aspect='auto',
origin='lower',
cmap=plt.cm.Blues)
ax2.plot(maxes, color='r')
ax2.set_xlim([0, N])
ax2.set_ylim([0, ax2.get_ylim()[1]])
plt.draw()
return beliefs, maxes
def test(data_input='random'):
# First, we will specify the prior. We will then generate some fake data
# from the prior specification. We will then perform inference. Then
# we'll plot some things.
hazard_func = lambda r: constant_hazard(r, _lambda=200)
if data_input == 'random':
# generate test data
N = 100 # how many data points to generate?
x, changepoints = generate_test_data(N, hazard_func)
elif data_input == 'ones':
x = np.ones(N)
changepoints = []
elif data_input == 'signature':
from pda.channel import Channel
from os import path
DATA_DIR = '/data/mine/domesticPowerData/BellendenRd/wattsUp'
#SIG_DATA_FILENAME = 'breadmaker1.csv'
SIG_DATA_FILENAME = 'washingmachine1.csv'
chan = Channel()
chan.load_wattsup(path.join(DATA_DIR, SIG_DATA_FILENAME))
x = chan.series.values[142:1647]
N = x.size
# plot
fig = plt.figure()
ax = fig.add_subplot(2,1,1)
ax.plot(x)
ylim = ax.get_ylim()
for cp in changepoints:
ax.plot([cp, cp], ylim, color='k')
# do inference
beliefs, maxes = inference(x, hazard_func)
# plot beliefs
beliefs = beliefs.astype(np.float32)
#print(beliefs)
ax2 = fig.add_subplot(2,1,2, sharex=ax)
ax2.imshow(-np.log(beliefs), interpolation='none', aspect='auto',
origin='lower', cmap=plt.cm.Blues)
ax2.plot(maxes, color='r')
ax2.set_xlim([0, N])
ax2.set_ylim([0, ax2.get_ylim()[1]])
plt.draw()
return beliefs, maxes
def load_dataset(data_dir=DD,
ignore_chans=None,
only_load_chans=None,
start_date=None,
end_date=None):
"""Loads an entire dataset directory.
Args:
data_dir (str)
ignore_chans (list of ints or label strings): optional.
Don't load these channels.
only_load_chans (list of ints or label strings): optional.
Returns:
list of Channels
"""
if ignore_chans is not None:
assert (isinstance(ignore_chans, list))
channels = []
labels = load_labels(data_dir)
print("Found", len(labels), "entries in labels.dat")
for chan, label in labels.iteritems():
if ignore_chans is not None:
if chan in ignore_chans or label in ignore_chans:
print("Ignoring chan", chan, label)
continue
if only_load_chans is not None:
if chan not in only_load_chans and label not in only_load_chans:
print("Ignoring chan", chan, label)
continue
print("Attempting to load chan", chan, label, "...", end=" ")
sys.stdout.flush()
try:
c = Channel(data_dir,
chan,
start_date=start_date,
end_date=end_date)
except IOError:
print("FAILED!")
else:
channels.append(c)
print("success.")
return channels
from __future__ import print_function, division
from pda.channel import Channel
import matplotlib.pyplot as plt
from matplotlib.ticker import MaxNLocator
import numpy as np
import os, datetime
import setupPlottingForLaTeX as spfl
NORMALISED_BAR_COLOR = 'gray'
UNNORMALISED_LINE_COLOR = 'k'
DATA_DIR = '/data/mine/vadeec/merged/house1'
FIGURE_PATH = os.path.expanduser('~/Dropbox/MyWork/imperial/PhD/writing'
'/papers/tetc2013/figures/')
LATEX_PDF_OUTPUT_FILENAME = os.path.join(FIGURE_PATH, 'power_histograms.pdf')
voltage = Channel()
voltage.load_high_freq_mains(os.path.join(DATA_DIR, 'mains.dat'), 'volts')
CHAN_IDS = [24,5,12,22,7,9,8,11,42,14,16,4]
spfl.setup(columns=2)
TITLE_Y = 0.7
MINIMUM_BIN_COUNT = 100
chans = []
normalised = []
for chan_id in CHAN_IDS:
# Get channel data
print("loading channel", chan_id)
c = Channel(DATA_DIR, chan_id)
chans.append(c)
TIMESPAN = 'W' # D (daily) or W (weekly)
CHAN_IDS = [14,22]
spfl.setup()
GRID = False
TITLE_Y = 0.75
XTICKS_ON = True
LATEX_PDF_OUTPUT_FILENAME = os.path.join(FIGURE_PATH,
'weekly_usage_histograms'+FIGURE_SUFFIX)
else:
CHAN_IDS = []
CHANS = []
for chan_id in CHAN_IDS:
# Get channel data
print("Loading channel", chan_id)
c = Channel(DATA_DIR, chan_id)
c = c.crop(START_DATE, END_DATE)
CHANS.append(c)
if FIGURE_PRESET == 'boiler seasons':
BIN_SIZE = 'T' # D (daily) or H (hourly) or T (minutely)
TIMESPAN = 'D' # D (daily) or W (weekly)
spfl.setup()
GRID = False
TITLE_Y = 0.7
XTICKS_ON = True
LATEX_PDF_OUTPUT_FILENAME = os.path.join(FIGURE_PATH,
'seasonal_variation'+FIGURE_SUFFIX)
print("Loading winter boiler data...")
