,
Ricarda Rosemann [aut] | Title: | Building sector model with heterogeneous renovation and construction of the stock |
|---|---|
| Description: | This building stock model represents residential and commercial buildings at customisable regional and temporal resolution. The building stock is quantified in floor area and distinguished by building type (SFH/MFH) and location (rural/urban). In each building category, construction cohorts are tracked explicitly. This allows to characterise buildings specifically for each subset of buildings. The evolution of the building stock follows from the flows of constructed, renovated and demolished buildings and is optimised under cost minimisation with a benefit for heterogeneity in the choice of construction and renovation alternatives. This benefit captures heterogeneity in the preferences of the agents and the building structure. |
| Authors: | Robin Hasse [aut, cre] ,
Ricarda Rosemann [aut] |
| Maintainer: | Robin Hasse <[email protected]> |
| License: | LGPL-3 |
| Version: | 0.6.2 |
| Built: | 2025-03-20 06:33:23 UTC |
| Source: | https://github.com/pik-piam/brick |
This building stock model represents residential and commercial buildings at customisable regional and temporal resolution. The building stock is quantified in floor area and distinguished by building type (SFH/MFH) and location (rural/urban). In each building category, construction cohorts are tracked explicitly. This allows to characterise buildings specifically for each subset of buildings. The evolution of the building stock follows from the flows of constructed, renovated and demolished buildings and is optimised under cost minimisation with a benefit for heterogeneity in the choice of construction and renovation alternatives. This benefit captures heterogeneity in the preferences of the agents and the building structure.
Maintainer: Robin Hasse [email protected] (ORCID)
Authors:
Ricarda Rosemann [email protected] (ORCID)
Useful links:
Write the specific costs to the input.gdx
.addSpecCostToInput( m, path, optimVarCon, optimVarRen, xinitCon, xinitRen, tcalib, varName = "x" ).addSpecCostToInput( m, path, optimVarCon, optimVarRen, xinitCon, xinitRen, tcalib, varName = "x" )
m |
Gams transfer container with previous input data |
path |
character, path to this run |
optimVarCon |
data frame with optimization variables of construction |
optimVarRen |
data frame with optimization variables of renovation |
xinitCon |
data frame with initial specific intangible costs of construction |
xinitRen |
data frame with initial specific intangible costs of renovation |
tcalib |
numeric, calibration time steps |
varName |
character, optimization variable to calculate specific costs from. Needs to be a column of optimVarCon and optimVarRen, should be one of 'x', 'xA' or 'xMin'. |
Add the calibration targets to the input gdx
.addTargetsToInput(mInput, path, calibTarget).addTargetsToInput(mInput, path, calibTarget)
mInput |
gamstransfer container of the input gdx |
path |
character, path to output folder of this run |
calibTarget |
list of data frames of calibration targets |
Check if the Armijo condition holds. Return only data combinations for which it does not hold.
.checkArmijoStep( prevStep, stepSizeParams, outerObjective, sensitivityArmijo, varName = "fA", factorMin = 1 ).checkArmijoStep( prevStep, stepSizeParams, outerObjective, sensitivityArmijo, varName = "fA", factorMin = 1 )
prevStep |
data frame with data combinations that did not satisfy the Armijo condition in the previous step |
stepSizeParams |
data frame with step size and related parameters |
outerObjective |
data frame containing the value of the outer objective function.
Needs to contain the column |
sensitivityArmijo |
numeric, parameter of the Armijo condition specifying how strict the condition is |
varName |
character, outer objective variable to test the Armijo condition on |
factorMin |
numeric, additional factor to modify the Armijo condition |
Check if the condition of the heuristic step size adaptation holds: Check if the outer objective does not increase. Return only data combinations for which the condition does not hold, i.e. the outer objective increases
.checkHeuristicStep(prevStep, outerObjective).checkHeuristicStep(prevStep, outerObjective)
prevStep |
data frame with data combinations that did not satisfy the heuristic condition in the previous step |
outerObjective |
data frame containing the value of the outer objective function.
