solardash/fakedata.py

import arrow
import random
import pprint
import json

from datetime import datetime, timedelta

rand = random.Random('a seed')

tz = 'US/Pacific'

startdate = arrow.Arrow(2019, 11, 1, tzinfo=tz)
enddate = arrow.Arrow(2019, 12, 10, tzinfo=tz)

meanwhprod = 20000
sigwhprod = 2000

def drange(s, e, interval):
	cmpfun = lambda s, e, ts, te: te < e
	if e < s:
		interval = -interval
		cmpfun = lambda s, e, ts, te: te > e
	ts = s.clone()
	te = ts + interval
	#print('dr:', repr((s, e, ts, te, cmpfun(s, e, ts, te), interval)))
	while cmpfun(s, e, ts, te):
		yield ts + (te - ts)
		ts, te = te, te + interval

# idea:
# first hour linear ramp up (25% in first half, 75% in second half)
# middle 5 hours near constant generation
# first/tailing linear is equiv of an hour total, so total power / 6

# approx 7 hours time
def makestartend(t):
	s = t.replace(hour=9).shift(minutes=rand.gauss(30, 10))
	e = t.replace(hour=16).shift(minutes=rand.gauss(30, 10))

	return (s, e)

def normdist(small, big, amount, minsize):
	'''Distribute most twoards the big side, total distribute amount
	over the entire range, [small, big].
	'''

	ret = []

	timediff = abs(big - small)
	if timediff < minsize:
		scaledamount = amount * (timedelta(hours=1) / timediff)
		midpnt = small + (big - small) / 2
		#print('ndf:', repr((small, big, midpnt, amount, scaledamount)))
		return [ (midpnt, scaledamount) ]

	#print('nd:', repr((small, big, amount)))
	dist = big - small
	halfpoint = small + (dist / 9 * 5)
	ret.extend(normdist(small, halfpoint, amount / 2, minsize))
	ret.extend(normdist(halfpoint, big, amount / 2, minsize))

	#print('ndr:')
	#pprint.pprint(ret)
	return ret

def linramp(start, end, wtarget, minsize):
	mid = start + (end - start) / 2
	timediff = abs(end - start)
	ret = []

	ndates = abs((end - start) / minsize)
	#print('lr', ndates)
	for x, i in enumerate(drange(start, end, minsize)):
		#print(repr((x, i)))
		yield (i, x / ndates * wtarget)

def makeconsumption(s, mean, sig, interval):
	s = s.replace(hour=0, minute=0, second=0)
	e = (s + timedelta(days=1)) - interval

	ret = [ (i, rand.gauss(mean, sig)) for i in drange(s, e, interval) ]

	return ret

def distribute(s, e, prod, minsize):
	onehour = timedelta(hours=1)
	totaltime = e - s
	mid = s + totaltime / 2
	startrampend = s + onehour
	endrampstart = e - onehour

	# prod == wh
	wtarget = prod / ((totaltime + onehour).seconds / 60 / 60)

	ret = []

	#print('d:', repr((s, e)))
	ret.extend(linramp(s, startrampend, wtarget, minsize))

	for i in drange(startrampend, endrampstart, minsize):
		ret.append((i, wtarget))

	ret.extend(linramp(e, endrampstart, wtarget, minsize))

	ret.sort()
	#pprint.pprint(ret)

	return ret

#print('start')

prodpoints = []
prodindex = []
conspoints = []
consindex = []

def serializearrowasmili(obj):
	if not isinstance(obj, arrow.Arrow):
		raise TypeError

	return int(obj.float_timestamp*1000)

for i in arrow.Arrow.range('day', startdate, enddate):
	whprod = rand.gauss(meanwhprod, sigwhprod)
	s, e = makestartend(i)
	noon = i.replace(hour=12)

	prodindex.append((noon, whprod))
	#print(repr(i), whprod)

	dist = []

	tmppoints = makeconsumption(s, rand.uniform(400,800), 100, timedelta(minutes=1))
	consindex.append((noon, sum((i[1] / 60 for i in tmppoints), 0)))
	conspoints.extend(tmppoints)

	# zero points for non-production
	import sys
	dist.extend(distribute(i, s, 0, timedelta(minutes=5)))

	# production
	dist.extend(distribute(s, e, whprod, timedelta(seconds=20)))

	# zero points for non-production
	eod = i.replace(hour=23, minute=55)
	dist.extend(distribute(e, eod, 0, timedelta(minutes=5)))

	# print timestamps as miliseconds
	if False:
		dist = ((int(a.float_timestamp*1000), b) for a, b in dist)

	# print space delimited time, else json
	if False:
		print('\n'.join('%s %s' % (a, b) for a, b in dist))
	else:
		#print(json.dumps(tuple(dist), indent=2))
		prodpoints.extend(dist)

for i in { 'prodpoints', 'prodindex', 'conspoints', 'consindex',}:
	v = locals()[i]
	v = [ (x, int(y)) for x, y in v ]
	locals()[i] = v

#conspoints = [ (x, -y) for x, y in conspoints ]

print('fakedata =', json.dumps(dict(production=prodpoints, prodindex=prodindex, consumption=conspoints, consindex=consindex), default=serializearrowasmili, indent=1))