Simulation context and state variable saving#

Simulation objects act as wrappers for a model and one or multiple event, and thus sets a context for a simulation. It handles the following aspect:

  • The access to the different variables timeseries.

  • Which outputs to save.

  • The length in unit step of the simulation.

  • The list of events to simulate.

  • The looping process of the simulation.

Simulation length#

The length of the simulation can be set by the n_temporal_units_to_sim argument at initialisation.

See also

Focus on the temporal dimension

Running multiple simulations#

At the moment we recommend redefining both a model and a simulation object when running multiple simulations.

Reading the outputs and saving files#

Monitoring the model variables#

By default, a simulation records the evolution of variables in numpy arrays, which are accessible directly as attributes as long as the Simulation object exists. Optionally the records can also be set to memmaps instead, which are saved as files at the end of the simulation (note that these files are temporary by default).

These records contain the variables values for each regions for each sector for each temporal unit and can also be accessed as formatted DataFrames, where each row represents a temporal unit and the columns represent all the possible (region,sector) tuples, i.e., industries, ordered in lexicographic order.

Here is a commented list of the different variables accessible:

# The production actually realised at each step
sim.production_realised

# The production capacity
sim.production_capacity

# The share of realised production distributed to rebuilding
sim.rebuild_prod

# The overproduction factor
sim.overproduction

# The (total) intermediate demand (ie how much intermediate demand was addressed to sector i in region j)
sim.intermediate_demand

# The (total) final demand (note that the final demand is currently fix in the model)
sim.final_demand

# The (total) rebuild demand
sim.rebuild_demand

# The amount of final demand that couldn't be satisfied
sim.final_demand_unmet

# The remaining amount of destroyed (ie not recovered/rebuilt) capital
sim.productive_capital_to_recover

# Note that the following array have one more dimension,
# its shape is (temporal units, sectors, regions*sectors)
# This one states for each temporal unit, for each input, for each (region,sector)
# if the input was limiting production. For efficiency, information is stored as a
# byte, -1 for False, 1 for True
sim.limiting_inputs

It is also possible to record the inputs stocks, but this is disabled by default as its shape is the same as limiting_inputs, but its dtype is float64, which can very rapidly lead to huge arrays difficult to have in memory.

# Setup the recording of stocks
sim = simulation(model, register_stocks=True)

# Access the array
sim.inputs_stocks

Hint

These DataFrames can easily be saved using any of pandas writers. BoARIO also makes it possible to save the raw arrays (see below).

Saving indexes, parameters and events simulated#

In order to keep experiments organized and reproducible, the following arguments can be used when instantiating a Simulation object:

  • "save_index" : True|False, if True, saves a file boario_output_dir/results/jsons/indexes.json, where the indexes (regions, sectors, final demand categories, etc.) are stored.

  • "save_params" : True|False, if True, saves a file boario_output_dir/results/jsons/simulated_params.json, where the simulation parameters are stored.

  • "save_events" : True|False, if True, saves a file boario_output_dir/results/jsons/simulated_events.json, where the indexes (regions, sectors, final demand categories, etc.) are stored.

Record files#

By defaults the arrays recording the evolution of variables are temporary files, which are deleted when the Simulation object is destroyed.

It is however possible to ask the Simulation object to save any selection of these raw arrays, by giving a list and an output directory when instantiating. Here is the complete list of variables than can be saved:

['production_realised', 'production_capacity', 'final_demand', 'intermediate_demand', 'rebuild_demand', 'overproduction', 'final_demand_unmet', 'rebuild_prod', 'inputs_stocks', 'limiting_inputs', 'kapital_to_recover']

Attention

inputs_stocks still requires the argument register_stocks to be True in order for the file to be saved.

For example the following code will create the files "production_realised" and "final_demand_unmet" in the specified folder (or to a temporary directory prefixed by "boario" by default).

sim = Simulation(
    model,
    save_records=["production_realised", "final_demand_unmet"],
    boario_output_dir="folder of your choosing/",
)

Files saved like this are raw numpy arrays and can then be read with:

import numpy as np

# For all records except limiting_inputs and inputs_stocks
np.memmap(
    "path/to/file",
    mode="r+",
    dtype="float64",
    shape=(n_temporal_units, n_sectors * n_regions),
)

# For limiting_inputs
np.memmap(
    "path/to/file",
    mode="r+",
    dtype="byte",
    shape=(n_sectors * n_temporal_units, n_sectors * n_regions),
)

# For inputs_stocks
np.memmap(
    "path/to/file",
    mode="r+",
    dtype="float64",
    shape=(n_sectors * n_temporal_units, n_sectors * n_regions),
)
On this page
Show Source