02 — Time series and resampling¶

Concept: a TimeSeries is one value per step of a grid, tagged with what it measures (kW, kWh, €/kWh or a dimensionless shape). Profiles are held as average power, which does not depend on the step, and resampling between any two grids conserves energy.

You will learn: how quantities combine, why power and energy resample differently, and what happens when a grid reaches outside the data.

In [1]:

from datetime import UTC, datetime, timedelta

import numpy as np

from myenergy.domain.errors import CoverageError, SeriesError
from myenergy.domain.series import Quantity, TimeSeries
from myenergy.domain.time import TimeGrid

start = datetime(2026, 1, 1, tzinfo=UTC)
quarter = TimeGrid(start=start, end=start + timedelta(hours=2), step=timedelta(minutes=15))
load = TimeSeries(
    grid=quarter,
    values=np.array([0.4, 0.4, 0.5, 2.1, 2.3, 0.6, 0.5, 0.4]),
    quantity=Quantity.POWER_KW,
)
load.values, load.energy().total()

from datetime import UTC, datetime, timedelta import numpy as np from myenergy.domain.errors import CoverageError, SeriesError from myenergy.domain.series import Quantity, TimeSeries from myenergy.domain.time import TimeGrid start = datetime(2026, 1, 1, tzinfo=UTC) quarter = TimeGrid(start=start, end=start + timedelta(hours=2), step=timedelta(minutes=15)) load = TimeSeries( grid=quarter, values=np.array([0.4, 0.4, 0.5, 2.1, 2.3, 0.6, 0.5, 0.4]), quantity=Quantity.POWER_KW, ) load.values, load.energy().total()

Out[1]:

(array([0.4, 0.4, 0.5, 2.1, 2.3, 0.6, 0.5, 0.4]), 1.8)

Power (kW) times the step (h) gives energy (kWh): energy() does exactly that, and total() only exists for energy, because summing kW has no meaning.

In [2]:

try:
    load.total()
except SeriesError as error:
    print(error)

try: load.total() except SeriesError as error: print(error)

total() is for energy; use time_mean() for kW

Resampling¶

One overlap-weighted method serves every pair of steps. Power becomes a time-weighted mean; energy is summed or split in proportion to the overlap. Either way, the energy over the period is unchanged, even for steps that do not divide each other (15 → 20 minutes).

In [3]:

for minutes in (60, 20, 5):
    other = load.resample(quarter.with_step(timedelta(minutes=minutes)))
    print(
        f"{minutes:>2} min: {len(other):>2} steps, "
        f"{other.energy().total():.6f} kWh, mean {other.time_mean():.4f} kW"
    )

for minutes in (60, 20, 5): other = load.resample(quarter.with_step(timedelta(minutes=minutes))) print( f"{minutes:>2} min: {len(other):>2} steps, " f"{other.energy().total():.6f} kWh, mean {other.time_mean():.4f} kW" )

60 min:  2 steps, 1.800000 kWh, mean 0.9000 kW
20 min:  6 steps, 1.800000 kWh, mean 0.9000 kW
 5 min: 24 steps, 1.800000 kWh, mean 0.9000 kW

Going to a finer grid holds the power and splits the energy:

In [4]:

hourly = load.resample(quarter.with_step(timedelta(hours=1)))
hourly.values, hourly.resample(quarter).values

hourly = load.resample(quarter.with_step(timedelta(hours=1))) hourly.values, hourly.resample(quarter).values

Out[4]:

(array([0.85, 0.95]), array([0.85, 0.85, 0.85, 0.85, 0.95, 0.95, 0.95, 0.95]))

Outside the data: an error, never a guess¶

A grid that reaches beyond the series is refused with both ranges in the message (SPEC decision 9).

In [5]:

later = TimeGrid(
    start=start + timedelta(hours=1), end=start + timedelta(hours=3), step=timedelta(minutes=15)
)
try:
    load.resample(later)
except CoverageError as error:
    print(error)

later = TimeGrid( start=start + timedelta(hours=1), end=start + timedelta(hours=3), step=timedelta(minutes=15) ) try: load.resample(later) except CoverageError as error: print(error)

cannot resample to [2026-01-01T01:00:00+00:00, 2026-01-01T03:00:00+00:00): the series covers [2026-01-01T00:00:00+00:00, 2026-01-01T02:00:00+00:00)

Next: 03 — Load and PV profiles (with the next part of M1) turns Enedis shapes into kW series for a household and a solar array.