Needs to contain the columns |
Read in brick results and compute the total outer objective function by combining resuls from construction and renovation.
.combineOuterObjective( m, outerObjective, p_constructionCalibTarget, p_renovationCalibTarget, tcalib, varName = "f" ).combineOuterObjective( m, outerObjective, p_constructionCalibTarget, p_renovationCalibTarget, tcalib, varName = "f" )
m |
Gams transfer container to read brick results from |
outerObjective |
data frame to write the outer objective to |
p_constructionCalibTarget |
data frame with historic construction flows |
p_renovationCalibTarget |
data frame with historic renovation flows |
tcalib |
numeric, calibration time steps |
varName |
character, column name in |
Combine the step size paramters delta and phi-derivative from construction and renovation flows
.combineStepSizeParams(paramsCon, paramsRen).combineStepSizeParams(paramsCon, paramsRen)
paramsCon |
data frame with step size parameters from construction |
paramsRen |
data frame with step size parameters from renovation |
Read in objective function values for slightly shifted optimization variables
and compute the gradient of the objective function.
Set the adjustment term d to the negative of the gradient.
.computeDescentDirection( m, dims, tcalib, flow = c("construction", "renovation") ).computeDescentDirection( m, dims, tcalib, flow = c("construction", "renovation") )
m |
Gams transfer container with current Brick results |
dims |
character, dimensions of the desired result object |
tcalib |
numeric, time steps to calibrate on |
flow |
character, either 'construction' or 'renovation' |
d
Compute the deviation between historic data and Brick results.
Then compute the adjustment term for the calibration d
.computeDeviation( m, target, dims, tcalib, flow = c("construction", "renovation"), renAllowed = NULL ).computeDeviation( m, target, dims, tcalib, flow = c("construction", "renovation"), renAllowed = NULL )
m |
Gams transfer container with current Brick results |
target |
data frame of historical data |
dims |
character, dimensions of historic data and Brick results |
tcalib |
numeric, time steps to calibrate on |
flow |
character, either 'construction' or 'renovation' |
renAllowed |
data frame with allowed renovation transitions |
In case of the logit calibration, this outer objective function is virtual: We are not actually performing an optimization algorithm on this function, but only use it for the step size adaptation.
.computeOuterObjective(m, gamsVar, target, dims, tcalib).computeOuterObjective(m, gamsVar, target, dims, tcalib)
m |
Gams transfer container object to read set values from |
gamsVar |
data frame with brick results for the given variable |
target |
data frame with historic data |
dims |
character, dimensions of data |
tcalib |
numeric, calibration time steps |
Evaluate the outer objective function for the stock
.computeOuterObjectiveStock( m, outerObjective, p_stockCalibTarget, tcalib, varName = "f" ).computeOuterObjectiveStock( m, outerObjective, p_stockCalibTarget, tcalib, varName = "f" )
m |
Gams transfer object to read the stock variable from |
outerObjective |
data frame to write the outer objective to |
p_stockCalibTarget |
data frame with historic stock |
tcalib |
numeric, calibration time steps |
varName |
character, column name in |
Compute the step size adaptation paramters delta and phi-derivative
.computeStepSizeParams(deviation).computeStepSizeParams(deviation)
deviation |
data frame with deviation and adjustment term 'd' |
remove all records of temporal parameters that are outside of thist
and make zero values explicit by filling with EPS.
