Moves a model's time-varying data from the calendar it was written against
onto calendar, so the model can be built and solved at another time
resolution. The time mirror of subset_model_regions() and
aggregate_model_regions().
Usage
recast_model_calendar(
object,
calendar,
rules = NULL,
year = NULL,
verbose = isVerbose()
)Arguments
- object
a model.
- calendar
the target calendar; coarser or finer than the model's own.
- rules
optional named character vector of timescales::CALENDAR_RULES keyed by value column, overriding the rate default for those columns (e.g.
c(demand = "sum")).- year
model year used to materialise both calendars on a shared datetime grid.
NULL(default) takes the model's first horizon milestone. It matters only where the calendars align to real dates (days of the year, weekdays), and then only across a leap year.- verbose
print per-table messages.
Value
the recast model, carrying calendar.
Details
Every timeslice-indexed value in energyRt is a rate: the equations multiply
it by the timeslice weight to obtain an annual quantity. Aggregating
therefore takes the weighted mean, which conserves the weighted annual
total exactly, and disaggregating copies the rate into each child timeslice.
rules overrides this per column where a value is not a rate (a declared
quantity to be summed, a standard deviation).
Rows are recast by their own timeframe, not by their table: only rows at the
source calendar's leaf timeframe move. A row declared at ANNUAL, or at any
timeframe calendar also has, is left exactly as it is.
A @timeframe naming a timeframe calendar does not have falls back to the
finest timeframe that survives, and is reported – otherwise the model
references a timeframe its calendar cannot resolve.
Comparability
A recast model is not a smaller answer to the same question. Collapsing timeslices removes the variation inside them, which is what a model prices: peaks flatten, variable resources become firmer than they are, and storage loses the cycle it was built to serve. Its dispatch, costs and objective are its own. Use it to run, plot and teach the shape of a solution, not to report the source model's numbers.
See also
Other interpolation:
interpolate_model(),
subset_model_regions(),
with_solver_log()
Examples
mod <- newModel("TOPIA", data = topia$modules$electricity$repo,
calendar = topia$modules$calendars$s4_h24,
region = topia$modules$electricity$regions,
horizon = topia$modules$horizons$BY2025,
discount = 0.05)
# 96 hourly timeslices collapse to 4 seasons: the same model, small enough
# to solve in a second
light <- recast_model_calendar(mod, calendar = calendars$s4)
#> recast onto calendar 's4' (96 -> 4 timeslices): 4 table(s), 8,448 -> 352 row(s); 6 timeframe declaration(s) -> SEASON (verbose = TRUE lists them).
#> Dispatch, costs and the objective are the recast model's own, not the source model's.
getCalendar(light)
#> An object of class "calendar"
#> Slot "name":
#> [1] "s4"
#>
#> Slot "desc":
#> [1] "SEASON calendar (4 timeslices; complete, regular) (timescales 0.6.0)"
#>
#> Slot "timeframes":
#> $ANNUAL
#> [1] "ANNUAL"
#>
#> $SEASON
#> [1] "WIN" "SPR" "SUM" "FAL"
#>
#>
#> Slot "year_fraction":
#> [1] 1
#>
#> Slot "year_start":
#> $month
#> [1] 1
#>
#> $day
#> [1] 1
#>
#>
#> Slot "utc_offset_minutes":
#> [1] 0
#>
#> Slot "timetable":
#> ANNUAL SEASON timeslice share weight
#> <char> <char> <char> <num> <num>
#> 1: ANNUAL WIN WIN 0.2465753 1
#> 2: ANNUAL SPR SPR 0.2520548 1
#> 3: ANNUAL SUM SUM 0.2520548 1
#> 4: ANNUAL FAL FAL 0.2493151 1
#>
#> Slot "timeslice_share":
#> timeslice share weight
#> <char> <num> <num>
#> 1: ANNUAL 1.0000000 1
#> 2: WIN 0.2465753 1
#> 3: SPR 0.2520548 1
#> 4: SUM 0.2520548 1
#> 5: FAL 0.2493151 1
#>
#> Slot "default_timeframe":
#> [1] "SEASON"
#>
#> Slot "timeframe_rank":
#> ANNUAL SEASON
#> 1 2
#>
#> Slot "timeslices_in_frame":
#> ANNUAL SEASON
#> 1 4
#>
#> Slot "timeslice_family":
#> parent child
#> <char> <char>
#> 1: ANNUAL WIN
#> 2: ANNUAL SPR
#> 3: ANNUAL SUM
#> 4: ANNUAL FAL
#>
#> Slot "timeslice_ancestry":
#> parent child
#> <char> <char>
#> 1: ANNUAL WIN
#> 2: ANNUAL SPR
#> 3: ANNUAL SUM
#> 4: ANNUAL FAL
#>
#> Slot "next_in_timeframe":
#> timeslice timeslicep
#> <char> <char>
#> 1: WIN SPR
#> 2: SPR SUM
#> 3: SUM FAL
#> 4: FAL WIN
#>
#> Slot "next_in_year":
#> timeslice timeslicep
#> <char> <char>
#> 1: WIN SPR
#> 2: SPR SUM
#> 3: SUM FAL
#> 4: FAL WIN
#>
#> Slot "misc":
#> $pTimesliceWeight
#> NULL
#>
#>
# the hourly demand profile, and the four seasonal rates it becomes
dem <- getObject(light, name = "DEM_ELC", drop = TRUE)
head(dem@demand)
#> cluster region year timeslice demand
#> 1 <NA> C1 2025 WIN 432.5774
#> 2 <NA> C1 2025 SPR 391.8829
#> 3 <NA> C1 2025 SUM 405.4693
#> 4 <NA> C1 2025 FAL 395.0703
#> 5 <NA> C1 2030 WIN 475.8351
#> 6 <NA> C1 2030 SPR 431.0712