Skip to content

Run from Python

Programmatic control over runs, from a single baseline to large ensembles of scenarios, sensitivity sweeps, and Monte Carlo analyses.

Prerequisites: Python environment set up. If not done yet, see Installation.

Quick start

Open a terminal from inside your EPM repository folder and run:

conda activate epm_env
python epm.py --folder_input data_test --config data_test/config.csv
  • --folder_input: your input data folder (inside epm/input/)
  • --config: the configuration file that tells EPM which CSV files to use

Results are written to output/simulations_run_<timestamp>/.

Common workflows

Run EPM with your own input folder:

python epm.py --folder_input my_country --config my_country/config.csv

Use --simple to relax integer constraints (faster, good for first tests):

python epm.py --folder_input my_country --config my_country/config.csv --simple

Run all scenarios defined in a scenarios.csv file, using 4 CPU cores in parallel:

python epm.py --folder_input my_country --config my_country/config.csv --scenarios --cpu 4

Run only a subset of scenarios:

python epm.py --folder_input my_country --config my_country/config.csv --scenarios -S baseline HighDemand

Run 50 Monte Carlo samples using an uncertainties definition file:

python epm.py --folder_input my_country --config my_country/config.csv \
  --montecarlo --montecarlo_samples 50 \
  --uncertainties my_country/uncertainties.csv \
  --cpu 8 --reduced_output

Windows prerequisite

Monte Carlo requires chaospy. On Windows, install C++ Build Tools first, then pip install chaospy==4.3.18.

Re-run postprocessing on an existing results folder without re-solving the model:

python epm.py --postprocess simulations_run_2025-05-01_12-00-00

Assess the value of a specific project by running a counterfactual (model without that project):

python epm.py --folder_input my_country --config my_country/config.csv \
  --generator_assessment SolarProject

Assess the value of an interconnection:

python epm.py --folder_input my_country --config my_country/config.csv \
  --interco_assessment Angola-Zambia

All options

The tables below list every argument you can pass to python epm.py. Arguments are optional unless stated otherwise (the model uses default values for anything not specified).

Core inputs

Argument Default Description
--folder_input data_test Input data folder (inside epm/input/)
--config config.csv Configuration file path, relative to folder_input
--modeltype None Override solver type: MIP or RMIP

Scenarios

Argument Default Description
--scenarios None Scenario file. Use without value to load scenarios.csv, or provide a custom path
-S / --selected_scenarios None Run only a subset of scenarios. Example: -S baseline HighDemand
--simple None Relax integer constraints. Use without arguments for default ['DiscreteCap', 'y']

Performance

Argument Default Description
--cpu 1 Number of CPU cores for parallel scenario runs
--reduced_output False Generate smaller output files (recommended for large Monte Carlo runs)
--output_zip False Zip the output folder after processing

Monte Carlo & sensitivity

Argument Default Description
--montecarlo False Enable Monte Carlo simulation
--montecarlo_samples 10 Number of random samples
--uncertainties None CSV file defining uncertain parameters and their distributions
--sensitivity False Enable built-in sensitivity analysis

Assessment

Argument Default Description
--generator_assessment None Exclude named generators to assess their counterfactual value
--project_assessment None Create individual and combined assessment scenarios per project
--interco_assessment None Assess the value of a named interconnection

Regional & country filtering

Argument Default Description
--country None Filter zones to include only those belonging to this country
--focus_country None After a regional run, generate single-country inputs for this country

Postprocessing & output

Argument Default Description
--postprocess None Run postprocessing only on an existing results folder
--generate_figures False Generate figures during postprocessing
--no_plot_dispatch Disable dispatch plots (saves time on large runs)
--plot_selected_scenarios all Restrict plots to a subset of scenarios
--graphs_folder img Folder where postprocessing plots are saved
--reduce_definition_csv False Generate reduced yearly CSV files for Tableau
--simulation_label timestamp Custom name for the results folder

Debugging

Argument Default Description
--debug False Enable verbose DEBUG mode in GAMS
--trace False Enable TRACE mode in GAMS
  • Input Setup: how config.csv, pSettings, and scenarios work together
  • Input Catalog: full reference of all input parameters
  • Debugging: what to do when the model does not converge