.cropParamsToThist(gdx, thist).cropParamsToThist(gdx, thist)
gdx |
character, file path to gdx |
thist |
numeric vector of historic periods |
Assemble specific costs from initial specific costs and the optimization variable
.determineSpecCost(optimVar, xinit, dims, tcalib, varName = "x").determineSpecCost(optimVar, xinit, dims, tcalib, varName = "x")
optimVar |
data frame with the optimization variable 'x' and optionally 'xMin' and 'xA' |
xinit |
data frame with initial specific intangible costs |
dims |
character, dimensions of the optimization variable |
tcalib |
numeric, calibration time steps |
varName |
character, optimization variable to calculate specific costs from. Should be one of 'x', 'xA' or 'xMin'. |
Replace zeros in value column with the gams special value EPS. This way, the value is saved in gdx files by gams.
.explicitZero(x, value = "value").explicitZero(x, value = "value")
x |
data.frame |
value |
character, name of value column |
data.frame with explicit zeros in value column
Robin Hasse
Performs successive semi_joins with passed data frames by all common columns. Useful to filter allowed renovations or vintages that exist.
.filter(x, ...).filter(x, ...)
x |
data.frame |
... |
data frames that only have columns existing also in |
filtered data frame
Robin Hasse
used to select a specifc level or scenario from a data frame with alternative values.
.filterLevel(df, lvl, switchName = "", lvlCol = "level").filterLevel(df, lvl, switchName = "", lvlCol = "level")
df |
data.frame |
lvl |
value used to select rows |
switchName |
character, corresponding BRICK switch name (only used for more informative error message) |
lvlCol |
character, colname containing |
data.frame with selected rows without lvlCol
Search for config file in multiple steps. The file is found if
* config is a full file path already
* config is the name of a file in the config folder
* there is exactly one file in configFolder matching the pattern
passed via config
.findCfg(config, configFolder, isFinalCfg).findCfg(config, configFolder, isFinalCfg)
config |
character, config file, either a path to a yaml file or the name of the file in 'inst/config/' |
configFolder |
character, directory to search for configs. If NULL, the BRICK-internal config folder is used. |
isFinalCfg |
logical, is this the final config and not an intermediate? |
file path to config
Depending on what is passed via originGDX, this function returns the
file path if it exists, looks for recognised file names in the given
directory if it exists or looks for the latest run with the given name in the
outputFolder.
.findOriginGdxFile(originGdx, outputFolder).findOriginGdxFile(originGdx, outputFolder)
originGdx |
character, file path to run or gdx file used as historical or scenario name |
outputFolder |
directory of output folder, only required if
|
file path to origin gdx file
Gives the column names of a list of data frames except the first and the last
one, i.e. qty and value.
.getDims(calibTarget).getDims(calibTarget)
calibTarget |
list, calibration targets |
Initialize the data frame for the optimization variable of construction
.initOptimVarCon(mInput, tcalib).initOptimVarCon(mInput, tcalib)
mInput |
Gamstransfer container with the input data |
tcalib |
numeric, calibration time periods |
Initialize the data frame for the optimization variable of renovation
.initOptimVarRen(mInput, tcalib, renAllowed).initOptimVarRen(mInput, tcalib, renAllowed)
mInput |
Gamstransfer container with the input data |
tcalib |
numeric, calibration time periods |
renAllowed |
data frame with allowed renovation transitions |
Initialize the data frame for the objective function value
.initOuterObjective(mInput, tcalib).initOuterObjective(mInput, tcalib)
mInput |
Gamstransfer container with the input data |
tcalib |
numeric, calibration time periods |
Compute the initial step size of the step size iteration algorithm
.initStepSize(i, stepSizeParams, outerObjective, stepSizeInit).initStepSize(i, stepSizeParams, outerObjective, stepSizeInit)
i |
numeric, iteration number of overall calibration procedure |
stepSizeParams |
data frame with parameters for the step size adaptation |
outerObjective |
data frame with previous and current values of outer objective function;
needs to contain the columns |
stepSizeInit |
numeric, lower bound for the initial step size |
Calculate log-Likelihood of historical values
.logLikelihood(res, target).logLikelihood(res, target)
res |
numeric, result values |
target |
numeric, historic values |
Just a check to get a helpful error if gamstransfer is missing. Once the
package is available on GitHub, it should be listed as a dependency and this
file can be removed. The check is disabled during contiunous integration (CI)
i.a. GitHub actions and during devtools:check which is called by
lucode2::buildLibrary.