winter_boiler = Channel(DATA_DIR, 2)
winter_boiler = winter_boiler.crop(datetime.datetime(year=2013, month=2, day=1),
def init_aggregate_and_appliance_dataset_figure(
start_date,
end_date,
n_subplots=2,
aggregate_type='one second',
plot_both_aggregate_signals=False,
data_dir=DD,
plot_appliance_ground_truth=True,
ignore_chans=None,
**kwargs):
"""Initialise a basic figure with multiple subplots. Plot aggregate
data. Optionally plot appliance ground truth dataset.
Args:
start_date, end_date (str): Required. e.g. '2013/6/4 18:00'
n_subplots (int): Required. Must be >= 1. Includes aggregate and
appliance ground truth plots.
aggregate_type (str): 'one second' or 'current cost'. The flavour of
aggregate data to load, plot and return.
plot_both_aggregate_signals (bool): Default==False. Plot both flavours
of aggregate data? Has no effect on which flavour is returned.
data_dir (str): Default=DD
plot_appliance_ground_truth (bool): Default==True
ignore_chans (list of strings or ints): Defaults to a standard list of
channels to ignore.
**kwargs: passed to ax.plot
Returns:
subplots (list of axes),
chan (pda.Channel)
"""
if plot_appliance_ground_truth:
assert (n_subplots >= 2)
else:
assert (n_subplots >= 1)
# Initialise figure and subplots
fig = plt.figure()
fig.canvas.set_window_title(start_date + ' - ' + end_date)
subplots = [fig.add_subplot(n_subplots, 1, 1)]
for i in range(2, n_subplots + 1):
subplots.append(fig.add_subplot(n_subplots, 1, i, sharex=subplots[0]))
# Load and plot aggregate channel(s)
if aggregate_type == 'one second' or plot_both_aggregate_signals:
print('Loading high freq mains...')
one_sec = Channel()
one_sec.load_normalised(data_dir,
high_freq_param='active',
start_date=start_date,
end_date=end_date)
one_sec.plot(subplots[0], color='k', **kwargs)
if aggregate_type == 'current cost' or plot_both_aggregate_signals:
print('Loading Current Cost aggregate...')
cc = Channel(data_dir,
'aggregate',
start_date=start_date,
end_date=end_date) # cc = Current cost
cc.plot(subplots[0], color='r', **kwargs)
subplots[0].set_title('Aggregate. 1s active power, normalised.')
subplots[0].legend()
chan = one_sec if aggregate_type == 'one second' else cc
if plot_appliance_ground_truth:
print('Loading appliance ground truth dataset...')
if ignore_chans is None:
ignore_chans = [
'aggregate', 'amp_livingroom', 'adsl_router',
'livingroom_s_lamp', 'gigE_&_USBhub', 'livingroom_s_lamp2',
'iPad_charger', 'subwoofer_livingroom', 'livingroom_lamp_tv',
'DAB_radio_livingroom', 'kitchen_lamp2',
'kitchen_phone&stereo', 'utilityrm_lamp', 'samsung_charger',
'kitchen_radio', 'bedroom_chargers', 'data_logger_pc',
'childs_table_lamp', 'baby_monitor_tx', 'battery_charger',
'office_lamp1', 'office_lamp2', 'office_lamp3', 'gigE_switch'
]
ds = load_dataset(data_dir,
ignore_chans=ignore_chans,
start_date=start_date,
end_date=end_date)
print("Removing inactive channels...")
ds = remove_inactive_channels(ds)
print("Plotting dataset ground truth...")