.onLoad(libname, pkgname).onLoad(libname, pkgname)
libname |
not used |
pkgname |
not used |
Robin Hasse
Overwrite a named list with another named list. The result corresponds to the overwritten list unless a value is overwritten by the overwriting list.
.overwriteList(x, y, isFinalCfg, defaultCfgPath).overwriteList(x, y, isFinalCfg, defaultCfgPath)
x |
named list, provides the structure and default values that can be
overwritten by |
y |
named list, overwrites |
isFinalCfg |
logical, is this the final config and not an intermediate? |
defaultCfgPath |
character, path to default config. Only used for more helpful error message. |
This function is called recursively until the default config is reached. Therefor, the list structure will always correspond to the default config. No config based directly or indirectly on the default can have list keys that the default config doesn't have.
named list with the structure of x that is (partly)
overwritten by y.
Read calibration targets from input folder
.readCalibTarget().readCalibTarget()
Read yaml file from given path, check minimum requirement (title exists) and save the file path as a attribute.
.readCfg(file).readCfg(file)
file |
character, path to config file |
named list with config parameters
Only applicable for optimization calibration.
.readOuterObjectiveOptim(m, outerObjective, varName = "f").readOuterObjectiveOptim(m, outerObjective, varName = "f")
m |
Gams transfer container to read brick results from |
outerObjective |
data frame to write the outer objective to |
varName |
character, column name in |
Calculate sum of squared differences between results and historic values
.sumSquare(res, target).sumSquare(res, target)
res |
numeric, result values |
target |
numeric, historic values |
Update the step size for the selected combinations
.updateStepSize(totalStep, stepSizeParams, stepReduction).updateStepSize(totalStep, stepSizeParams, stepReduction)
totalStep |
data frame with combinations to perform the adjustment ot the step size on |
stepSizeParams |
data frame with the parameters of the step size adjustment algorithm |
stepReduction |
numeric, factor applied to stepSize to reduce the step size. Should be < 1. |
Adjust 'x' by adding the adjustment term 'd' multiplied by the step size
.updateX( optimVar, deviation, stepSizeParams, dims, nameTo = "x", factorMin = 1 ).updateX( optimVar, deviation, stepSizeParams, dims, nameTo = "x", factorMin = 1 )
optimVar |
data frame with the optimization variable 'x' and optionally 'xMin' and 'xA' |
deviation |
data frame with deviation and adjustment term 'd' |
stepSizeParams |
data frame with parameters of step size adaptation, including step size 'stepSize' |
dims |
character, dimensions of the optimization variable |
nameTo |
character, optimization variable to write the result to |
factorMin |
numeric, additional factor to multiply the adjustment term with |
Update the optimization variable 'x' for selected combinations only
.updateXSelect( totalStep, optimVar, deviation, stepSizeParams, dims, nameTo = "x" ).updateXSelect( totalStep, optimVar, deviation, stepSizeParams, dims, nameTo = "x" )
totalStep |
data frame with combinations to perform the adjustment of 'x' on |
optimVar |
data frame with the optimization variables |
deviation |
data frame with the deviation from historic data and the adjustment parameter 'd' |
stepSizeParams |
data frame with the parameters of the step size adaptatation procedure |
dims |
character, dimensions of the optimization variable |
nameTo |
character, column name to write the updated optimization variable to. Should be one of 'x', 'xA', 'xMin'. |
Write the intangible costs to a .csv file
.writeCostIntang(file, optimVar, xinit, dims, tcalib).writeCostIntang(file, optimVar, xinit, dims, tcalib)
file |
character, path to file to which results should be written |
optimVar |
data frame with optimization variable, needs to contain column 'x' |
xinit |
data frame with initial intangible specific costs |
dims |
character, dimensions of the optimization variable |
tcalib |
numeric, calibration time steps |
Add assumed intangible costs
addAssump(df, assumpFile)addAssump(df, assumpFile)
df |
data.frame for the cost of construction or renovation |
assumpFile |
character, file path to assumption file |
data frame with added intangible cost
Robin Hasse
Temporary function that prepares historic stock and flows for calibration
aggregateMatching(path, config, overwrite = FALSE)aggregateMatching(path, config, overwrite = FALSE)
path |
character, path to calibration run |
config |
named list, configuration of calibration run |
overwrite |
logical, should existing data be overwritten? |
This functionality will migrate to mredgebuidlings but this infrastructure requires more time to be developed.
Robin Hasse
This is meant to work with the installed package BRICK but also when loading
the package via devtools::load_all("path/to/brick")
brick.file(..., mustWork = TRUE)brick.file(..., mustWork = TRUE)
... |
character vectors, specifying subdirectory and files within brick |
mustWork |
if TRUE, an error is given if there are no matching files |
A character vector of positive length, containing the file paths
that matched ... in BRICK.
Robin Hasse
Check which output file was written and derive state of Gams run
checkGamsSuccess(path)checkGamsSuccess(path)
path |
character, path to search for gams output files |
If output.gdx was written, give success message.
If abort.gdx was written or no output file exists, stop with error message.
Ricarda Rosemann
Copy gams scripts to output folder
copyGamsFiles(path, overwrite = FALSE)copyGamsFiles(path, overwrite = FALSE)
path |
character vector with folders to write input data into |
overwrite |
logical, should existing input.gdx be overwritten? |
Robin Hasse
Copy history gdx to output folder
copyHistoryGdx( path, outputFolder = NULL, config, overwrite = FALSE, thistOnly = TRUE )copyHistoryGdx( path, outputFolder = NULL, config, overwrite = FALSE, thistOnly = TRUE )
path |
character vector with folders to write input data into |
outputFolder |
directory of output folder |
config |
named list with run configuration |
overwrite |
logical, should existing input.gdx be overwritten? |
thistOnly |
logical, crop temporal parameters to historic periods |
Robin Hasse
Copy initial gdx to output folder
copyInitialGdx(path, config, overwrite = FALSE)copyInitialGdx(path, config, overwrite = FALSE)
path |
character vector with folders to write input data into |
config |
named list with run configuration |
overwrite |
logical, should existing input.gdx be overwritten? |
Robin Hasse
Create a complete set of input data for the gams optimisation.
createInputData(path, config, overwrite = FALSE)createInputData(path, config, overwrite = FALSE)
path |
character vector with folders to write input data into |
config |
named list with run configuration |
overwrite |
logical, should existing input.gdx be overwritten? |
This function reads static input data according to the input data revision in the config and creates all required sets and parameters for the gams optimisation depending on the switches in the config.
Robin Hasse
Create data for reference matching
createMatchingData(path, config, overwrite = FALSE)createMatchingData(path, config, overwrite = FALSE)
path |
character vector with folders to run the model in |
config |
run configurations |
overwrite |
logical, should existing data be overwritten? |
Robin Hasse
Add all parameters to gams container based on config
createParameters(m, config, inputDir)createParameters(m, config, inputDir)
m |
gams Container, central object to store all data for input.gdx |
config |
named list with run configuration |
inputDir |
directory of input folder |
gams Container with parameters added
Robin Hasse
Create a folder for the model to run in and copy required gams files there.