plot_each_channel_activity(subplots[1], ds)
return subplots, chan
from __future__ import print_function, division
from pda.channel import Channel
import matplotlib.pyplot as plt
from matplotlib.ticker import MaxNLocator
import numpy as np
import os, datetime
import setupPlottingForLaTeX as spfl
NORMALISED_BAR_COLOR = 'gray'
UNNORMALISED_LINE_COLOR = 'k'
DATA_DIR = '/data/mine/vadeec/merged/house1'
FIGURE_PATH = os.path.expanduser('~/Dropbox/MyWork/imperial/PhD/writing'
'/papers/tetc2013/figures/')
LATEX_PDF_OUTPUT_FILENAME = os.path.join(FIGURE_PATH, 'power_histograms.pdf')
voltage = Channel()
voltage.load_high_freq_mains(os.path.join(DATA_DIR, 'mains.dat'), 'volts')
CHAN_IDS = [24, 5, 12, 22, 7, 9, 8, 11, 42, 14, 16, 4]
spfl.setup(columns=2)
TITLE_Y = 0.7
MINIMUM_BIN_COUNT = 100
chans = []
normalised = []
for chan_id in CHAN_IDS:
# Get channel data
print("loading channel", chan_id)
c = Channel(DATA_DIR, chan_id)
chans.append(c)
DATA_DIR = '/data/mine/vadeec/merged/house1'
FIGURE_PATH = os.path.expanduser('~/Dropbox/MyWork/imperial/PhD/writing'
'/papers/tetc2013/figures/')
LATEX_PDF_OUTPUT_FILENAME = os.path.join(FIGURE_PATH, 'on_durations.pdf')
CHAN_IDS = [6,5,12,22,39,9,8,11,42,13,16,4]
spfl.setup(columns=2)
TITLE_Y = 0.7
chans = []
for chan_id in CHAN_IDS:
# Get channel data
print("loading channel", chan_id)
c = Channel(DATA_DIR, chan_id)
chans.append(c)
#-------------------------------------------
fig = plt.figure()
n_subplots = len(chans)
for c in chans:
subplot_index = chans.index(c) + 1
ignore_n_off_samples = {'breadmaker': 600,
'washing_machine': 10,
'dishwasher': 10}
on_durations = c.durations('on',
ignore_n_off_samples=ignore_n_off_samples.get(c.name))
#!/bin/python
from __future__ import print_function, division
from pda.channel import Channel
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
from matplotlib import animation
import numpy as np
import datetime
BIN_SIZE = 'H' # H (hourly) or T (minutely)
c = Channel('/data/mine/vadeec/jack-merged', 2)
START_PERIOD = c.series.index[0].to_period('W')
width = 1440 if BIN_SIZE=='T' else 24
fig = plt.figure()
ax = fig.add_subplot(111)
COLOR = 'b'
x = np.arange(width)
y = np.zeros(width)
rects = ax.bar(x, y, facecolor=COLOR, edgecolor=COLOR)
ax.set_xlim([0, width])
ax.set_ylim([0, 10])
ax.xaxis.set_major_locator(ticker.MultipleLocator(width / 12))
def format_time(x, pos=None):
if BIN_SIZE == 'T': #minutely
hours = x // 60
else:
hours = x
return '{:d}'.format(int(hours))
DATA_DIR = '/data/mine/vadeec/merged/house1'
FIGURE_PATH = os.path.expanduser('~/Dropbox/MyWork/imperial/PhD/writing'
'/papers/tetc2013/figures/')
LATEX_PDF_OUTPUT_FILENAME = os.path.join(FIGURE_PATH, 'on_durations.pdf')
CHAN_IDS = [6, 5, 12, 22, 39, 9, 8, 11, 42, 13, 16, 4]
spfl.setup(columns=2)
TITLE_Y = 0.7
chans = []
for chan_id in CHAN_IDS:
# Get channel data
print("loading channel", chan_id)
c = Channel(DATA_DIR, chan_id)
chans.append(c)
#-------------------------------------------
fig = plt.figure()
n_subplots = len(chans)
for c in chans:
subplot_index = chans.index(c) + 1
ignore_n_off_samples = {
'breadmaker': 600,
'washing_machine': 10,
'dishwasher': 10
}
on_durations = c.durations('on',
def init_aggregate_and_appliance_dataset_figure(
start_date, end_date, n_subplots=2,
aggregate_type='one second', plot_both_aggregate_signals=False,
data_dir=DD, plot_appliance_ground_truth=True, ignore_chans=None,
**kwargs):
"""Initialise a basic figure with multiple subplots. Plot aggregate
data. Optionally plot appliance ground truth dataset.
Args:
start_date, end_date (str): Required. e.g. '2013/6/4 18:00'
n_subplots (int): Required. Must be >= 1. Includes aggregate and
appliance ground truth plots.
aggregate_type (str): 'one second' or 'current cost'. The flavour of
aggregate data to load, plot and return.
plot_both_aggregate_signals (bool): Default==False. Plot both flavours
of aggregate data? Has no effect on which flavour is returned.
data_dir (str): Default=DD
plot_appliance_ground_truth (bool): Default==True
ignore_chans (list of strings or ints): Defaults to a standard list of
channels to ignore.
**kwargs: passed to ax.plot
Returns:
subplots (list of axes),
chan (pda.Channel)
"""
if plot_appliance_ground_truth:
assert(n_subplots >= 2)
else:
assert(n_subplots >= 1)
# Initialise figure and subplots
fig = plt.figure()
fig.canvas.set_window_title(start_date + ' - ' + end_date)
subplots = [fig.add_subplot(n_subplots, 1, 1)]
for i in range(2, n_subplots+1):
subplots.append(fig.add_subplot(n_subplots, 1, i, sharex=subplots[0]))
# Load and plot aggregate channel(s)
if aggregate_type=='one second' or plot_both_aggregate_signals:
print('Loading high freq mains...')
one_sec = Channel()
one_sec.load_normalised(data_dir, high_freq_param='active',
start_date=start_date, end_date=end_date)
one_sec.plot(subplots[0], color='k', **kwargs)
if aggregate_type=='current cost' or plot_both_aggregate_signals:
print('Loading Current Cost aggregate...')
cc = Channel(data_dir, 'aggregate',
start_date=start_date, end_date=end_date) # cc = Current cost
cc.plot(subplots[0], color='r', **kwargs)
subplots[0].set_title('Aggregate. 1s active power, normalised.')
subplots[0].legend()
chan = one_sec if aggregate_type=='one second' else cc
if plot_appliance_ground_truth:
print('Loading appliance ground truth dataset...')
if ignore_chans is None:
ignore_chans=['aggregate', 'amp_livingroom', 'adsl_router',
'livingroom_s_lamp', 'gigE_&_USBhub',
'livingroom_s_lamp2', 'iPad_charger',
'subwoofer_livingroom', 'livingroom_lamp_tv',
'DAB_radio_livingroom', 'kitchen_lamp2',
'kitchen_phone&stereo', 'utilityrm_lamp',
'samsung_charger', 'kitchen_radio',
'bedroom_chargers', 'data_logger_pc',
'childs_table_lamp', 'baby_monitor_tx',
'battery_charger', 'office_lamp1', 'office_lamp2',
'office_lamp3', 'gigE_switch']
ds = load_dataset(data_dir, ignore_chans=ignore_chans,
start_date=start_date, end_date=end_date)
print("Removing inactive channels...")
ds = remove_inactive_channels(ds)
print("Plotting dataset ground truth...")
plot_each_channel_activity(subplots[1], ds)
return subplots, chan
TIMESPAN = 'W' # D (daily) or W (weekly)
CHAN_IDS = [14, 22]
spfl.setup()
GRID = False
TITLE_Y = 0.75
XTICKS_ON = True
LATEX_PDF_OUTPUT_FILENAME = os.path.join(
FIGURE_PATH, 'weekly_usage_histograms' + FIGURE_SUFFIX)
else:
CHAN_IDS = []
CHANS = []
for chan_id in CHAN_IDS:
# Get channel data
print("Loading channel", chan_id)
c = Channel(DATA_DIR, chan_id)
c = c.crop(START_DATE, END_DATE)
CHANS.append(c)
if FIGURE_PRESET == 'boiler seasons':
BIN_SIZE = 'T' # D (daily) or H (hourly) or T (minutely)
TIMESPAN = 'D' # D (daily) or W (weekly)
spfl.setup()
GRID = False
TITLE_Y = 0.7
XTICKS_ON = True
LATEX_PDF_OUTPUT_FILENAME = os.path.join(
FIGURE_PATH, 'seasonal_variation' + FIGURE_SUFFIX)
print("Loading winter boiler data...")
winter_boiler = Channel(DATA_DIR, 2)
winter_boiler = winter_boiler.crop(