createRunFolder( path, config = NULL, overwrite = FALSE, recursive = FALSE, showWarnings = TRUE )createRunFolder( path, config = NULL, overwrite = FALSE, recursive = FALSE, showWarnings = TRUE )
path |
character vector, containing |
config |
list with run configuration |
overwrite |
logical; Should exiting folders be overwritten? |
recursive |
logical; Should exiting folders be overwritten? |
showWarnings |
logical; Should exiting folders be overwritten? |
Robin Hasse
Add all sets to gams container based on config
createSets(m, config)createSets(m, config)
m |
gams Container, central object to store all data for input.gdx |
config |
named list with run configuration |
gams Container with sets added
Robin Hasse
Create a data frame that has a column for each set and the full crossing of all set entries as rows
expandSets(..., .m = NULL)expandSets(..., .m = NULL)
... |
gams sets |
.m |
gams Container with sets referenced in |
data.frame with full crossing of set entries
Robin Hasse
Search the given output folder for the run with the most recent time stamp and return the path to this run. If the output folder contains only one run, return its path, also if it does not contain a time stamp.
findLastRun(outputFolder)findLastRun(outputFolder)
outputFolder |
character, output folder to search in |
Ricarda Rosemann
Retrieve mapping file from BRICK
getBrickMapping( name, type = "sectoral", error.missing = TRUE, returnPathOnly = FALSE )getBrickMapping( name, type = "sectoral", error.missing = TRUE, returnPathOnly = FALSE )
name |
character, file name of the mapping file |
type |
character, Mapping type (e.g. "regional", or "sectoral") |
error.missing |
logical, return error if the file does not exist |
returnPathOnly |
logical, file name of the mapping file |
Robin Hasse
guess column names based on column values
guessColnames(x, m)guessColnames(x, m)
x |
data.frame with unknown column names |
m |
gams Conatiner with sets |
data.frame with guessed column names
Robin Hasse
Preparations of a model run, send the model to SLURM if desired
initModel( config = NULL, path = NULL, configFolder = NULL, outputFolder = "output", references = NULL, restart = FALSE, runReporting = TRUE, sendToSlurm = NULL, slurmQOS = NULL, tasksPerNode = NULL, timeLimit = NULL, tasks32 = FALSE )initModel( config = NULL, path = NULL, configFolder = NULL, outputFolder = "output", references = NULL, restart = FALSE, runReporting = TRUE, sendToSlurm = NULL, slurmQOS = NULL, tasksPerNode = NULL, timeLimit = NULL, tasks32 = FALSE )
config |
run configurations |
path |
character vector with folders to run the model in |
configFolder |
character, directory to search for configs. If NULL, the BRICK-internal config folder is used. |
outputFolder |
directory of output folder |
references |
named character vector of matching references |
restart |
logical or character vector of elements to be restarted. If FALSE (default), then no restart is initiated. If TRUE, then the run in the given path or the latest run is restarted with default settings. Allowed elements of the character vector are:
|
runReporting |
logical, whether to run the reporting, i.e. write the mif |
sendToSlurm |
boolean whether or not the run should be started via SLURM |
slurmQOS |
character, slurm QOS to be used |
tasksPerNode |
numeric, number of tasks per node to be requested |
timeLimit |
character, time limit of the slurm job given in the format hh:mm:ss |
tasks32 |
boolean whether or not the SLURM run should be with 32 tasks |
This function creates the run folder with the necessary config and gams files. It then either calls the function to start the model directly or passes the model to SLURM.
path (invisible)
Ricarda Rosemann
Checks whether slurm is available so that jobs can be submitted, e.g., via sbatch ....
isSlurmAvailable()isSlurmAvailable()
logical(1). Whether slurm is available.
Convert nested named list to long data.frame
listToDf(x, n = 1)listToDf(x, n = 1)
x |
named list with identical depth on each branch |
n |
Integer, number of nested function calls, leave default |
data.frame with column for each level of the named list
Robin Hasse
Load Input data from mredgebuildings
loadMadratData(config)loadMadratData(config)
config |
named list with run configuration |
directory of input folder
Robin Hasse
Helper function to turn a named list into a string appended to the gams command line call
makeHandle(lst, type = c("gams", "model"))makeHandle(lst, type = c("gams", "model"))
lst |
named list of flags |
type |
character, type of flags (either '"gams"' or '"model') |
Gams parameters are lower case flags and follow one minus '-'. Model switches are upper case flags and follow two minuses '–'.
character of flags
Robin Hasse
extract different periods from a given BRICK config
periodFromConfig(config, periodType)periodFromConfig(config, periodType)
config |
named list with run configuration |
periodType |
type of period(s) |
numeric vector with periods
Robin Hasse
Plot heat map of reference deviation
plotRefDeviation(path)plotRefDeviation(path)
path |
character; directory of output folder |
Robin Hasse
Plot an overview of the stock and flows
plotSummary(path, facet = "typ", showHistStock = FALSE, splitRen = FALSE)plotSummary(path, facet = "typ", showHistStock = FALSE, splitRen = FALSE)
path |
character, path to the run |
facet |
character, dimension to resolve as facets |
showHistStock |
logical, show given historic next to the modeled stock |
splitRen |
logical, plot renovation with identical replacement semi-transparent |
Robin Hasse
Read config file in yaml format
readConfig(config = NULL, configFolder = NULL, readDirect = FALSE)readConfig(config = NULL, configFolder = NULL, readDirect = FALSE)
config |
character, config file, either a path to a yaml file or the
name of the file in |
configFolder |
character, directory to search for configs. If NULL, the BRICK-internal config folder is used. |
readDirect |
logical, specify whether |
If no argument is given, the default config is used.
named list with run config
Robin Hasse
Read madrat input files from input folder
readInput(filename, dims, inputDir = NULL)readInput(filename, dims, inputDir = NULL)
filename |
character, filename |
dims |
character vector, column names |
inputDir |
character, directory of input folder |
data.frame with data of the given input file
Robin Hasse
Read symbol from gams container
readSymbol(x, symbol = NULL, selectArea = TRUE, stringAsFactor = TRUE)readSymbol(x, symbol = NULL, selectArea = TRUE, stringAsFactor = TRUE)
x |
gams Container, Parameter, Variable or Set |
symbol |
character, name of gams object if x is a Container else NULL |
selectArea |
logical, select area quantity and remove this dimension |
stringAsFactor |
logical, keep default factors from gams |
Robin Hasse
The mif file contains reporting variables for the given model run.
reportMif(path, file = NULL, tmpl = NULL)reportMif(path, file = NULL, tmpl = NULL)
path |
character, path to the run |
file |
character, path of mif file. If |
tmpl |
character, BRICK reporting template. There has to be a brickSets
mapping named with the same suffix: |
Robin Hasse
Two methods for the Brick calibration are available: The logit calibration and the optimization calibration.
runCalibration( path, parameters, tcalib, gamsOptions = NULL, switches = NULL, fileName = "main.gms", gamsCall = "gams" )runCalibration( path, parameters, tcalib, gamsOptions = NULL, switches = NULL, fileName = "main.gms", gamsCall = "gams" )
path |
character vector with folders to run gams in |
parameters |
named list of calibration parameters |
tcalib |
numeric, time periods to calibrate on |
gamsOptions |
named list of GAMS options |
switches |
named list of model switches |
fileName |
character vector with gams file names |
gamsCall |
system command to call gams |
The general procedure is as follows:
We aim to minimize the deviation between historic values and Brick results.
An intangible component of the specific construction and renovation costs is iteratively
adjusted to achieve this. Thus the intangible costs serve as the optimization variable x
and a functional evaluation of the deviation represents the objective function f.
Brick is run iteratively and from each run, an adjustment term d is computed.
The optimization variable is adjusted according to x = x + stepSize * d, where
the step size is determined according to the Armijo step size adaptation algorithm.
The implementation is centered around the following objects:
The optimization variable x and its variant(s) required in the step size adaptation
are gathered in the data frame optimVarCon for construction and optimVarRen
for renovation.
The objective function value f and its variant(s) are gathered in the data
frame outerObjective
The adjustment term d and related variables if applicable are gathered in the
data frame deviationCon for construction and deviationRen for renovation.
The step size stepSize and related parameters required for the step size adaptation
algorithm are gathered in the data frame stepSizeParams.
Ricarda Rosemann
The adjustment term is computed as if Brick were a pure logit model:
d = log(<brick result>/<historic value>).
runCalibrationLogit( path, parameters, tcalib, calibTarget, gamsOptions = NULL, switches = NULL, fileName = "main.gms", gamsCall = "gams" )runCalibrationLogit( path, parameters, tcalib, calibTarget, gamsOptions = NULL, switches = NULL, fileName = "main.gms", gamsCall = "gams" )
path |
character vector with folders to run gams in |
parameters |
named list of calibration parameters |
tcalib |
numeric, time periods to calibrate on |
calibTarget |
list of data frames of calibration targets |
gamsOptions |
named list of GAMS options |
switches |
named list of model switches |
fileName |
character vector with gams file names |
gamsCall |
system command to call gams |
Here the armijo step size algorithm may not be applicable; if this is the case, a heuristic step size adaptation is applied.
This implements a gradient descent method to perform the minimization of the deviation. The adjustment term is computed as the negative of the gradient of the objective function.
runCalibrationOptim( path, parameters, tcalib, calibTarget, gamsOptions = NULL, switches = NULL, fileName = "main.gms", gamsCall = "gams" )runCalibrationOptim( path, parameters, tcalib, calibTarget, gamsOptions = NULL, switches = NULL, fileName = "main.gms", gamsCall = "gams" )
path |
character vector with folders to run gams in |
parameters |
named list of calibration parameters |
tcalib |
numeric, time periods to calibrate on |
calibTarget |
list of data frames of calibration targets |
gamsOptions |
named list of GAMS options |
switches |
named list of model switches |
fileName |
character vector with gams file names |
gamsCall |
system command to call gams |
Run the gams model in a given directory.
runGams( path, gamsOptions = NULL, switches = NULL, fileName = "main.gms", gamsCall = "gams" )runGams( path, gamsOptions = NULL, switches = NULL, fileName = "main.gms", gamsCall = "gams" )
path |
character vector with folders to run gams in |
gamsOptions |
named list of GAMS options |
switches |
named list of model switches |
fileName |
character vector with gams file names |
gamsCall |
system command to call gams |
The file 'input.gdx' has to exist in the given directory.
Robin Hasse
Based on user input, construct the SLURM configuration string.
setSlurmConfig(slurmQOS, tasksPerNode = 16, tasks32 = FALSE, timeLimit = NULL)setSlurmConfig(slurmQOS, tasksPerNode = 16, tasks32 = FALSE, timeLimit = NULL)
slurmQOS |
string, name of the desired QOS (Quality of Service) |
tasksPerNode |
numeric, number of tasks per node to be requested |
tasks32 |
boolean, specify whether a node with 32 tasks should be requested |
timeLimit |
character, time limit of the slurm job given in the format hh:mm:ss |
Check if SLURM configuration is admissible.
string with SLURM configuration
Ricarda Rosemann
Run the model with given configuration.
startModel(path, runReporting = TRUE)startModel(path, runReporting = TRUE)
path |
character vector with folders to run the model in |
runReporting |
logical, whether to run the reporting, i.e. write the mif |
This function creates a run folder with necessary gams files if missing. It then computes the input data and finally runs the optimisation.
Robin Hasse
Missing periods are interpolated linearly and extrapolated constantly, additional periods are removed and all other dimensions are filtered to the elements dined in the model sets.
toModelResolution(x, m, value = "value")toModelResolution(x, m, value = "value")
x |
data.frame with temporal dimension |
m |
gams Container with sets as known dimensions |
value |
character, name of value column |
data.frame with temporal resolution according to model
Robin Hasse